Data / MCP / Skills

slug · title · url · videoId · event · durationMs · speakers · topics · summary · slideText · slides

[
  {
    "slug": "agent-skills",
    "title": "Don't Build Agents, Build Skills Instead – Barry Zhang & Mahesh Murag, Anthropic",
    "url": "https://ai.engineer/talks/CEvIs9y1uog-agent-skills",
    "videoId": "CEvIs9y1uog",
    "event": "AI Engineer Code 2025",
    "durationMs": 982000,
    "speakers": [
      {
        "name": "Barry Zhang"
      },
      {
        "name": "Mahesh Murag"
      }
    ],
    "topics": [
      {
        "slug": "agent-engineering",
        "name": "Agent engineering"
      },
      {
        "slug": "ai-engineering-leadership",
        "name": "AI engineering leadership"
      },
      {
        "slug": "rag-and-knowledge",
        "name": "Retrieval-augmented generation"
      },
      {
        "slug": "model-context-protocol",
        "name": "MCP"
      },
      {
        "slug": "coding-agents",
        "name": "Coding agents"
      }
    ],
    "summary": "Barry Zhang and Mahesh Murag of Anthropic argue that instead of building domain-specific agents, developers should build reusable Skills—organized folders of files that package procedural knowledge for agents. They explain that skills are progressively disclosed to protect the context window, use scripts as self-documenting tools, and have already grown to thousands in five weeks, including foundational, partner, and enterprise skills. Skills complement MCP servers by providing expertise while MCP handles connectivity. The future includes treating skills like software with testing and versioning, and enabling agents to create their own skills for continuous learning, ultimately creating a collective knowledge base that makes agents more capable and reliable.",
    "slideText": "Year of agents\nYear of agents\n\nTimeline: FEB — TODAY\n\nANTHROPIC\nHow We Build Effective Agents\nBarry Zhang\n\nANTHROPIC\nModel Context Protocol\nIntegrating LLMs with your tools & data\nMahesh Murag\n\nModel Context Protocol\n\nStar History\nLegend: [illegible]\nY-axis: GitHub Stars\nY-axis ticks: 20K, 40K, 60K\nX-axis: Date\nTimeline: 2025, April, July, October\nstar-history.com\n\nWelcome to Claude Code research preview\nCLAUDE CODE\nPress Enter to continue...\n\nClaude Agent SDK\n\nANTHROPIC\nA timeline traces agent-building talks, MCP’s rising repository stars, Claude Code and the Claude Agent SDK. The speaker sees a shift toward tighter coupling between the model and its runtime environment.\nHow we used to think about agents\nHow we used to think about agents\n\nCoding Agent\nResearch Agent\nFinance Agent\nMarketing Agent\n\nANTHROPIC\n\nInterface text:\nAIE / CODE\nPRESENTED BY\nGoogle DeepMind\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nThe earlier model gave each domain its own agent, tools, and scaffolding. The slide shows separate model–tool loops for coding, research, finance, and marketing.\nCode is the universal interface\nCode is the universal interface\n\nCoding Agent\nUse cases\n\nANTHROPIC\n\nInterface text:\nAIE / CODE\nPRESENTED BY Google DeepMind\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nA coding agent can serve multiple use cases through code. For a financial report, it can pull data from APIs, organize files, and analyze the data with Python.\nWhat’s missing?\nWhat’s missing?\n\n1040 U.S. Individual Income Tax Return\n2024\n[illegible]\n\nMahesh\nMathematical Genius\n(Hypothetical)\n\nBarry\nTax Professional\n(Hypothetical)\n\nANTHROPIC\n\nInterface text:\nAIE / CODE\nPRESENTED BY\nGoogle DeepMind\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nThe speaker compares agents to a mathematical genius tackling taxes: both need domain expertise to execute reliably, as a tax professional would.\nIntroducing Agent Skills\nIntroducing Agent Skills\n\nANTHROPIC\n\nAIE / CODE\nPRESENTED BY\nGoogle DeepMind\n\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nAgent Skills package procedural knowledge in files for agents.\nWhat are Agent Skills?\nWhat are Agent Skills?\n\nSkills are organized collections of files that package composable procedural knowledge for agents\n\nanthropic_brand/\n├ SKILL.md\n├ docs.md\n├ slide-decks.md\n└ apply_template.py\n\nANTHROPIC\n\nAIE / CODE\nPRESENTED BY\nGoogle DeepMind\n\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nAgent Skills package procedural knowledge in folders. The example combines a SKILL.md file with supporting documents and a Python script.\nSkills are just folders\nSkills are just folders\n\nanthropic_brand/\n├ SKILL.md\n├ docs.md\n├ slide-decks.md\n└ apply_template.py\n\nANTHROPIC\n\nInterface text:\nAIE / CODE\nPRESENTED BY\nGoogle DeepMind\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nA skill folder can hold SKILL.md, documents, and a Python script. Teams can version the folder in Git or zip it for sharing.\nSkills can include scripts as tools\nSkills can include scripts as tools\n\nanthropic/brand_styling/slide-decks.md\n\n## Anthropic Slide Decks\n\n- Intro/outro slides\n  - background color: `#141413`\n  - foreground color: oat\n- Section slides:\n  - background color: `#da7857`\n  - foreground color: `#141413`\n\nUse the `./apply_template.py` script to update a pptx file in-place.\n\nanthropic/brand_styling/apply_template.py\n\nimport sys\nfrom pptx import Presentation\n\nif len(sys.argv) != 2:\n    print(\"USAGE: apply_template.py <pptx>\")\n    sys.exit(1)\n\nprs = Presentation(sys.argv[1])\nfor slide in prs.slides:\n    ...\n\nANTHROPIC\n\nInterface text:\nAIE / CODE\nPRESENTED BY Google DeepMind\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nThe slide styling instructions point to a Python script that updates a PowerPoint file in place. The speaker says Claude saved the script in the skill after repeatedly writing it to style slides.\nFoundational Skills\nThe Skills ecosystem\nFoundational Skills\n\nMy big day is coming! Help me make a detailed wedding planning checklist for the next 12 months. Organize by timeframe with tasks, vendor categories, and budget considerations.\n\nI want to surprise my wife with a trip to Italy! I have 6 months to save - here's my recent finances. Help me create a plan and a budget in Excel. Where can I save the money?!\n\nI need a business report for my food truck. Create a PDF with charts - best sellers, margin, etc- whatever you think is useful. Make it look great and help me make decisions!\n\nDocument Skills\nAnthropic\n\nClaude Scientific Skills\nLicense MIT  Skills 120\n\nA comprehensive collection of 120+ ready-to-use scientific skills for Claude, created by the K-Dense team. Transform Claude into your AI research assistant capable of executing complex multi-step scientific workflows across biology, chemistry, medicine, and beyond.\n\nThese skills enable Claude to seamlessly work with specialized scientific libraries, databases, and tools across multiple scientific domains:\n• Bioinformatics & Genomics - Sequence analysis, single-cell RNA-seq, gene regulatory networks, variant annotation, phylogenetic analysis\n• Cheminformatics & Drug Discovery - Molecular property prediction, virtual screening, ADMET analysis, molecular docking, lead optimization\n• Proteomics & Mass Spectrometry - LC-MS/MS processing, peptide identification, spectral matching, protein quantification\n• Clinical Research & Precision Medicine - Clinical trials, pharmacogenomics, variant interpretation, drug safety, precision therapeutics\n• Healthcare AI & Clinical ML - EHR analysis, physiological signal processing, medical imaging, clinical [cut off]\n\nScientific Skills\nK-Dense-AI\n\nEmbedded document text: WEDDING PLANNING JOURNEY; Financial Performance Analysis; [illegible]\n\nANTHROPIC\n\nInterface text:\nAI Engineer\nEngineering the future of AI\nFoundational skills give agents general or domain-specific capabilities. Anthropic’s document skills create and edit Office documents; K-Dense’s scientific skills support research tasks such as EHR analysis.\nPartner Skills\nThe Skills ecosystem\nPartner Skills\n\nAutomate anything in the browser, reliably\nResearch agent\nTask agent\nAuthenticated agent\nGet detailed research with live information\nWhat are some ongoing and active clinical trials for Parkinson’s?\n[illegible]\nFind me available apartments and their floorplans\n\nNotion Skills for Claude\n\nNotion Skills for Claude are step-by-step guides that teach Claude how to do real work in Notion. Instead of just answering questions, Skills help Claude complete entire workflows—structuring pages the way your team does, updating databases with the right info, following your brand guidelines, and working through your actual patterns.\n\nUse them to turn a spec into an implementation plan, run research that ends up as a beautiful Notion doc, prep for meetings with context already pulled together, or capture decisions into your wiki or task database. It’s a faster way to go from idea to done—and everything ends up organized, shareable, and ready for your team to use.\n\nTo get started: Download the .zip below, then upload each Skill to Claude. Once loaded, Claude [illegible]\n\nBrowserbase\nNotion\n\nANTHROPIC\n\nInterface text:\nAI Engineer\nEngineering the future of AI\nBrowserbase’s Stagehand skill helps Claude navigate the web; Notion’s skills help it understand and research a Notion workspace.\nTrends in the Skills ecosystem\nTrends in the Skills ecosystem\n\nMore complex, production-grade skills\nSkills complementing MCP servers\n\nANTHROPIC\n\nInterface text:\nAI Engineer\nEngineering the future of AI\nTrends in the Skills ecosystem\n\nMore complex, production-grade skills\nSkills complementing MCP servers\nNon-developers building high-value skills\n\nANTHROPIC\n\nInterface text:\nAI Engineer\nEngineering the future of AI\nSkills are becoming more complex and reaching non-developers. They also orchestrate MCP tools: MCP provides external connections, while skills provide expertise.\nSkills: the complete picture\nSkills: the complete picture\n\nAgent\n</>\n\nMCP server 1\nMCP server 2\nMCP server 3\n\nANTHROPIC\n\nInterface text:\nAIE / CODE\nPRESENTED BY Google DeepMind\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nThe agent loop couples a model with a code runtime and connects to MCP servers for outside tools and data. The agent can pull relevant skills into context at runtime for a particular task.\nExploring how Skills evolve: domain offerings\nExploring how Skills evolve\n\nClaude for Financial Services\nClaude for Life Sciences\n\nANTHROPIC\n\nInterface text:\nAIE / CODE\nPRESENTED BY\nGoogle DeepMind\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nAnthropic's financial services and life sciences offerings each paired Claude with MCP servers and skills for professionals in those domains.\nExploring how Skills evolve: development practices\nExploring how Skills evolve\n\nEvaluation\nBrand Style Skill\nTriggering accuracy\n8/10 cases correctly triggered\nOutput quality\nMatch expected format\nScript execution\napply_template.py failed\nSKILL.md structure\nValid YAML frontmatter\n\nVersioning\n2025-10-11\n2025-10-31\n2025-11-21\n\nComposability\nBranded Decks\nBrand Style Skill + PowerPoint Skill\nPowerPoint Skill\nMCP\npandas\n\nANTHROPIC\n\nInterface text:\nAIE / CODE\nPRESENTED BY\nGoogle DeepMind\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nAnthropic is exploring evaluation, version history and composition for skills. The example combines brand and PowerPoint skills for branded decks, while a PowerPoint skill references MCP and pandas.\nExploring how Skills evolve\nExploring how Skills evolve\n\nUser 1\nUser 2\nUser 3\nUser 4\nUser 5\n\nANTHROPIC\n\nInterface text:\nAIE / CODE\nPRESENTED BY\nGoogle DeepMind\nDON'T BUILD AGENTS, BUILD SKILLS INSTEAD\nBARRY ZHANG\nMember of Technical Staff\nMAHESH MURAG\nMember of Technical Staff\nANTHROPIC\nFive users point toward a shared toolbox. The speakers envision an evolving knowledge base of capabilities, curated by people and agents and shared across an organization.\nSkills are a concrete step towards continuous learning\nSkills are a concrete step towards continuous learning\n\nDAY 1\nNo skills\nIntelligent\n\nDAY 5\nA few skills\nCapable\n\nANTHROPIC\n\nInterface text:\nAI Engineer\nEngineering the future of AI\nSkills are a concrete step towards continuous learning\n\nDAY 1\nNo skills\nIntelligent\n\nDAY 5\nA few skills\nCapable\n\nDAY 30\nMany skills\nUseful\n\nANTHROPIC\n\nInterface text:\nAI Engineer\nEngineering the future of AI\nThe curve sketches a goal: Claude progresses from intelligent with no skills on day 1 to useful with many skills on day 30. Skills retain procedural knowledge for specific tasks.\nMoving up the stack\nMoving up the stack\n\nSkills\nApplications\n\nAgents\nOperating Systems\n\nModels\nProcessors\n\nANTHROPIC\n\nInterface text:\nAI Engineer\nEngineering the future of AI\nThe slide compares models to processors, agents to operating systems, and skills to applications. The speaker hopes skills will open that application layer to people solving concrete problems.\nStop Building Agents, Build Skills Instead\nANTHROPIC\n\nStop Building Agents\nBuild Skills Instead\n\nBarry Zhang\nMahesh Murag\n\nCome work on Skills\nanthropic.com/careers\n\nBuild Skills\ngithub.com/anthropics/skills\n\nAI Engineer\nEngineering the future of AI\nThe speakers propose skills as a way to ship and share capabilities for general agents, and invite viewers to build skills or join Anthropic.\nAI Engineer Code Summit\nAI Engineer\nCode Summit\nhttps://ai.engineer",
    "slides": {
      "url": "https://assets.ai.engineer/sites/aiecode2025/knowledge-corpus/objects/sha256/9c/9c9ac6fd2f996d6e2a26195a10968b2ecdaf0e74a56b73b2d506313083f23183.json",
      "sha256": "9c9ac6fd2f996d6e2a26195a10968b2ecdaf0e74a56b73b2d506313083f23183",
      "bytes": 27316
    }
  },
  {
    "slug": "ai-coding-workflow",
    "title": "AI Coding Workflow: From Product Idea to Tested Implementation",
    "url": "https://ai.engineer/talks/-QFHIoCo-Ko-ai-coding-workflow",
    "videoId": "-QFHIoCo-Ko",
    "event": "AI Engineer Europe 2026",
    "durationMs": 5790000,
    "speakers": [
      {
        "name": "Matt Pocock"
      }
    ],
    "topics": [
      {
        "slug": "infrastructure-and-deployment",
        "name": "Infrastructure & deployment"
      },
      {
        "slug": "coding-agents",
        "name": "Coding agents"
      },
      {
        "slug": "safety-and-governance",
        "name": "Safety & governance"
      },
      {
        "slug": "agent-engineering",
        "name": "Agent engineering"
      },
      {
        "slug": "design-engineering-and-ai-interfaces",
        "name": "AI interface design"
      },
      {
        "slug": "forward-deployed-engineering",
        "name": "Forward-deployed engineering"
      }
    ],
    "summary": "Matt Pocock presents a hands-on workshop on building a full AI-assisted coding workflow, arguing that software engineering fundamentals—not hype—make agents effective. He introduces the 'smart zone' and 'dumb zone' of LLMs (performance drops after ~100k tokens) and the 'Memento problem' (agents forget between sessions). His process starts with a 'Grill Me' skill that relentlessly questions the user until shared understanding is reached, then produces a PRD without reading it, slices work into vertical 'tracer bullet' issues, and runs agents AFK using TDD. He advocates designing codebases with deep, testable modules and shows Sandcastle, a TypeScript library for parallel agent execution with separate implementer (Sonnet) and reviewer (Opus). The workshop transforms ambiguous briefs into shippable features while keeping humans in the loop for QA and taste.",
    "slideText": null,
    "slides": null
  }
]

slug · name · url · biography · jobTitle · organization · portraitUrl · profiles · affiliations · topics · talkSlugs

[
  {
    "slug": "swyx",
    "name": "swyx",
    "url": "https://ai.engineer/speakers/swyx",
    "biography": "swyx, also known as Shawn Wang, is the **co-founder and chief executive of [AI Engineer](https://ai.engineer/about)**, editor and podcast co-host of [Latent Space](https://www.latent.space/about), and founder of Smol AI. His 2023 essay **[The Rise of the AI Engineer](https://www.latent.space/p/ai-engineer)** described an emerging profession of developers turning foundation-model capabilities into products. Through publishing, conferences, and software experiments, he has helped organize that community and explain its work: building the software, managing the product-specific data, and evaluating what models actually do.\n\n### From finance to building software\n\nA Singaporean based in San Francisco, swyx began his career in finance. He worked as a derivatives trader at Standard Chartered and a technology, media, and telecommunications equities analyst at Balyasny Asset Management. Programming was already part of those jobs: he wrote options-pricing tools in Haskell and used Python for financial analysis.\n\nIn 2016, he moved to New York to become a nontechnical product manager at Sentieo. Creating its Similar Tables feature exposed a limitation of that role: implementing his ideas depended on persuading engineers to build them. In his [career retrospective](https://swyx.io/decade), he connects that frustration to his decision to learn programming seriously. During 2017, he studied through freeCodeCamp and online tutorials, including an intensive period in Chiang Mai, before concluding that he needed more structured training and attending Fullstack Academy. Software work at Two Sigma followed, then developer advocacy and experience roles at Netlify, Amazon Web Services, Temporal, and Airbyte.\n\nWriting and speaking became part of that transition. In 2018, his writing about React Suspense brought him to Dan Abramov’s attention and led to moderating the React subreddit. He later co-founded Svelte Society and helped build the React TypeScript Cheatsheet community, contributing to shared resources for developers learning unfamiliar tools.\n\nAt Temporal, where he became Head of Developer Experience, his interests expanded into distributed systems. His [argument for workflows as code](https://swyx.io/why-temporal) starts with a business process crossing several services: one rate-limits requests, another stops responding, and unfinished work needs to survive both. He explained the appeal of preserving progress and standardizing retries and timeouts while keeping the process readable as an ordinary program.\n\n### Learning in public and finding an AI audience\n\n**[Learn In Public](https://swyx.io/learn-in-public)**, published in 2018, asks developers to create the notes, examples, tutorials, and contributions they wish they had found while learning. The intended reader can be their own earlier self. Explaining a problem exposes gaps in understanding; publishing invites correction and gives experienced developers something specific to respond to. Contributing to the people and projects one learns from can turn that feedback into collaboration and mentorship. Audience size is secondary to the learning itself.\n\nHis **[The Coding Career Handbook](https://learninpublic.org/)** extends that approach into the decisions surrounding a developer’s working life. Published in 2020 and made free and open source in August 2026, it covers the first job hunt, specialization, technical growth, negotiation, writing, and advancement. Its premise is that learning to code alone does not teach someone how to choose opportunities, communicate their work, or grow into broader responsibilities.\n\nswyx started Latent Space’s newsletter in 2022 and launched its podcast with Alessio Fanelli in 2023. Detailed conversations about models, infrastructure, and applications became a way to explore a new field. His retrospective also records missed opportunities to join an AI lab and a deep-learning education he did not pursue further; the move into AI was less settled than a smooth career chronology might suggest.\n\nIn June 2023, **The Rise of the AI Engineer** gave that audience a professional identity. swyx argued that model APIs and open models let software developers explore capabilities and validate products without first assembling a research team or training a model. That access brought responsibilities beyond prompting: product-specific data, evaluation, orchestration, and the software needed to turn a prototype into a product.\n\nHe explicitly described a trend he was calling attention to, rather than work he had invented. The distinction from machine-learning research was also permeable: AI engineers could fine-tune or host models, while research engineers could build applications. With Benjamin Dunphy, he co-founded AI Engineer to give these builders a place to meet and exchange practical knowledge.\n\n### smol developer and the move to Cognition\n\nHis software experiments accompanied the argument. In 2023, swyx launched **[smol developer](https://github.com/smol-ai/developer)**, a project for generating a codebase from a product specification and embedding a developer agent inside another application. It separated planning, identifying files, and generating code into reusable functions, allowing a developer to inspect or revise the plan before continuing.\n\nAn early retest undermined his confidence. A week after his initial experiments, the same agent against the same endpoints could no longer do what had impressed him. He turned to other projects, including AI News. The failure became part of his later explanation for [joining Cognition in September 2025](https://swyx.io/cognition): it represented the kind of software-agent company he wished smol developer had become.\n\nHe announced that he and most of Smol AI would join Cognition while AI Engineer and Latent Space continued independently. He retained the Smol AI domain, existing code IP, and AI News. He subsequently became an adviser to Cognition, with work including coding-agent evaluation and codebase understanding through Codemaps.\n\n### Patterns, team workflows, and standards for output\n\n- **Reusable patterns for AI-intensive applications.** swyx’s [SPADE framework](https://www.ai.engineer/talks/IHkyFhU6JEY-designing-ai-intensive-applications) describes recurring work: synchronize inputs, plan and parallelize tasks, analyze and consolidate results, deliver an output, and evaluate it. AI News makes the pattern concrete. Collecting material from multiple communities and reducing it into a useful digest requires coordinated processing across many model calls. SPADE is a proposed mental model for that coordination, paired with his preference for measuring valuable output against the human effort required rather than debating whether a workflow qualifies as an agent.\n\n- **Agents for everything else.** Using agents inside AI Engineer changed swyx’s understanding of their value. After he connected Devin to Figma with help from Claude Cowork, the team’s designer could turn designs into a functioning website and continue iterating while swyx was asleep. Removing the wait for a developer also made room for animations and polish the designer could pursue directly. His [account of agents beyond coding](https://www.youtube.com/watch?v=zepu8Kk6FBQ) extends that approach to schedules maintained as code, external-data synchronization, and purchasing research. These uses inform his emphasis on APIs, command-line tools, and MCP integrations that let agents operate existing software. He also acknowledges the cost of failed automation: when proposing to replace a SaaS tool, he describes identifying teammates’ three leading concerns and working through them, because those teammates will inherit the mistakes.\n\n- **No more slop.** swyx defines slop as low-quality, inaccurate, or inauthentic work, whether a human or a model produces it. His [case for accountability in AI-assisted work](https://www.youtube.com/watch?v=IoiHI7p12Ao) challenges claims about hours of autonomous execution that leave the quality of the resulting code unanswered. He favors keeping human attention close to difficult problems while delegating more routine work asynchronously, and using AI to improve codebase understanding through work such as Codemaps. AI News applies the editorial version of that standard: it should be willing to tell readers when little deserves their attention.\n\n### Extending AI engineering into science\n\nIn January 2026, swyx helped launch [Latent Space’s AI for Science podcast](https://www.latent.space/p/science), hosted by RJ Honicky and Brandon Anderson. He argued that engineers’ skills could contribute to scientific work alongside domain experts, while acknowledging that he and Alessio brought developer-tools experience and needed hosts with scientific training and experience building science companies.\n\nHis interest in automated research has also involved a change of mind. In an [August 2026 post about human simulation](https://x.com/swyx/status/2090948945753076141), swyx described moving from skepticism about simulation as a scaling-law claim to taking simulated human feedback seriously as a possible component of increasingly automated research. He presented that shift as a revision of his own expectations, rather than an established technical result.",
    "jobTitle": "Curator",
    "organization": "AI Engineer",
    "portraitUrl": "https://ai.engineer/wf26/speakers/by-id/spk_shawn_wang.jpg",
    "profiles": {
      "website": null,
      "github": null,
      "linkedin": null,
      "x": null
    },
    "affiliations": [
      {
        "organization": "AI Engineer",
        "eventSlug": "ai-engineer-world-s-fair-2026",
        "eventYear": 2026
      },
      {
        "organization": "AI Engineer",
        "eventSlug": "ai-engineer-europe-2026",
        "eventYear": 2026
      },
      {
        "organization": "AI Engineer",
        "eventSlug": "ai-engineer-code-2025",
        "eventYear": 2025
      },
      {
        "organization": "Latent.Space",
        "eventSlug": "ai-engineer-world-s-fair-2025",
        "eventYear": 2025
      },
      {
        "organization": "Latent.Space",
        "eventSlug": "ai-engineer-summit-2025",
        "eventYear": 2025
      },
      {
        "organization": "Latent.Space",
        "eventSlug": "ai-engineer-world-s-fair-2024",
        "eventYear": 2024
      },
      {
        "organization": "Latent.Space",
        "eventSlug": "ai-engineer-summit-2023",
        "eventYear": 2023
      }
    ],
    "topics": [
      {
        "slug": "agent-engineering",
        "name": "Agent engineering"
      },
      {
        "slug": "agent-memory",
        "name": "Agent memory"
      },
      {
        "slug": "agent-orchestration",
        "name": "Agent orchestration"
      },
      {
        "slug": "ai-engineering-leadership",
        "name": "AI engineering leadership"
      },
      {
        "slug": "ai-in-finance",
        "name": "AI in finance"
      },
      {
        "slug": "ai-platform-engineering",
        "name": "AI platform engineering"
      },
      {
        "slug": "ai-startups-and-small-teams",
        "name": "AI startups & small teams"
      },
      {
        "slug": "coding-agents",
        "name": "Coding agents"
      },
      {
        "slug": "enterprise-ai",
        "name": "Enterprise AI"
      },
      {
        "slug": "fine-tuning",
        "name": "Fine-tuning"
      },
      {
        "slug": "generative-media",
        "name": "Generative media"
      },
      {
        "slug": "human-and-agent-accessibility",
        "name": "Human & agent accessibility"
      },
      {
        "slug": "human-in-the-loop-validation",
        "name": "Human-in-the-loop validation"
      },
      {
        "slug": "infrastructure-and-deployment",
        "name": "Infrastructure & deployment"
      },
      {
        "slug": "model-context-protocol",
        "name": "MCP"
      },
      {
        "slug": "machine-learning-fundamentals",
        "name": "ML fundamentals"
      },
      {
        "slug": "multi-agent-systems",
        "name": "Multi-agent systems"
      },
      {
        "slug": "multimodal-models-and-applications",
        "name": "Multimodal AI"
      },
      {
        "slug": "product-engineering",
        "name": "Product engineering"
      },
      {
        "slug": "rag-and-knowledge",
        "name": "Retrieval-augmented generation"
      },
      {
        "slug": "safety-and-governance",
        "name": "Safety & governance"
      },
      {
        "slug": "workflow-automation",
        "name": "Workflow automation"
      }
    ],
    "talkSlugs": [
      "6-things-to-know-about-aie-world-s-fair-2026",
      "agents-for-everything-else-swyx",
      "ai-engineering-without-borders",
      "define-ai-engineer",
      "designing-ai-intensive-applications",
      "no-more-slop-swyx",
      "software-engineering-ai",
      "the-1-000x-ai-engineer-swyx",
      "why-agent-engineering",
      "youtube-1719dc0dd72889b66132",
      "youtube-8888f18cd1c77795639c",
      "youtube-95bc1caadf36a40e16d7",
      "youtube-b01c3c14dd8f8e90af4c"
    ]
  },
  {
    "slug": "charles-frye",
    "name": "Charles Frye",
    "url": "https://ai.engineer/speakers/charles-frye",
    "biography": "Charles Frye is a [member of technical staff at Modal](https://www.linkedin.com/in/charles-frye-38654abb/details/experience/), where he helps teams deploy and optimize neural-network inference. He primarily authored Modal’s [GPU Glossary](https://modal.com/gpu-glossary), explaining how accelerator hardware and model execution shape application design. His work connects two practical questions: how to make better use of expensive computing resources, and how to tell whether the outputs are worth producing.\n\n### From neuroscience research to complete applications\n\nCharles studied biology and computational neuroscience at the University of Chicago before pursuing doctoral research at UC Berkeley’s Redwood Center for Theoretical Neuroscience, supported by an NSF Graduate Research Fellowship. He [completed his neuroscience doctorate in 2020](https://charlesfrye.github.io), studying the geometry of neural-network optimization.\n\nHis [first-author paper on gradient-flat regions](https://arxiv.org/abs/2003.10397), written with five collaborators, identified a problem with methods used to investigate that geometry. Algorithms searching for critical points—where a loss function’s gradient vanishes—could instead converge to regions where the gradient norm was stationary while the gradient itself remained nonzero. In these **gradient-flat regions**, the gradient lies approximately in the Hessian’s null space, making the loss locally approximately linear along the gradient direction. Mistaking these regions for critical points could distort conclusions about neural-network loss landscapes and the optimization methods operating on them.\n\nTeaching developed alongside his research. Charles created browser-accessible mathematics and statistics notebooks and designed a [course in data science for research psychology](https://www.linkedin.com/in/charles-frye-38654abb) in which students constructed probabilistic models using Python and PyMC3. Interactive notes, labs, and automated feedback made the material available beyond the classroom.\n\nAt [Weights & Biases](https://www.linkedin.com/in/charles-frye-38654abb/details/experience/), he worked first as a deep-learning instructor and then as an educator, producing material on machine learning, Python, and applied mathematics. From February 2022 to February 2024, he was a deep-learning educator at The Full Stack. Through **Full Stack Deep Learning**, he taught engineers to build complete machine-learning applications, from GPUs and training through deployment, user experience, and monitoring.\n\nHis [askFSDL assistant](https://fullstackdeeplearning.com/llm-bootcamp/spring-2023/askfsdl-walkthrough/) put that approach into practice with a bot that answered questions using retrieved course materials. Initially, scraping and chopping up the material produced poor results; preserving source-data structure improved the system. In his [2023 account of building with LLMs](https://www.youtube.com/watch?v=zl4EdALzktU), Charles described another adjustment prompted by actual use: people asked what the bot could do and how it operated, questions he had not anticipated. He added prompt instructions for that class of question. His approach to retrieval extends beyond vector similarity to keyword search, metadata filtering, and combinations suited to the information being sought.\n\n### When self-hosting became worthwhile\n\nIn January 2023, Charles [ran GLM-130B on an eight-GPU machine](https://charlesfrye.github.io/programming/2023/01/12/running-glm-130b.html). He wanted experience operating a language model himself and was concerned about concentrating a new computing capability in privately controlled services. Earlier openly available models had not delivered the free-form generation he needed.\n\nThe experiment worked, but it exposed difficult software dependencies, weaker instruction following than the OpenAI models he was accustomed to, and serving costs substantially above the API after a day of work. He also found an error in his own German prompt that complicated judging a translation. His interest in self-hosting survived those results without turning them into a claim that it was already the better option.\n\nIn his [2023 inference workshop](https://www.youtube.com/watch?v=N7lJY5IKVLE), Charles argued that open and proprietary systems could coexist: accessible weights permit deployment and inference customization, while hosted services offer convenience, capability, and enterprise support. His conditional prediction was that open models would become especially competitive when applications needed a sufficient level of capability rather than the strongest available model. By [2025](https://www.youtube.com/watch?v=DeFF3J8T5Pk), he saw improvements in both model quality and inference engines making **open-weight model self-hosting** practical for more applications.\n\nCharles joined Modal as a developer advocate in February 2024 and became a member of technical staff in February 2026. His [account of Modal as a distributed computer](https://modal.com/blog/modal-is-a-computer) explains its architecture through operating-system concepts: container images become running programs, shared files are reused, storage forms a hierarchy of caches, and routing carries information to and from containers. The analogy explains how the platform combines and shares cloud resources without assigning its construction to Charles alone.\n\n### Making expensive hardware do useful work\n\n- **GPU literacy for application decisions.** Charles’s [GPU teaching](https://www.youtube.com/watch?v=y-UGrYbJsJk) emphasizes **arithmetic intensity**—computation performed per memory load—and low-precision matrix-matrix multiplication. Sequential token generation can repeatedly move model weights while leaving arithmetic capacity underused. Batching requests, generating multiple candidates, or checking proposed tokens together can expose more parallel work. In their [Whisper batching article](https://modal.com/blog/batching-whisper), Charles and Cathy Zhou explain the serving tradeoff: process requests together, but impose a waiting-time limit so sporadic traffic does not leave users waiting indefinitely for a full batch.\n\n- **Benchmarks that answer deployment questions.** Charles’s [LLM Engineer’s Almanac benchmarking work](https://www.youtube.com/watch?v=DeFF3J8T5Pk) grew from repeated customer questions about particular models and workloads. It compared vLLM, SGLang, and TensorRT-LLM across model configurations and input/output lengths, relating throughput to first-token latency. The initial results measured out-of-the-box performance rather than exhaustively tuned engines. They made the cost difference between parallel prompt processing and sequential output generation visible. Charles recommends considering better context before automatically requesting longer generated reasoning; the performance results do not establish which choice improves every task’s answers.\n\n- **Keeping the CPU out of the GPU’s way.** In his [explanation of inference engines](https://www.youtube.com/watch?v=woIYJYd_etI), Charles identifies scheduling and host-device synchronization as potential bottlenecks. Faster GPU arithmetic does little good if CPU work leaves the accelerator idle. CUDA graphs address one part of this problem by recording a sequence of GPU operations that can then be launched together, reducing repeated host-side launch work. He treats correctness as a separate obligation: performance optimizations must also be checked for changes in model behavior.\n\nHis [speculative-decoding article with Shankha Biswas](https://modal.com/blog/achieve-sota-specdec) develops a complementary approach. A smaller draft model proposes tokens that the target model checks in parallel, turning some sequential decoding into small parallel processing steps. Task-specific training can increase the number of accepted tokens per iteration. Their account describes collaborative work involving Modal colleagues, Decagon, Z Lab, and SGLang, including training on synthetic data when production data is restricted and checking for failures to generalize beyond the training distribution.\n\nCharles also explains where batching intuition stops applying. His [discussion of KV-cache arithmetic intensity](https://x.com/charles_irl/status/2090169189189448123) distinguishes shared model weights from context-specific cached information. Independent requests reuse model weights, but attention generally loads different cached context for each request unless prefixes overlap. Batching therefore does not improve every component of inference equally.\n\n### When a beautiful output still fails\n\nThe [QArt codes project account](https://modal.com/blog/qart-codes-evals), coauthored with Will Shainin and Erik Dunteman, gives Charles’s evaluation work a concrete test: an attractive generated image must also transmit the information encoded in its QR pattern.\n\nThe team compared automated scanning and aesthetic evaluators with human judgments before using them for larger experiments. Will performed the laborious manual scanning that helped establish whether automated checks matched phone-based use. A configuration that produced better-looking images initially reduced scan success, so choosing parameters alone did not resolve the tradeoff.\n\nThe team then applied **inference-time compute scaling**: generate eight candidates per request, evaluate them, and display four ranked first by scannability and then by aesthetics. They reported reaching a 95% scan-rate objective measured per batch while keeping end-to-end latency below 20 seconds at the 95th percentile. Additional generation helped because the system could select usable results. Charles explores the same mechanism for coding tasks, where executable tests can identify successful candidates among multiple smaller-model outputs.\n\nWith Joy Liu, Charles coauthored [Modal Vibe](https://modal.com/blog/modal-vibe), a coding-platform demonstration built around a sandbox manager, a frontend, and isolated execution environments. Their account examines the startup and scaling demands of running many generated applications concurrently and explicitly cautions against copying the demo directly into production. His [satire about ever more elaborate agent harnesses](https://x.com/charles_irl/status/2091329494909120959) questions the expectation that additional orchestration alone will turn unreliable generation into dependable software.\n\nCharles’s applications also include a [camera-driven installation](https://www.youtube.com/watch?v=woIYJYd_etI) that turns whatever it sees into a museum-style descriptive placard, using a Qwen vision-language model served through SGLang on Modal. It supplies the curatorial language that makes an ordinary scene sound like an artwork—a playful use of the same inference systems he measures and explains.",
    "jobTitle": "Member of Technical Staff",
    "organization": "Modal",
    "portraitUrl": "https://ai.engineer/wf26/speakers/by-id/spk_charles_frye.jpg",
    "profiles": {
      "website": null,
      "github": null,
      "linkedin": null,
      "x": null
    },
    "affiliations": [
      {
        "organization": "Modal",
        "eventSlug": "ai-engineer-world-s-fair-2026",
        "eventYear": 2026
      },
      {
        "organization": "Modal",
        "eventSlug": "ai-engineer-world-s-fair-2025",
        "eventYear": 2025
      },
      {
        "organization": "Modal",
        "eventSlug": "ai-engineer-world-s-fair-2024",
        "eventYear": 2024
      }
    ],
    "topics": [
      {
        "slug": "agent-engineering",
        "name": "Agent engineering"
      },
      {
        "slug": "agent-memory",
        "name": "Agent memory"
      },
      {
        "slug": "ai-cost-and-performance-engineering",
        "name": "AI cost & performance"
      },
      {
        "slug": "ai-engineering-leadership",
        "name": "AI engineering leadership"
      },
      {
        "slug": "ai-platform-engineering",
        "name": "AI platform engineering"
      },
      {
        "slug": "coding-agents",
        "name": "Coding agents"
      },
      {
        "slug": "developer-workflows-and-testing",
        "name": "Developer workflows & testing"
      },
      {
        "slug": "enterprise-ai",
        "name": "Enterprise AI"
      },
      {
        "slug": "evals",
        "name": "Evals"
      },
      {
        "slug": "generative-media",
        "name": "Generative media"
      },
      {
        "slug": "gpu-programming-and-kernel-optimization",
        "name": "GPU programming & kernels"
      },
      {
        "slug": "inference",
        "name": "Inference"
      },
      {
        "slug": "infrastructure-and-deployment",
        "name": "Infrastructure & deployment"
      },
      {
        "slug": "kv-caching-and-prefix-reuse",
        "name": "KV caching & prefix reuse"
      },
      {
        "slug": "local-inference",
        "name": "Local inference"
      },
      {
        "slug": "model-context-protocol",
        "name": "MCP"
      },
      {
        "slug": "machine-learning-fundamentals",
        "name": "ML fundamentals"
      },
      {
        "slug": "model-routing-and-llm-gateways",
        "name": "Model routing & gateways"
      },
      {
        "slug": "observability-and-reliability",
        "name": "Observability & reliability"
      },
      {
        "slug": "prompting-and-in-context-learning",
        "name": "Prompting & in-context learning"
      },
      {
        "slug": "quantization",
        "name": "Quantization"
      },
      {
        "slug": "rag-and-knowledge",
        "name": "Retrieval-augmented generation"
      },
      {
        "slug": "safety-and-governance",
        "name": "Safety & governance"
      },
      {
        "slug": "speculative-decoding",
        "name": "Speculative decoding"
      },
      {
        "slug": "structured-outputs",
        "name": "Structured outputs"
      }
    ],
    "talkSlugs": [
      "ai-engineering-201-inference",
      "ai-engineering-201-the-rest-of-the-owl",
      "how-fast-are-llm-inference-engines-anyway",
      "what-every-ai-engineer-needs-to-know-about-gpus",
      "what-we-learned-from-a-year-of-building-with-llms",
      "youtube-a1ed0627f76116a7ee43"
    ]
  }
]

slug · name · url · description · talkSlugs

[
  {
    "slug": "infrastructure-and-deployment",
    "name": "Infrastructure & deployment",
    "url": "https://ai.engineer/topics/infrastructure-and-deployment",
    "description": null,
    "talkSlugs": [
      "1-ai-guardrails-the-unreasonable-effectiveness-of-finetuned-modernberts-diego-carpentero",
      "12-factor-agents",
      "200-million-patient-interactions-later-what-the-generic-voice-stack-misses",
      "360brew-llm-based-personalized-ranking-and-recommendation-hamed-firooz-and-maziar-sanjabi-linked",
      "a-piece-of-pi-embedding-the-openclaw-coding-agent-in-your-product",
      "a-practical-guide-to-efficient-ai",
      "a-song-of-types-and-agents",
      "a2a-mcp-automating-business-processes-with-llms",
      "accelerating-ai-on-edge-chintan-parikh-and-weiyi-wang-google-deepmind",
      "accelerating-mixture-of-experts-training-with-rail-optimized-infiniband-networking-in-crusoe-clo",
      "active-graph-agent-runtime-babyagi-4",
      "adaption-labs-gradient-free-continual-learning",
      "agent-output-is-not-ux-rendering-layer-your-llm-pipeline-is-missing-bala-ramdoss-amazon-lens",
      "agent-reinforcement-fine-tuning",
      "agentcraft-putting-the-orc-in-agent-orchestration",
      "agentic-enterprise-what-your-ceo-must-know-about-ai",
      "agents-at-scale-inside-minimax-s-model-and-the-infrastructure-behind-it",
      "agents-building-agents",
      "agents-don-t-do-standups-building-the-post-engineer-engineering-org",
      "agents-in-production-how-opengov-built-and-scaled-og-assist",
      "agents-need-receipts-not-more-tool-calls",
      "agents-need-receipts-not-more-tool-calls-505635",
      "agents-redacted-location-address-access-redacted-location-address-and-the-future-of-machine-iden",
      "agents-reported-thousands-of-bugs-how-many-were-real-ian-butler-and-nick-gregory",
      "agents-vs-workflows-why-not-both",
      "agi-the-path-forward",
      "ai-agent-harnesses",
      "ai-agents-for-performance-ship-faster-pay-less",
      "ai-agents-meet-test-driven-development",
      "ai-coding-workflow",
      "ai-consulting-in-practice-nathaniel-whittemore-nlw-superintelligent",
      "ai-copilots-for-tech-architecture-the-highest-roi-use-case-you-re-not-building-boris-bogatin-and",
      "ai-engineering-201-inference",
      "ai-kernel-generation-what-s-working-what-s-not-what-s-next",
      "ai-platform-engineering",
      "ai-system-design-from-idea-to-production",
      "ai-tools-for-forward-deployed-engineering",
      "alphalab-autonomous-multi-agent-research-across-optimization-domains-with-frontier-llms-brendan",
      "always-on-agents-run-production-without-the-on-call-tax",
      "anchoring-enterprise-genai-with-knowledge-graphs",
      "anthropic-for-vps-of-ai",
      "anthropic-s-applied-ai-team-on-the-evolution-of-agentic-surfaces",
      "anthropic-s-cca-exam-as-a-field-guide-for-agentic-engineering",
      "architecting-agent-memory-principles-patterns-and-best-practices",
      "architecting-and-testing-controllable-agents",
      "arrakis-how-to-build-an-ai-sandbox-from-scratch",
      "automating-escrow-with-usdc-and-ai",
      "autonomous-agents-for-scientific-tasks-sina-shahandeh-radicait",
      "ax-is-the-only-experience-that-matters",
      "balancing-innovation-with-security-safety",
      "bdd-adr-prd-wtf-capturing-decisions-for-humans-and-ai-alike-michal-cichra-safe-intelligence",
      "benchmarking-semantic-code-retrieval-on-claude-code",
      "beyond-apis-how-ai-web-agents-are-automating-the-long-tail-of-knowledge-work",
      "beyond-the-consensus-navigating-ai-s-frontier-in-2025",
      "botdojo-launch-enhancing-ai-assistants-with-evaluations-and-synthetic-data",
      "breaking-ai-s-1-gigahertz-barrier",
      "bringing-continual-learning-into-enterprises",
      "build-deploy-ai-powered-apps-18cc1b",
      "build-enterprise-generative-ai-apps-using-llama-3-at-1-000-tokens-s-on-the-sambanova-ai-platform",
      "build-evals-that-actually-matter-nick-ung-akshay-sharma-lyft",
      "build-systems-not-code",
      "build-the-ai-gtm-agent-that-knows-the-buyer-before-the-first-message",
      "building-a-10-person-unicorn",
      "building-a-smarter-ai-agent-with-neural-rag",
      "building-agent-interfaces-lessons-from-chrome-devtools-mcp-for-agents",
      "building-agentic-applications-with-heroku-managed-inference-and-agents-julian-duque-and-anush-ds",
      "building-agents-at-cloud-scale-antje-barth-aws",
      "building-agents-is-trivial-now-context-is-the-next-frontier",
      "building-agents-the-hard-parts",
      "building-agents-with-mcp",
      "building-ai-powered-developer-tools-at-jane-street",
      "building-ai-products-that-actually-work",
      "building-alice-s-brain-an-ai-sales-rep-that-learns-like-a-human-sherwood-satwik-11x",
      "building-an-agentic-platform",
      "building-applications-with-ai-agents",
      "building-ciso-approved-agent-fleet-architecture",
      "building-context-aware-reasoning-applications-with-langchain-and-langsmith",
      "building-cursor-composer",
      "building-deterministic-infrastructure-for-non-deterministic-ai-agents",
      "building-durable-agents-with-workflow-devkit-ai-sdk",
      "building-durable-production-ready-agents-with-openai-sdk-and-temporal",
      "building-effective-voice-agents",
      "building-enterprise-llm-agents-that-work",
      "building-generative-image-video-models-at-scale",
      "building-great-agent-skills",
      "building-multi-agent-systems-with-finite-state-machines",
      "building-multimodal-ai-agents-from-scratch",
      "building-redacted-username-gergely-orosz-simon-eskildsen",
      "building-reliable-support-agents-using-the-effect-typescript-library-michael-fester",
      "building-scaling-an-ai-agent-swarm-of-low-latency-real-time-voice-bots",
      "building-sota-open-weights-tool-use-the-command-r-family",
      "building-your-own-software-factory",
      "challenges-to-scaling-agents-for-generative-ai-products",
      "ci-cd-is-dead-agents-need-continuous-compute-and-computers-hugo-santos-and-madison-faulkner",
      "ci-in-the-era-of-ai-from-unit-tests-to-stochastic-evals",
      "claude-agent-sdk-workshop",
      "code-world-model-building-world-models-for-computation",
      "codex-and-subagents",
      "codex-behind-the-harness",
      "cognitive-shield-real-time-real-smart-rachna-srivastava",
      "comfyui-workshop-with-comfyanonymous-and-jedrick-kosinski",
      "compilers-in-the-age-of-llms",
      "compression-at-the-edge",
      "compute-system-design-for-next-generation-frontier-models",
      "computer-use-2-0-agents-just-got-multi-cursor",
      "containing-agent-chaos",
      "context-platform-engineering-to-reduce-token-anxiety-val-bercovici-and-callan-fox-weka",
      "continual-learning-for-ai-agents-from-failures-to-durable-improvements-soheil-feizi-relai",
      "cooking-with-fire-without-burning-down-the-kitchen",
      "covalent-launch-the-gpu-cheatcode-fine-tune-20-llama-models-in-5-minutes",
      "critical-ai-inference-your-cio-can-trust",
      "customized-production-ready-inference-with-open-source-models-dmytro-dima-dzhulgakov",
      "dark-factory-openclaw-ships-faster-than-you-can-read-the-diff",
      "data-quality-is-the-compute-multiplier",
      "data-readiness-is-a-myth-reliable-ai-with-an-agentic-semantic-layer-anushrut-gupta-promptql",
      "decoding-mistral-ai-s-large-language-models",
      "define-ai-engineer",
      "designing-agents-the-floor-is-the-frontier",
      "designing-voice-agents-for-real-conversations",
      "domain-adaptation-and-fine-tuning-for-domain-specific-llms",
      "don-t-build-slop-4-levels-of-ai-agent-maturity",
      "dream-machine-scaling-to-1m-users-in-4-days-keegan-mccallum-luma-ai",
      "e-values-evaluating-the-values-of-ai",
      "effective-agent-design-patterns-in-production",
      "effective-ai-agents",
      "effective-ai-agents-need-data-flywheels-not-the-next-biggest-llm-sylendran-arunagiri-nvidia",
      "efficient-reinforcement-learning",
      "engineering-voice-agents-latency-quality-and-scale",
      "ensure-ai-agents-work-evaluation-frameworks-for-scaling-success",
      "evals-are-not-unit-tests",
      "evals-driven-development-engineering-a-mental-health-ai-coach-ethically-safely",
      "evaluating-ai-agents",
      "events-are-the-wrong-abstraction-for-your-ai-agents",
      "every-solo-agent-builder-eventually-reinvents-a-worse-version-of-ci-cd",
      "everything-i-learned-training-frontier-small-models",
      "everything-we-knew-about-software-has-changed-theo-browne",
      "everything-you-need-to-know-about-agent-observability",
      "exposing-agents-as-mcp-servers-with-mcp-agent-sarmad-qadri",
      "fast-models-need-slow-developers",
      "feedback-loops-are-all-you-need",
      "fighting-slop-with-slop",
      "forward-deployed-engineering-101",
      "forward-deployed-engineering-at-cursor-pauline-brunet",
      "foundry-local-cutting-edge-ai-experiences-on-device-with-onnx-runtime-and-olive-emma-ning-micros",
      "from-46-to-90-fine-tuning-tiny-llms-for-on-device-agents-cormac-brick-google",
      "from-blind-spots-to-merged-prs-continuous-agentic-performance-optimization",
      "from-chaos-to-choreography-multi-agent-orchestration-patterns-that-actually-work-sandipan-bhaumi",
      "from-fork-to-fleet-designing-an-agent-sandbox-cloud-abhishek-bhardwaj-openai",
      "from-mcp-to-scale-pipelines-that-build-themselves",
      "from-mixture-of-experts-to-mixture-of-agents-with-super-fast-inference",
      "from-model-weights-to-api-endpoint-with-tensorrt-llm",
      "from-stateless-nightmares-to-durable-agents",
      "from-tokens-to-cells-foundation-models-for-single-cell-biology-akram-baharlouei-altos-labs",
      "from-writing-code-to-designing-systems-how-the-developer-role-is-changing-chris-noring-microsoft",
      "frontier-ai-at-home-literally",
      "frontier-feud",
      "frontier-results-on-device-rl-nabors-arize",
      "fun-stories-from-building-openrouter-and-where-all-this-is-going",
      "future-of-knowledge-assistants",
      "future-proof-coding-agents",
      "gadgets-personal-app-vibe-coding-that-is-actually-safe",
      "gemini-nano-on-device-florina-muntenescu-oli-gaymond-google-deepmind",
      "gemma-4-deep-dive-cassidy-hardin-google-deepmind",
      "gemma-deepmind-s-family-of-open-models",
      "generative-video-at-the-speed-of-light",
      "git-push-get-an-ai-api",
      "giving-a-voice-to-ai-agents",
      "going-beyond-rag-extended-mind-transformers",
      "government-agents-ai-agents-meet-tough-regulations-mark-myshatyn-los-alamos-national-laboratory",
      "gpt-web-app-generator-10-000-apps-created-in-a-month-matija-sosic",
      "gpu-less-trust-less-limit-less-reimagining-the-confidential-ai-cloud",
      "graphrag",
      "grounded-reasoning-systems-for-cloud-architecture",
      "guardrails-first-engineering-member-facing-health-ai",
      "hacking-the-inference-pareto-frontier",
      "harness-engineering-is-not-enough-why-software-factories-fail",
      "harnessing-the-power-of-llms-locally",
      "healthcare-s-agent-bytecode-x12-as-the-harness-for-ai-agents",
      "how-agents-broke-app-level-infrastructure",
      "how-agents-will-unlock-the-500b-promise-of-ai",
      "how-autoresearch-is-changing-ml-research-zhengyao-jiang-weco-ai",
      "how-claude-code-works",
      "how-deep-research-works",
      "how-fast-are-llm-inference-engines-anyway",
      "how-forward-deployed-engineering-is-done-at-cognition",
      "how-forward-deployed-engineering-is-done-at-decagon",
      "how-forward-deployed-engineering-is-done-at-factory",
      "how-forward-deployed-engineering-is-done-at-ramp",
      "how-forward-deployed-engineering-is-done-redacted-username-kepler",
      "how-google-deepmind-is-researching-the-next-frontier-of-ai-for-gemini-raia-hadsell-vp-of-researc",
      "how-i-automate-my-own-job-at-hugging-face-using-agents",
      "how-instacart-transformed-its-search-and-discovery-using-an-llm-driven-approach",
      "how-llms-work-for-web-devs-gpt-in-600-lines-of-vanilla-js",
      "how-lovable-self-improves-every-hour",
      "how-to-build-agents-that-run-for-hours-without-losing-the-plot",
      "how-to-build-an-ai-native-health-company",
      "how-to-build-an-ai-strategy-that-fails",
      "how-to-build-enterprise-aware-agents",
      "how-to-build-planning-agents-without-losing-control-yogendra-miraje-factset",
      "how-to-build-the-world-s-fastest-voice-bot",
      "how-to-build-world-class-ai-products-sarah-sachs-notion-and-carlos-esteban-braintrust",
      "how-to-build-your-own-ai-data-center-in-2025",
      "how-to-defend-your-sites-from-ai-bots",
      "how-to-hire-ai-engineers-when-everyone-is-cheating-with-ai",
      "how-to-leverage-domain-expertise-chris-lovejoy-notius-labs",
      "how-transformers-finally-ate-vision",
      "how-we-built-zeta2-training-an-edit-prediction-model-in-production-ben-kunkle-zed",
      "how-we-hacked-yc-spring-2025-batch-s-ai-agents",
      "how-web-data-infrastructure-powers-the-next-generation-of-ai",
      "hyperspace-more-nodes-is-all-you-need",
      "i-run-a-fleet-of-ai-agents-across-three-machines-here-s-what-broke",
      "in-the-land-of-ai-agents-the-verifiers-are-king",
      "infinite-software-crisis",
      "infra-behind-krea-2-how-to-train-and-serve-at-scale",
      "infra-that-fixes-itself-thanks-to-coding-agents-mahmoud-abdelwahab-railway",
      "insights-from-snorkel-ai-running-azure-ai-infrastructure",
      "introducing-strands-agents-an-open-source-ai-agents-sdk",
      "introduction-to-llm-serving-with-sglang",
      "invisible-users-invisible-interfaces-accelerating-design-iteration-with-ai-simulation",
      "it-admin-for-the-ai-workforce-sarthak-aggarwal-decawork",
      "iterating-on-llm-apps-at-scale-learnings-from-discord",
      "keynote-the-ai-developer-experience-doesn-t-have-to-suck-why-and-how-we-built-modal",
      "keynote-why-people-think-agent-is-a-buzzword-but-it-isn-t",
      "latent-space-paper-club-aiewf-special-edition-test-of-time-deepseek-r1-v3-vibhu-sapra",
      "learning-on-the-job-the-future-of-post-training",
      "lessons-from-building-genai-based-applications-juan-peredo",
      "lessons-from-studying-every-memory-system",
      "let-s-build-an-agent-from-scratch-kam-lasater",
      "let-s-integrate-ai-agents-in-event-sourced-systems",
      "let-s-talk-about-fomat-fear-of-missing-agent-time",
      "level-up-your-automation-game-agentic-workflows-on-vertex-ai",
      "llama-3-at-1-redacted-password-000-tok-s-on-the-sambanova-ai-platform",
      "llamafile-bringing-ai-to-the-masses-with-fast-cpu-inference",
      "llm-pelican-benchmark",
      "llm-quality-optimization-bootcamp",
      "lobster-trap-openclaw-in-containers-from-local-to-k8s-and-back",
      "local-models-trust-control-optimization",
      "long-tasks-and-experienced-open-source-dev-productivity",
      "low-level-technicals-of-llms",
      "luminal-search-based-deep-learning-compilers-joe-fioti",
      "machines-of-buying-selling-grace",
      "magic-editor-under-the-hood-weaving-generative-ai-into-a-billion-user-app",
      "make-your-own-event-sourced-agent-harness-using-stream-processors",
      "making-open-models-10x-faster-and-better-for-modern-application-innovation",
      "mastering-ai-pricing-mayank-pant-stripe",
      "mastering-llm-inference-optimization-from-theory-to-cost-effective-deployment",
      "maximize-gpu-efficiency-with-continuous-profiling-for-gpus",
      "mcp-tasks-async-why-the-heck-aren-t-any-agents-supporting-mcp-tasks-async",
      "measuring-agi-interactive-reasoning-benchmarks",
      "medic-for-apache-spark-first-aid-for-failing-jobs-drasko-profirovic-pinterest",
      "memory-harnesses-for-long-running-research-agents",
      "mentoring-the-machine",
      "milliseconds-to-magic-real-time-workflows-using-the-gemini-live-api-and-pipecat",
      "minimax-m2",
      "mission-critical-evals-at-scale-learnings-from-100-000-medical-decisions",
      "move-fast-break-nothing",
      "multi-model-multimodal-and-multi-agent-innovations-in-azure-ai",
      "navigating-rag-optimization-with-an-evaluation-driven-compass",
      "neil-zeghidour-voice-ai-when-is-the-her-moment",
      "no-more-bad-outputs-with-structured-generation",
      "notion-s-token-town",
      "on-ai-and-knowledge",
      "on-engineering-ai-systems-that-endure-the-bitter-lesson",
      "one-model-to-rule-recommendations-netflix-s-big-bet",
      "one-registry-to-rule-them-all-sonny-merla-mauro-luchetti-mattia-redaelli-quantyca",
      "ontologies-for-agentic-systems",
      "open-challenges-for-ai-engineering",
      "open-rag-eval-rag-evaluation-without-golden-answers",
      "openai-for-vps-of-ai",
      "openclaw-in-your-hand-building-a-physical-ai-terminal-for-local-llm-agents",
      "openrag-an-open-source-stack-for-rag-phil-nash",
      "optimizing-inference-for-voice-models-in-production",
      "personalization-in-the-era-of-llms",
      "pi-coding-agent",
      "pipecat-cloud-enterprise-voice-agents-built-on-open-source",
      "playground-in-prod-optimising-agents-in-production-environments",
      "poc-to-prod-hard-lessons-from-200-enterprise-genai-deployments",
      "practical-graphrag-making-llms-smarter-with-knowledge-graphs",
      "practical-tactics-to-build-reliable-ai-apps-dmitry-kuchin-multinear",
      "privacy-first-enterprise-ai-building-ai-agents-that-never-leave-your-security-boundary",
      "privacy-preserving-intelligence-steve-korshakov-bee-acq-amazon",
      "proactive-agents",
      "production-multi-agent-architecture",
      "production-ready-rag",
      "productionizing-genai-models-lessons-from-the-world-s-best-ai-teams",
      "prompt-engineering-tactics",
      "prototyping-as-leadership-how-a-cto-ships-with-ai-agents",
      "rag-and-the-mongodb-document-model",
      "rag-for-vps-of-ai",
      "real-ai-agents-need-planning-not-just-prompting",
      "real-roi-lessons-from-enterprises-that-have-already-succeeded-with-llms-redacted-username-scale",
      "real-time-experiments-with-an-ai-co-scientist",
      "realtime-conversational-video-with-pipecat-and-tavus-chad-bailey-and-brian-johnson-daily-tavus",
      "recsys-keynote-improving-recommendation-systems-search-in-the-age-of-llms",
      "recursive-model-improvement",
      "remobi-app-don-t-change-your-terminal-workflow-for-mobile",
      "research-to-reality-bringing-frontier-ml-research-to-production",
      "research-to-reality-with-google-deepmind",
      "retrieval-augmented-generation-in-the-wild",
      "revenue-engineering-how-to-price-and-reprice-your-ai-product",
      "reverse-engineering-a-viking-voip-phone-protocol-with-claude-code-boris-starkov-elevenlabs",
      "rise-of-the-ai-architect-clay-bavor-and-alessio-fanelli",
      "rl-for-autonomous-coding-aakanksha-chowdhery-reflection-ai",
      "rl-quantization-agents",
      "road-to-5-million-tokens-breaking-barriers-in-long-context-training-max-ryabinin-together-ai",
      "running-ai-application-in-minutes-quick-start-with-ai-templates",
      "running-gemma-4-on-device-40-tokens-s-on-iphone-with-mlx",
      "running-llms-locally-practical-llm-performance-on-dgx-spark-mozhgan-kabiri-chimeh-nvidia",
      "scaffold-wisely",
      "scaling-ai-agents-without-breaking-reliability",
      "scaling-ai-in-education-a-khanmigo-case-study",
      "scaling-compute-on-context",
      "scaling-enterprise-grade-rag-systems-lessons-from-the-legal-frontier",
      "scaling-reinforcement-learning-lessons-from-trillion-token-deployments-at-fortune-500s",
      "scaling-the-next-paradigm-of-heterogeneous-intelligence",
      "scaling-to-long-horizons",
      "scaling-up-continual-learning",
      "second-order-effects",
      "see-hear-speak-draw",
      "self-training-agents-hermes-agent-hf-traces-skills-mcp-finetuning-merve-noyan-hugging-face",
      "serving-2-million-models-without-melting-scaling-the-hugging-face-hub",
      "serving-voice-ai-at-1-hr-open-source-loras-latency-load-balancing",
      "serving-voice-ai-at-scale-arjun-desai-cartesia-rohit-talluri-aws",
      "shift-left-how-to-become-an-ai-engineer-from-a-full-stack-background",
      "ship-it-building-production-ready-agents",
      "simulation-maxxing-how-nubank-ships-agents-20-faster-with-simulations",
      "skills-at-scale",
      "small-ai-teams-with-huge-impact-vik-paruchuri-datalab",
      "small-bets-big-impact-building-genbi-at-a-fortune-100",
      "software-engineering-is-becoming-plan-and-review",
      "sovereign-escape-velocity-ownership-with-open-models-gus-martins-and-ian-ballantyne-google-deepm",
      "spec-driven-development-agentic-coding-at-faang-scale-and-quality-al-harris-amazon-kiro",
      "special-topics-in-kernels-rl-reward-hacking-in-agents",
      "spreadsheets-are-all-you-need-decoding-the-decoder-llm-without-de-code",
      "state-of-data",
      "state-of-the-union-why-local-why-now",
      "state-space-models-for-realtime-multimodal-intelligence",
      "stop-ai-agent-hallucinations-5-techniques-production-patterns-elizabeth-fuentes-leone-aws",
      "stop-guessing-build-robust-ai-with-layered-cot",
      "stop-ordering-ai-takeout-a-cookbook-for-winning-when-you-build-in-house",
      "stop-renting-your-cognitive-infrastructure",
      "stop-writing-tone-instructions-layer-them",
      "storyteller-building-multi-modal-apps-with-ts-modelfusion-lars-grammel-phd",
      "substrate-launch-the-api-for-modular-ai",
      "taking-reinforcement-learning-cross-datacenter",
      "text-diffusion-brendan-o-donoghue-google-deepmind",
      "text-to-speech-data-preparation-and-fine-tuning-workshop-ronan-mcgovern",
      "the-1-000x-ai-engineer-swyx",
      "the-100-tool-agent-is-a-trap-scaling-with-semantic-routers-and-jit-context",
      "the-3-pillars-of-autonomy-michele-catasta-replit",
      "the-adversarial-path-to-the-personal-assistant",
      "the-agent-ready-web-simplify-user-actions-with-webmcp",
      "the-agentic-ai-engineer",
      "the-agentic-web-and-the-bazaar-era-of-ai-ramesh-raskar-mit-media-lab",
      "the-benchmarks-game-why-it-s-rigged-and-how-you-can-really-win",
      "the-build-operate-divide-bridging-product-vision-and-ai-operational-reality",
      "the-current-state-of-browser-agents",
      "the-demo-i-wish-i-d-had-openai-s-agents-sdk-serverless",
      "the-desktop-frontier-ahmad-osman-osmantic",
      "the-dirty-secret-of-forward-deployed-engineering",
      "the-emerging-skillset-of-wielding-coding-agents",
      "the-end-of-awkward-ai-transcriptions",
      "the-engineer-of-the-future-is-the-person-who-is-able-to-choose-what-is-worth-doing-addy-osmani",
      "the-era-of-unbounded-products-designing-for-multimodal-i-o",
      "the-eyes-are-the-context-window-to-the-soul-how-windsurf-gets-to-know-you",
      "the-factory-that-dreams-39-ai-agents-no-framework",
      "the-friction-is-your-judgment",
      "the-future-is-domain-specific-agents",
      "the-future-of-qwen-a-generalist-agent-model",
      "the-genai-maturity-curve-or-you-probably-don-t-need-fine-tuning",
      "the-geopolitics-of-ai-infrastructure",
      "the-golden-age-of-ai-engineering",
      "the-hidden-costs-of-building-your-own-rag-stack-ofer-mendelevitch-vectara",
      "the-hidden-life-of-embeddings",
      "the-hierarchy-of-needs-for-training-dataset-development",
      "the-infrastructure-for-the-singularity",
      "the-knowledge-graph-mullet-trimming-graphrag-complexity",
      "the-llm-triangle-engineering-principles-for-robust-ai-applications",
      "the-log-is-the-agent",
      "the-maturity-phases-of-running-evals",
      "the-messy-reality-of-scale-synthetic-data-and-pre-training",
      "the-missing-primitive-for-agent-swarms",
      "the-new-application-layer",
      "the-next-unicorns-7-top-ai-startups-from-the-hf0-residency",
      "the-price-of-intelligence-ai-agent-pricing-in-2025",
      "the-production-ai-playbook-deploying-agents-at-enterprise-scale",
      "the-prompt-is-the-platform",
      "the-rag-stack-we-landed-on-after-37-fails",
      "the-rise-of-open-models-in-the-enterprise",
      "the-small-model-infrastructure-nobody-built-so-we-did-filip-makraduli-superlinked",
      "the-state-of-model-routing-nvidia-cognition-openrouter",
      "the-unbearable-lightness-of-agent-optimization",
      "the-unofficial-guide-to-apple-s-private-cloud-compute",
      "the-unreasonable-effectiveness-of-separating-the-task-from-the-model",
      "the-web-browser-is-all-you-need",
      "the-weekend-ai-engineer",
      "the-z-l-continuum-should-ai-engineers-still-read-code",
      "thinner-agents-on-a-smarter-substrate-the-ontology-based-semantic-layer",
      "this-video-was-edited-with-ai-agent-but-how",
      "tlms-tiny-llms-and-agents-on-edge-devices-with-litert-lm",
      "training-an-llm-from-scratch-locally",
      "training-krea-2-what-matters-in-generative-model-training",
      "trends-across-the-ai-frontier",
      "two-roads-to-durable-agents-replay-vs-snapshot-eric-allam-co-founder-trigger-dev",
      "under-5-minutes-to-a-deployed-llm-endpoint-audry-hsu-runpod",
      "understanding-ai-stakes-to-break-production-code",
      "unlock-agent-autonomy-the-runtime-for-ai-native-systems",
      "unlocking-developer-productivity-across-cpu-and-gpu-with-max",
      "unveiling-the-latest-gemma-model-advancements",
      "using-agents-to-build-an-agent-company",
      "using-oss-models-to-build-ai-apps-with-millions-of-users",
      "using-rl-based-agent-to-detect-and-remediate-etl-pipeline-failures",
      "vision-ai-in-2025-peter-robicheaux-roboflow",
      "vision-zero-bugs",
      "voice-ai-your-bot-isn-t-special",
      "voice-in-visuals-out-the-agony-and-the-ecstasy",
      "voiceops-fying-low-latency-intelligence-extraction-from-messy-audio-streams-dippu-kumar-singh",
      "waymo-s-emma-teaching-cars-to-think-jyh-jing-hwang-waymo",
      "what-breaks-when-you-build-ai-under-sovereignty-constraints",
      "what-data-from-20-million-pull-requests-reveal-about-ai-transformation",
      "what-does-it-take-to-build-a-personal-local-private-ai-agent-that-augments-you-deeply",
      "what-every-ai-engineer-needs-to-know-about-gpus",
      "what-it-actually-takes-to-deploy-genai-applications-to-enterprises",
      "what-lies-beneath-the-api-benjamin-cowen-modal",
      "what-rl-means-for-agents",
      "what-the-best-agents-share",
      "what-we-learned-deploying-ai-within-bloomberg-s-engineering-organization",
      "what-we-learned-from-a-year-of-building-with-llms",
      "what-we-learned-from-using-llms-in-pinterest",
      "when-will-the-benchmaxxing-plague-end",
      "why-building-eval-platforms-is-hard",
      "why-chatgpt-keeps-interrupting-you",
      "why-large-tiny-lms-agents-on-edge-robotics",
      "why-mlx",
      "why-more-context-makes-your-agent-dumber-and-what-to-do-about-it",
      "why-we-don-t-need-more-data-centers",
      "why-you-should-care-about-ai-interpretability",
      "why-your-agents-need-decision-traces-not-just-documents-zach-blumenfeld-neo4j",
      "why-your-ai-ux-is-broken-and-it-s-not-the-model-s-fault",
      "writing-principles-for-task-tuned-prompt-engineering",
      "wtf-do-people-use-open-models-for",
      "you-might-not-need-50-diffusion-steps",
      "your-agent-architecture-has-a-half-life-of-6-months",
      "your-agent-evolved-your-evals-didn-t",
      "your-agent-failed-in-prod-good-luck-reproducing-it",
      "your-attention-is-the-bottleneck-not-your-agents-zack-proser-workos",
      "your-fine-tuned-model-is-tech-debt-a-50x-roi-house-of-cards",
      "your-llm-deception-monitor-is-broken-the-fix-is-in-the-training-data-sachin-kumar-lexisnexis",
      "your-moat-is-your-data-model",
      "your-realtime-ai-is-ngmi-sean-dubois-openai-kwindla-hultman-kramer-daily",
      "youtube-0371da799e2c154968bd",
      "youtube-04e58fb37fee786d215c",
      "youtube-06ea394305681127392c",
      "youtube-08a1f0e2a4a4fcd18124",
      "youtube-10859790e6e7a182958f",
      "youtube-187a1637c7bf015de670",
      "youtube-1f17b815d5c9afa8a898",
      "youtube-276f859bcfe822e2f467",
      "youtube-31b0f9892fe237c0e318",
      "youtube-3837a774b03ab06f9e44",
      "youtube-3d84e7afa2e44f21bd3b",
      "youtube-3fa5dba9535b300e783d",
      "youtube-405d639eef1c391f04dd",
      "youtube-40c0a903db79223717ca",
      "youtube-422e1925484113104e43",
      "youtube-433b6c126cf30f52e304",
      "youtube-48ac285e9ace1ac33b2e",
      "youtube-502af1124ddc0a68f915",
      "youtube-52bf7b08bb7d4742ea8f",
      "youtube-54f48d11d6a791984276",
      "youtube-561cf6576bf9b94c7ffc",
      "youtube-61ead8e20b85bf99f1db",
      "youtube-649d40b14a294ff4e1a6",
      "youtube-6c0f6fd219675c3f58a1",
      "youtube-6e9bac1afcd6ff1d78de",
      "youtube-6fcfb5ae5a2bdc32a5b2",
      "youtube-708927ce515a512a39c4",
      "youtube-7be948df4cdb3236b72a",
      "youtube-814050f1b1446db25258",
      "youtube-835051dca330c322ddc9",
      "youtube-86707775be4ad567144b",
      "youtube-8acee04c409cbd23da4e",
      "youtube-8c138db6bc945eee916c",
      "youtube-8e0d9ceebdb22d4f0587",
      "youtube-93a23ec02e7457554f41",
      "youtube-9b2aaaa07c15a78315e7",
      "youtube-9c20be05456532b413c4",
      "youtube-9c3fa3c70590bbbf1cf1",
      "youtube-9d4469f54600ef8dcf68",
      "youtube-a1ed0627f76116a7ee43",
      "youtube-a726a9581120ceba75a8",
      "youtube-b006941a973d92c44b1c",
      "youtube-b259ecf426664455245f",
      "youtube-b3b88f589f2be887e368",
      "youtube-b4ae2559dea247af03e3",
      "youtube-b90ae020cfee1febfd51",
      "youtube-b92e3d4c3c3481044117",
      "youtube-bb69af9cef90a68feb53",
      "youtube-be37558ed6795ac3da2d",
      "youtube-bf4c09081cbc8efe51dd",
      "youtube-bfc00573674178488871",
      "youtube-c050c93126b379b95230",
      "youtube-c2212e3ab08fd37597bd",
      "youtube-ce30eba4cf31b984a4f0",
      "youtube-cf4e79c20a282ea11e37",
      "youtube-d24f548d7a232f37438b",
      "youtube-d8c3395d74bf5f529944",
      "youtube-d9470a6ec6f5a0a3b05f",
      "youtube-db094c6fd6d3fee927bb",
      "youtube-dd16943baa0178dffc78",
      "youtube-e75ead05705dd207eba6",
      "youtube-eb2fab9b37bd143966d0",
      "youtube-fc0d490e2e5e8a0d2627",
      "z-ai-glm-4-6-what-we-learned-from-100-million-open-source-downloads-yuxuan-zhang-z-ai"
    ]
  },
  {
    "slug": "rag-and-knowledge",
    "name": "Retrieval-augmented generation",
    "url": "https://ai.engineer/topics/rag-and-knowledge",
    "description": "Talks whose official titles focus on retrieval, GraphRAG, ontologies, or structured model outputs.",
    "talkSlugs": [
      "1-ai-guardrails-the-unreasonable-effectiveness-of-finetuned-modernberts-diego-carpentero",
      "12-factor-agents",
      "120k-players-in-a-week-lessons-from-the-first-viral-clip-app",
      "2026-the-year-the-ide-died",
      "360brew-llm-based-personalized-ranking-and-recommendation-hamed-firooz-and-maziar-sanjabi-linked",
      "a-practical-guide-to-efficient-ai",
      "a-practitioner-s-guide-to-graphs-tim-ainge-good-collective",
      "a-song-of-types-and-agents",
      "accelerate-your-ai-journey-with-azure-ai-model-catalog",
      "active-graph-agent-runtime-babyagi-4",
      "agent-evals-finally-with-the-map",
      "agent-skills",
      "agentic-graphrag-ai-s-logical-edge",
      "agentic-graphrag-simplifying-retrieval-across-structured-unstructured-data-zach-blumenfeld",
      "agentic-search-for-context-engineering",
      "agentic-security-permissions-provenance-and-the-agent-supply-chain",
      "agents-are-built-at-the-fringe-getting-from-90-to-100",
      "agents-are-robots-too-what-self-driving-taught-me-about-building-agents-jesse-hu-abundant",
      "agents-building-agents",
      "agents-in-production-how-opengov-built-and-scaled-og-assist",
      "agents-need-feature-flags",
      "agents-need-receipts-not-more-tool-calls",
      "agents-reported-thousands-of-bugs-how-many-were-real-ian-butler-and-nick-gregory",
      "ai-agents-for-performance-ship-faster-pay-less",
      "ai-agents-meet-test-driven-development",
      "ai-driven-multi-document-correlation-for-enterprise-financial-compliance-and-fraud-detection",
      "ai-engineering-101",
      "ai-engineering-201-inference",
      "ai-engineering-201-the-rest-of-the-owl",
      "ai-engineering-with-the-google-gemini-2-5-model-family",
      "ai-music-generation-from-prompt-to-production",
      "ai-on-your-lakehouse-context-comes-in-shapes-not-queries",
      "ai-pipelines-and-agents-in-pure-typescript-with-mastra-ai",
      "ai-red-teaming-agent-azure-ai-foundry-nagkumar-arkalgud-keiji-kanazawa-microsoft",
      "ai-sdk-v6",
      "ai-system-design-from-idea-to-production",
      "ai-tools-for-forward-deployed-engineering",
      "alphalab-autonomous-multi-agent-research-across-optimization-domains-with-frontier-llms-brendan",
      "amp-code-next-generation-ai-coding",
      "analyzing-10-000-sales-calls-with-ai-in-2-weeks",
      "anchoring-enterprise-genai-with-knowledge-graphs",
      "anthropic-s-cca-exam-as-a-field-guide-for-agentic-engineering",
      "any-to-any-building-native-multimodal-agents",
      "architecting-agent-memory-principles-patterns-and-best-practices",
      "architecting-and-testing-controllable-agents",
      "are-mcps-overhyped-a-rant-about-mcps",
      "automating-escrow-with-usdc-and-ai",
      "autonomous-agents-for-scientific-tasks-sina-shahandeh-radicait",
      "benchmarking-semantic-code-retrieval-on-claude-code",
      "benchmarks-the-good-the-bad-and-the-ugly",
      "beyond-apis-how-ai-web-agents-are-automating-the-long-tail-of-knowledge-work",
      "beyond-conversation-why-documents-transform-natural-language-into-code",
      "beyond-static-intelligence-evaluating-continual-learning",
      "blender-mcp-and-the-future-of-creative-tools-siddharth-ahuja",
      "books-reimagined-ai-to-create-new-experiences-for-things-you-know-ukasz-gandecki-thebrain-pro",
      "botdojo-launch-enhancing-ai-assistants-with-evaluations-and-synthetic-data",
      "bounded-autonomy-between-free-will-and-determinism",
      "break-it-til-you-make-it-building-the-self-improving-stack-for-ai-agents",
      "build-a-prompt-learning-loop",
      "build-deploy-ai-powered-apps",
      "build-deploy-ai-powered-apps-18cc1b",
      "build-dynamic-products-and-stop-the-ai-sideshow",
      "build-enterprise-generative-ai-apps-using-llama-3-at-1-000-tokens-s-on-the-sambanova-ai-platform",
      "build-for-the-memo-not-the-demo-notes-from-200-investment-committees",
      "build-systems-not-code",
      "build-the-ai-gtm-agent-that-knows-the-buyer-before-the-first-message",
      "build-your-first-demand-driven-context-base-let-ai-agents-tell-you-what-they-need",
      "build-your-own-deep-research-agent-technical-writer",
      "building-a-smarter-ai-agent-with-neural-rag",
      "building-agent-interfaces-lessons-from-chrome-devtools-mcp-for-agents",
      "building-agentic-applications-with-heroku-managed-inference-and-agents-julian-duque-and-anush-ds",
      "building-agents-is-trivial-now-context-is-the-next-frontier",
      "building-agents-with-mcp",
      "building-ai-agents-that-actually-automate-knowledge-work",
      "building-alice-s-brain-an-ai-sales-rep-that-learns-like-a-human-sherwood-satwik-11x",
      "building-an-agentic-platform",
      "building-an-ai-assistant-that-makes-phone-calls",
      "building-blocks-for-llm-systems-products",
      "building-context-aware-reasoning-applications-with-langchain-and-langsmith",
      "building-conversational-ai-agents-thor-schaeff-elevenlabs",
      "building-deterministic-infrastructure-for-non-deterministic-ai-agents",
      "building-durable-production-ready-agents-with-openai-sdk-and-temporal",
      "building-efficient-hybrid-context-query-for-llm-grounding",
      "building-enterprise-llm-agents-that-work",
      "building-great-agent-skills",
      "building-in-the-gemini-era-kat-kampf-ammaar-reshi-google-deepmind",
      "building-interactive-uis-in-vs-code-with-mcp-apps-marlene-mhangami-liam-hampton-github",
      "building-metrics-that-actually-work-david-karam-pi-labs",
      "building-multi-agent-systems-with-finite-state-machines",
      "building-multimodal-ai-agents-from-scratch",
      "building-reactive-ai-apps",
      "building-redacted-username-gergely-orosz-simon-eskildsen",
      "building-reliable-agentic-systems",
      "building-reliable-support-agents-using-the-effect-typescript-library-michael-fester",
      "building-security-around-ml",
      "building-self-coding-agents",
      "building-sota-open-weights-tool-use-the-command-r-family",
      "building-the-platform-for-agent-coordination",
      "building-trust-in-enterprise-ai-evaluating-domain-specific-llms-for-real-world-financial-scenari",
      "building-voice-agents-with-openai",
      "building-with-anthropic-s-claude-the-prompt-doctor-is-in",
      "building-your-own-secure-ai-workflows-human-in-the-loop-automation-with-n8n",
      "buy-now-maybe-pay-later-dealing-with-prompt-tax-while-staying-at-the-frontier-andrew-thompson",
      "bypassing-the-multimodal-tax-framework-free-hybrid-rag-raw-sql-rrf-and-live-ui-telemetry",
      "can-oncology-workflows-run-without-human-touch-anant-shankhdhar-risa-labs",
      "case-study-deep-dive-telemedicine-support-agents-with-langgraph-mcp",
      "citation-needed-provenance-for-llm-built-knowledge-graphs",
      "claude-agent-sdk-workshop",
      "claude-fable-claude-tag-and-anthropic-s-culture-cat-wu-thariq-shihipar-ft-simon-willison",
      "claude-for-long-horizon-tasks",
      "claude-plays-minecraft-introducing-a-real-world-serverless-ai-to-a-virtual-world",
      "cognitive-shield-real-time-real-smart-rachna-srivastava",
      "cohere-for-vps-of-ai",
      "combine-skills-and-mcp-to-close-the-context-gap",
      "compilers-in-the-age-of-llms",
      "compute-system-design-for-next-generation-frontier-models",
      "connecting-the-dots-with-context-graphs",
      "context-engineering-connecting-the-dots-with-graphs-stephen-chin-neo4j",
      "context-engineering-for-complex-codebases",
      "context-engineering-in-2026-compaction-memory-cost",
      "context-graphs-for-explainable-decision-aware-ai-agents",
      "context-is-the-new-code",
      "context-platform-engineering-to-reduce-token-anxiety-val-bercovici-and-callan-fox-weka",
      "convex-launch",
      "crabrag-why-automated-assistants-need-graph-memory-not-more-tokens",
      "creating-agents-that-co-create",
      "creating-and-scaling-your-own-custom-copilots-with-azure-ai-studio",
      "customized-production-ready-inference-with-open-source-models-dmytro-dima-dzhulgakov",
      "data-is-your-differentiator-building-secure-and-tailored-ai-systems",
      "data-readiness-is-a-myth-reliable-ai-with-an-agentic-semantic-layer-anushrut-gupta-promptql",
      "decoding-mistral-ai-s-large-language-models",
      "deepswe-a-contamination-resistant-coding-benchmark-james-shi-datacurve",
      "designing-ai-intensive-applications",
      "develop-at-idea-velocity",
      "developer-experience-in-the-age-of-ai-coding-agents",
      "devin-2-0-and-the-future-of-swe",
      "dispatch-from-the-future-building-an-ai-native-company-dan-shipper-every-ai-i",
      "does-genai-belong-to-data-scientists",
      "domain-adaptation-and-fine-tuning-for-domain-specific-llms",
      "don-t-build-slop-4-levels-of-ai-agent-maturity",
      "don-t-just-slap-on-a-chatbot-building-ai-that-works-before-you-ask",
      "don-t-ship-skills-without-evals",
      "dspy-the-end-of-prompt-engineering",
      "effective-agent-design-patterns-in-production",
      "effective-ai-agents-need-data-flywheels-not-the-next-biggest-llm-sylendran-arunagiri-nvidia",
      "embeddings-are-stunting-agents-how-codeium-breaks-through-the-ceiling-for-retrieval",
      "engineering-better-evals-scalable-llm-evaluation-pipelines-that-work",
      "ensure-ai-agents-work-evaluation-frameworks-for-scaling-success",
      "enterprise-agents-have-a-structure-problem-ishita-daga-tesla",
      "enterprise-deep-research-the-next-killer-app-for-enterprise-ai-ofer-mendelevitch-vectara",
      "evals-are-not-unit-tests",
      "evals-driven-development-engineering-a-mental-health-ai-coach-ethically-safely",
      "evals-workshop-mastering-ai-evaluation-from-playground-to-production",
      "evaluating-ai-search-a-practical-framework-for-augmented-ai-systems",
      "every-company-should-have-a-brain-garry-tan-y-combinator",
      "every-harness-will-become-a-claw",
      "everything-i-learned-training-frontier-small-models",
      "evolving-claude-apis-for-agents",
      "eyelevel-launch-your-rag-is-tripping-here-s-the-real-reason-why",
      "fast-models-need-slow-developers",
      "feedback-loops-are-all-you-need",
      "fighting-slop-with-slop",
      "first-steps-toward-automated-ai-research",
      "five-hard-earned-lessons-about-evals",
      "forget-rag-pipelines-build-production-ready-ai-agents-in-15-minutes",
      "form-factors-for-your-new-ai-coworkers",
      "forward-deployed-engineering-101",
      "from-agent-traces-to-agent-simulations-rustem-feyzkhanov-snorkel-ai",
      "from-arc-to-dia-lessons-learned-in-building-ai-browser",
      "from-chaos-to-choreography-multi-agent-orchestration-patterns-that-actually-work-sandipan-bhaumi",
      "from-copilot-to-colleague-building-trustworthy-productivity-agents-for-high-stakes-work",
      "from-software-developer-to-ai-engineer",
      "from-systems-of-record-to-systems-of-context",
      "from-writing-code-to-designing-systems-how-the-developer-role-is-changing-chris-noring-microsoft",
      "frontier-ai-at-home-literally",
      "frontier-results-on-device-rl-nabors-arize",
      "full-workshop-setting-yourself-up-for-success-jason-liu-openai-codex",
      "fun-stories-from-building-openrouter-and-where-all-this-is-going",
      "function-calling-is-all-you-need",
      "future-of-knowledge-assistants",
      "gemini-nano-on-device-florina-muntenescu-oli-gaymond-google-deepmind",
      "gemma-4-deep-dive-cassidy-hardin-google-deepmind",
      "git-push-get-an-ai-api",
      "giving-a-voice-to-ai-agents",
      "going-beyond-rag-extended-mind-transformers",
      "gpt-web-app-generator-10-000-apps-created-in-a-month-matija-sosic",
      "graph-intelligence-enhance-reasoning-and-retrieval-using-graph-analytics",
      "graphrag",
      "graphrag-methods-to-create-optimized-llm-context-windows-for-retrieval-jonathan-larson-microsoft",
      "guardrails-first-engineering-member-facing-health-ai",
      "guide-verify-solve-the-engineering-discipline-agentic-development-demands",
      "harness-engineering",
      "harnessing-the-power-of-llms-locally",
      "hasura-launch-realtime-data-connectivity-for-ai",
      "healthcare-s-agent-bytecode-x12-as-the-harness-for-ai-agents",
      "how-agent-o11y-differs-from-traditional-o11y",
      "how-blackrock-builds-custom-knowledge-apps-at-scale",
      "how-claude-code-works",
      "how-deep-research-works",
      "how-google-deepmind-is-researching-the-next-frontier-of-ai-for-gemini-raia-hadsell-vp-of-researc",
      "how-i-automate-my-own-job-at-hugging-face-using-agents",
      "how-instacart-transformed-its-search-and-discovery-using-an-llm-driven-approach",
      "how-intuit-uses-llms-to-explain-taxes-to-millions-of-taxpayers",
      "how-juries-and-librarians-can-solve-gtm-s-ai-trust-problem",
      "how-kepler-built-verifiable-ai-for-financial-services",
      "how-llms-work-for-web-devs-gpt-in-600-lines-of-vanilla-js",
      "how-to-build-agents-that-run-for-hours-without-losing-the-plot",
      "how-to-build-ai-agents-that-actually-work",
      "how-to-build-an-ai-strategy-that-fails",
      "how-to-build-enterprise-aware-agents",
      "how-to-build-trustworthy-ai",
      "how-to-build-world-class-ai-products-sarah-sachs-notion-and-carlos-esteban-braintrust",
      "how-to-construct-domain-specific-llm-evaluation-systems",
      "how-to-defend-your-sites-from-ai-bots",
      "how-to-improve-your-agents-academic-lit-review",
      "how-to-improve-your-vibe-coding-ian-butler",
      "how-to-look-at-your-data-what-to-look-for-how-to-measure",
      "how-we-hacked-yc-spring-2025-batch-s-ai-agents",
      "how-we-scaled-500m-ai-agents-in-production-with-2-engineers",
      "how-we-solved-context-management-in-agents-sally-ann-delucia",
      "how-we-taught-agents-to-use-good-retrieval-hanna-lichtenberg-mixedbread-ai",
      "how-web-data-infrastructure-powers-the-next-generation-of-ai",
      "html-is-all-you-need-for-agents-to-make-graphics",
      "human-seeded-evals-samuel-colvin-pydantic",
      "hybridrag-a-fusion-of-graph-and-vector-retrieval-to-enhance-data-interpretation",
      "hypermode-launch",
      "hyperspace-more-nodes-is-all-you-need",
      "i-gave-an-ai-agent-the-keys-to-my-life-here-s-what-happened",
      "i-run-a-fleet-of-ai-agents-across-three-machines-here-s-what-broke",
      "i-ve-never-seen-anything-scarier-than-an-llm-with-tool-calls-erik-meijer-aka-headinthebox",
      "improving-agents-is-a-data-mining-problem",
      "information-retrieval-from-the-ground-up",
      "infra-that-fixes-itself-thanks-to-coding-agents-mahmoud-abdelwahab-railway",
      "insights-from-snorkel-ai-running-azure-ai-infrastructure",
      "intelligence-continual-learning-expertise",
      "intro-to-graphrag",
      "introducing-strands-agents-an-open-source-ai-agents-sdk",
      "iterating-on-llm-apps-at-scale-learnings-from-discord",
      "judge-the-judge-building-llm-evaluators-that-actually-work-with-gepa-mahmoud-mabrouk-agenta-ai",
      "judging-llms",
      "just-do-it-let-your-tools-think-for-themselves-robert-chandler",
      "keynote-why-people-think-agent-is-a-buzzword-but-it-isn-t",
      "knowledge-graphs-graphrag-techniques-for-building-effective-genai-applications",
      "knowledge-graphs-in-litigation-agents-tom-smoker-whyhow-ai",
      "layering-every-technique-in-rag-one-query-at-a-time",
      "learned-execution-graphs-for-anomaly-detection-drift-in-apis-ritvik-pandya-jp-morgan-chase",
      "lessons-from-building-genai-based-applications-juan-peredo",
      "lessons-from-building-linkedin-s-genai-platform",
      "lessons-from-studying-every-memory-system",
      "let-s-build-an-agent-from-scratch-kam-lasater",
      "let-s-integrate-ai-agents-in-event-sourced-systems",
      "llama-3-at-1-redacted-password-000-tok-s-on-the-sambanova-ai-platform",
      "llm-codegen-fails-and-how-to-stop-em",
      "llm-evals-that-work-irl",
      "llm-knowledge-bases-a-practical-guide",
      "llm-observability-evaluation-experimentation-platform-dat-ngo-arize-ai",
      "llm-pelican-benchmark",
      "llm-safeguards-security-privacy-compliance-anti-hallucination",
      "llm-scientific-reasoning-how-to-make-ai-capable-of-nobel-prize-discoveries",
      "low-level-technicals-of-llms",
      "luminal-search-based-deep-learning-compilers-joe-fioti",
      "making-open-models-10x-faster-and-better-for-modern-application-innovation",
      "mastering-engineering-flow-with-windsurf",
      "maximize-gpu-efficiency-with-continuous-profiling-for-gpus",
      "mcp-agent-fine-tuning-workshop-ronan-mcgovern",
      "mcp-apps-primitives-discovery-and-the-future-of-software",
      "mcp-is-all-you-need",
      "medic-for-apache-spark-first-aid-for-failing-jobs-drasko-profirovic-pinterest",
      "memory-harnesses-for-long-running-research-agents",
      "memory-masterclass-make-your-ai-agents-remember-what-they-do-mark-bain-aius",
      "mentoring-the-machine",
      "mergeable-by-default-building-the-context-engine-to-save-time-and-tokens",
      "mind-the-gap-in-your-agent-observability",
      "minimax-m2",
      "model-maxxing-rft-dpo-sft-fine-tuning-with-openai-ilan-bigio-openai",
      "multi-agent-ai-and-network-knowledge-graphs-for-change-management-and-network-testing",
      "my-ai-thinks-i-m-eating-my-feelings-and-other-nutritional-insights",
      "navigating-challenges-and-technical-debt-in-llms-deployment",
      "navigating-rag-optimization-with-an-evaluation-driven-compass",
      "new-york-times-connections-a-case-study-on-nlp-in-word-games",
      "no-code-fine-tuning-mark-hennings",
      "no-more-bad-outputs-with-structured-generation",
      "on-ai-and-knowledge",
      "on-curiosity-sharif-shameem-lexica",
      "on-engineering-ai-systems-that-endure-the-bitter-lesson",
      "one-developer-two-dozen-agents-zero-alignment-why-we-need-collaborative-ai-engineering",
      "one-model-to-rule-recommendations-netflix-s-big-bet",
      "ontologies-for-agentic-systems",
      "open-challenges-for-ai-engineering",
      "open-questions-for-ai-engineering",
      "open-rag-eval-rag-evaluation-without-golden-answers",
      "open-source-is-dead-long-live-open-source",
      "openai-for-vps-of-ai",
      "openclaw-in-your-hand-building-a-physical-ai-terminal-for-local-llm-agents",
      "openllmetry-is-all-you-need",
      "openrag-an-open-source-stack-for-rag-phil-nash",
      "optimizing-llms-in-insurance-with-dspy-beyond-manual-tuning",
      "paperclip-open-source-human-control-plane-for-ai-labor-dotta",
      "perception-agents-antje-barth-amazon-agi-lab",
      "perceptual-evaluations-evals-for-aesthetics-diego-rodriguez-krea-ai",
      "persona-engineering-a-field-guide-to-ai-synthetic-personas-ishan-anand-insightsciences-ai",
      "personalization-in-the-era-of-llms",
      "pipecat-cloud-enterprise-voice-agents-built-on-open-source",
      "playground-in-prod-optimising-agents-in-production-environments",
      "poc-to-prod-hard-lessons-from-200-enterprise-genai-deployments",
      "practical-graphrag-making-llms-smarter-with-knowledge-graphs",
      "practical-tactics-to-build-reliable-ai-apps-dmitry-kuchin-multinear",
      "pragmatic-ai-with-typechat",
      "production-evals-for-agentic-ai-systems",
      "production-rag-agents",
      "production-ready-rag",
      "prompt-engineering-ai-red-teaming",
      "prompt-engineering-is-dead",
      "prompt-engineering-tactics",
      "pydantic-is-still-all-you-need",
      "rag-and-the-mongodb-document-model",
      "rag-at-scale-production-ready-genai-apps-with-azure-ai-search",
      "rag-evaluation-is-broken-here-s-why-and-how-to-fix-it",
      "rag-for-vps-of-ai",
      "rag-in-2025-state-of-the-art-and-the-road-forward",
      "rag-is-dead-right-kuba-rogut-redacted-username",
      "ralph-loops-build-dumb-ai-loops-that-ship",
      "real-ai-agents-need-planning-not-just-prompting",
      "real-roi-lessons-from-enterprises-that-have-already-succeeded-with-llms-redacted-username-scale",
      "realtime-multiplayer-automation-and-you",
      "recsys-keynote-improving-recommendation-systems-search-in-the-age-of-llms",
      "recursive-coding-agents",
      "reinforcement-learning-without-verifiable-rewards-will-brown-prime-intellect",
      "replacing-12k-loc-with-a-200-loc-skill",
      "research-to-reality-bringing-frontier-ml-research-to-production",
      "retrieval-augmented-generation-in-the-wild",
      "rlm-recursive-language-models-for-large-codebases",
      "road-to-5-million-tokens-breaking-barriers-in-long-context-training-max-ryabinin-together-ai",
      "running-ai-application-in-minutes-quick-start-with-ai-templates",
      "running-llms-locally-practical-llm-performance-on-dgx-spark-mozhgan-kabiri-chimeh-nvidia",
      "safety-and-security-for-code-executing-agents",
      "scaffold-wisely",
      "scaling-ai-agents-without-breaking-reliability",
      "scaling-compute-on-context",
      "scaling-enterprise-grade-rag-systems-lessons-from-the-legal-frontier",
      "scaling-github-for-your-agents-sam-morrow-github",
      "securing-agents-with-open-standards",
      "security-firewall-for-agents",
      "self-driving-products-product-signals-to-pull-requests",
      "self-training-agents-hermes-agent-hf-traces-skills-mcp-finetuning-merve-noyan-hugging-face",
      "semantic-blindness-500-000-sensors-confused-an-llm-raahul-singh-vanc-levstik-phaidra",
      "serving-2-million-models-without-melting-scaling-the-hugging-face-hub",
      "shift-left-how-to-become-an-ai-engineer-from-a-full-stack-background",
      "ship-real-agents-hands-on-evals-for-agentic-applications",
      "shipping-ai-that-works-an-evaluation-framework-for-pms",
      "shipping-an-enterprise-voice-ai-agent-in-100-days",
      "simulation-maxxing-how-nubank-ships-agents-20-faster-with-simulations",
      "skill-issue-lessons-from-skilling-up-coding-agents-to-use-langfuse",
      "skills-are-new-features-building-skill-centric-harness-yogendra-miraje-factset",
      "skills-are-the-new-sdks",
      "skills-at-scale",
      "small-ai-teams-with-huge-impact-vik-paruchuri-datalab",
      "small-bets-big-impact-building-genbi-at-a-fortune-100",
      "specifications-are-the-new-code",
      "spreadsheets-are-all-you-need-decoding-the-decoder-llm-without-de-code",
      "state-of-startups-and-ai-2025",
      "state-of-the-union-why-local-why-now",
      "state-space-models-for-realtime-multimodal-intelligence",
      "stateful-agents-full-workshop-with-redacted-username-packer-of-letta-and-memgpt",
      "stateful-environments-for-vertical-agents-josh-purtell-synth-labs",
      "stop-ai-agent-hallucinations-5-techniques-production-patterns-elizabeth-fuentes-leone-aws",
      "stop-babysitting-your-agents-building-a-context-engine-for-mergeable-code",
      "stop-burning-tokens-why-self-improvement-needs-domain-expertise-first-annabell-schafer-langfuse",
      "stop-guessing-build-robust-ai-with-layered-cot",
      "stop-renting-your-cognitive-infrastructure",
      "stop-using-rag-as-memory",
      "stop-writing-tone-instructions-layer-them",
      "storyteller-building-multi-modal-apps-with-ts-modelfusion-lars-grammel-phd",
      "strategies-for-llm-evals-guidellm-lm-eval-harness-openai-evals-workshop-taylor-jordan-smith",
      "structured-llm-outputs-with-pydantic",
      "structuring-the-unstructured-advanced-document-parsing-for-ai-workflows",
      "stuffing-context-is-not-memory-updating-weights-is",
      "substrate-launch-the-api-for-modular-ai",
      "supabase-vector-the-postgres-vector-database",
      "taming-rogue-ai-agents-with-observability-driven-evaluation",
      "task-fidelity-scaling-laws-kobie-crawford-snorkel-ai",
      "teaching-gemini-to-speak-youtube-adapting-llms-for-video-recommendations-to-2b-dau",
      "the-100-tool-agent-is-a-trap-scaling-with-semantic-routers-and-jit-context",
      "the-2025-ai-engineering-report-barr-yaron-amplify-partners",
      "the-a-to-z-of-building-ai-agents",
      "the-adversarial-path-to-the-personal-assistant",
      "the-agent-native-company",
      "the-agent-ready-web-simplify-user-actions-with-webmcp",
      "the-agentic-web-and-the-bazaar-era-of-ai-ramesh-raskar-mit-media-lab",
      "the-ai-pivot-with-chris-white-of-prefect-bryan-bischof-of-hex",
      "the-billable-hour-is-dead-long-live-the-billable-hour",
      "the-coherence-trap-why-llms-feel-smart-but-aren-t-thinking",
      "the-cure-for-the-vibe-coding-hangover",
      "the-dirty-secret-of-forward-deployed-engineering",
      "the-end-of-apps",
      "the-era-of-compound-engineering",
      "the-era-of-unbounded-products-designing-for-multimodal-i-o",
      "the-factory-that-dreams-39-ai-agents-no-framework",
      "the-friction-is-your-judgment",
      "the-future-of-evals-ankur-goyal-braintrust",
      "the-future-of-work",
      "the-genai-maturity-curve-or-you-probably-don-t-need-fine-tuning",
      "the-great-loops-debate-dex-horthy-geoff-huntley-ian-livingstone-greg-pstrucha-moderated-by-allie",
      "the-hidden-costs-of-building-your-own-rag-stack-ofer-mendelevitch-vectara",
      "the-hidden-life-of-embeddings",
      "the-hierarchy-of-needs-for-training-dataset-development",
      "the-intelligent-interface",
      "the-knowledge-graph-mullet-trimming-graphrag-complexity",
      "the-llm-triangle-engineering-principles-for-robust-ai-applications",
      "the-log-is-the-agent",
      "the-making-of-devin",
      "the-miranda-hypothesis-how-hamilton-the-musical-poisoned-your-persona-evals",
      "the-missing-layer-design-taste-in-ai-agents",
      "the-model-isn-t-wrong-you-re-just-bad-at-prompting",
      "the-next-unicorns-7-top-ai-startups-from-the-hf0-residency",
      "the-prompt-is-still-a-punch-card",
      "the-rag-stack-we-landed-on-after-37-fails",
      "the-rise-of-caas-context-as-a-service-for-agentic-ai",
      "the-small-model-infrastructure-nobody-built-so-we-did-filip-makraduli-superlinked",
      "the-state-of-ai-powered-search-and-retrieval",
      "the-state-of-mcp-observability-observable-tools-alex-volkov-and-benjamin-eckel-weights-biases-an",
      "the-state-of-model-routing-nvidia-cognition-openrouter",
      "the-unbearable-lightness-of-agent-optimization",
      "the-unreasonable-effectiveness-of-prompt-learning",
      "the-unreasonable-effectiveness-of-separating-the-task-from-the-model",
      "the-voice-first-ai-overlay-designing-conversational-co-pilots",
      "thinner-agents-on-a-smarter-substrate-the-ontology-based-semantic-layer",
      "tool-calling-is-not-just-plumbing-for-ai-agents",
      "training-albatross-an-expert-finance-llm",
      "trunk-tools-launch-disrupting-the-15-trillion-construction-industry-with-autonomous-agents",
      "trust-but-verify",
      "trust-but-verify-high-fidelity-reasoning-in-agentic-workflows",
      "turn-10-994-notes-into-your-agents-memory",
      "two-roads-to-durable-agents-replay-vs-snapshot-eric-allam-co-founder-trigger-dev",
      "understanding-ai-stakes-to-break-production-code",
      "useful-general-intelligence",
      "user-signal-die-at-the-retrieval-boundary",
      "using-ai-to-build-an-infinite-game",
      "using-llms-to-secure-source-code",
      "using-rl-based-agent-to-detect-and-remediate-etl-pipeline-failures",
      "vector-search-benchmark-eting",
      "vercel-ai-sdk-masterclass-from-fundamentals-to-deep-research",
      "vibe-coding-at-scale-customizing-ai-assistants-for-enterprise-environments",
      "video-has-no-memory-here-s-how-we-built-one",
      "voiceops-fying-low-latency-intelligence-extraction-from-messy-audio-streams-dippu-kumar-singh",
      "voicevision-rag-integrating-visual-document-intelligence-with-voice-response",
      "we-vetted-2-000-ai-skills-before-they-reached-developers",
      "wearing-the-agent-engineering-a-family-and-friends-personal-agent-from-group-chats-to-glasses",
      "what-do-models-still-suck-at",
      "what-if-the-harness-mattered-more-than-the-model-aditya-bhargava-etsy",
      "what-s-new-from-anthropic-and-what-s-next",
      "what-the-best-agents-share",
      "what-we-learned-from-using-llms-in-pinterest",
      "when-agents-meet-physical-data-the-other-physics-of-agent-harnesses",
      "when-all-context-matters-extended-cache-augmented-generation-ecag",
      "when-vectors-break-down-graph-based-rag-for-dense-enterprise-knowledge",
      "when-will-the-benchmaxxing-plague-end",
      "which-jobs-can-be-replaced-today",
      "why-and-how-you-need-to-sandbox-ai-generated-code-harshil-agrawal-cloudflare",
      "why-building-eval-platforms-is-hard",
      "why-can-t-anyone-answer-questions-about-the-business",
      "why-chatgpt-keeps-interrupting-you",
      "why-eval-is-the-next-great-compute-primitive-sunil-pai-matt-carey-cloudflare",
      "why-more-context-makes-your-agent-dumber-and-what-to-do-about-it",
      "why-senior-engineers-struggle-to-build-ai-agents",
      "why-should-anyone-care-about-evals",
      "why-the-best-ai-agents-are-built-without-frameworks-primitives-over-frameworks",
      "why-we-killed-our-multi-agent-pipeline-subbiah-sethuraman-and-abhilash-asokan-zs-associates",
      "why-your-agent-disagrees-with-itself-and-what-to-do-about-it",
      "why-your-agent-s-brain-needs-a-playbook-practical-wins-from-using-ontologies",
      "why-your-agents-need-decision-traces-not-just-documents-zach-blumenfeld-neo4j",
      "wisdom-driven-knowledge-augmented-generation-at-scale",
      "writing-principles-for-task-tuned-prompt-engineering",
      "wtf-is-the-context-layer-the-missing-infrastructure-for-production-agents",
      "your-agent-evolved-your-evals-didn-t",
      "your-agent-is-wasting-tokens-and-you-don-t-know-it",
      "your-agent-s-biggest-lie-i-searched-the-web",
      "your-coding-agent-just-got-cloned-and-your-brain-isn-t-ready",
      "your-coding-agent-should-do-ai-system-engineering",
      "your-fine-tuned-model-is-tech-debt-a-50x-roi-house-of-cards",
      "your-insecure-mcp-server-won-t-survive-production-tun-shwe-lenses",
      "your-llm-deception-monitor-is-broken-the-fix-is-in-the-training-data-sachin-kumar-lexisnexis",
      "your-llm-ran-out-of-knowledge-now-what",
      "your-mcp-server-is-bad-and-you-should-feel-bad",
      "your-moat-is-your-data-model",
      "youtube-0de56212c994927345ea",
      "youtube-187a1637c7bf015de670",
      "youtube-1ba2add1e7d665064c06",
      "youtube-1fb7da6111555ba92291",
      "youtube-289a4dcaf385a412c330",
      "youtube-2c1a22ae526d567bf524",
      "youtube-3cc00d2e4b8a31683663",
      "youtube-532f0bf69e9eae3a73fa",
      "youtube-53783b315e0664562cfb",
      "youtube-5cc518a90941e34dcf17",
      "youtube-6b7ded464089cc3cfd85",
      "youtube-7b892fd4d253077d46f0",
      "youtube-814050f1b1446db25258",
      "youtube-84a27e80fe3e588bed13",
      "youtube-8af753d08cc761c815d5",
      "youtube-932c82bf93c7a7b8ff03",
      "youtube-9a3275dac7d508a432ff",
      "youtube-9c3fa3c70590bbbf1cf1",
      "youtube-abf8b2b17b3665c83ff1",
      "youtube-bfc00573674178488871",
      "youtube-cf62a92c82ed1de99233",
      "youtube-eb68d1d97fc72ce2a488",
      "youtube-f33c9077df5560520779",
      "youtube-fb96a7d87f9fa9556b4c",
      "z-ai-glm-4-6-what-we-learned-from-100-million-open-source-downloads-yuxuan-zhang-z-ai"
    ]
  }
]

slug · name · url · summary · speakerSlugs · talkSlugs

[
  {
    "slug": "google-deepmind",
    "name": "Google DeepMind",
    "url": "https://ai.engineer/orgs/google-deepmind",
    "summary": "Google DeepMind develops AI models and applications for consumers, developers, businesses and scientists. Its **Gemini** models power an assistant app and capabilities across Google Search and Cloud. The Gemini app lets people work with voice, camera feeds and screen sharing, while small businesses can create images, video and audio for marketing. **Gemma** offers open models small enough to run on local devices, and **AlphaFold** helps scientists predict molecular structures and interactions.\n\nDeepMind was founded in 2010 by Demis Hassabis, Shane Legg and Mustafa Suleyman; Google DeepMind brings together DeepMind and Google Brain. In August 2026, Hassabis became Chair of Google DeepMind and Alphabet Chief Scientist, with Koray Kavukcuoglu taking operational leadership as SVP of Google DeepMind. Its research includes DQN, which combined deep learning and reinforcement learning to learn Atari games from pixels and scores, and AlphaFold 2, which predicts protein structures from amino acid sequences.\n\nThe organization connects research with both mass-market applications and scientific infrastructure. Google reported that the **Gemini app surpassed 1 billion monthly users** in August 2026. The freely available AlphaFold Protein Database, launched with EMBL-EBI in 2021, expanded to more than 200 million predicted structures in 2022. By November 2025, DeepMind reported that it had been used by over 3 million researchers across more than 190 countries.",
    "speakerSlugs": [
      "logan-kilpatrick",
      "nicholas-kang",
      "michael-aaron",
      "philipp-schmid",
      "patrick-lober",
      "paige-bailey",
      "guillaume-vernade",
      "ian-ballantyne",
      "thor-schaeff",
      "sander-dieleman",
      "kat-kampf",
      "ammaar-reshi",
      "kevin-hou",
      "shrestha-basu-mallick",
      "florina-muntenescu",
      "oli-gaymond",
      "cassidy-hardin",
      "omar-sanseviero",
      "mukund-sridhar",
      "aarush-selvan",
      "raia-hadsell",
      "kp-sawhney",
      "benoit-schillings",
      "gus-martins",
      "brendan-o-donoghue",
      "jack-rae",
      "kathleen-kenealy",
      "nidhi-kaushik-vyas",
      "valeria-wu-fon",
      "tom-ouyang",
      "dumitru-erhan",
      "shane-gu",
      "nicole-brichtova"
    ],
    "talkSlugs": [
      "a-year-of-gemini-progress-what-comes-next",
      "agentic-evaluations-at-scale-for-everybody",
      "ai-engineering-with-the-google-gemini-2-5-model-family",
      "any-to-any-building-native-multimodal-agents",
      "build-deploy-ai-powered-apps",
      "build-deploy-ai-powered-apps-18cc1b",
      "building-conversational-agents",
      "building-generative-image-video-models-at-scale",
      "building-in-the-gemini-era-kat-kampf-ammaar-reshi-google-deepmind",
      "defying-gravity",
      "don-t-ship-skills-without-evals",
      "from-transcription-to-live-music-gemini-s-audio-stack-thor-schaeff-google-deepmind",
      "frontier-feud",
      "gemini-nano-on-device-florina-muntenescu-oli-gaymond-google-deepmind",
      "gemma-4-deep-dive-cassidy-hardin-google-deepmind",
      "gemma-deepmind-s-family-of-open-models",
      "how-deep-research-works",
      "how-google-deepmind-is-researching-the-next-frontier-of-ai-for-gemini-raia-hadsell-vp-of-researc",
      "how-google-deepmind-runs-agents-at-scale-kp-sawhney-ian-ballantyne-google-deepmind",
      "let-s-go-bananas-with-genmedia",
      "milliseconds-to-magic-real-time-workflows-using-the-gemini-live-api-and-pipecat",
      "research-to-reality-with-google-deepmind",
      "sovereign-escape-velocity-ownership-with-open-models-gus-martins-and-ian-ballantyne-google-deepm",
      "text-diffusion-brendan-o-donoghue-google-deepmind",
      "thinking-deeper-in-gemini",
      "unveiling-the-latest-gemma-model-advancements",
      "veo-3-for-developers",
      "why-senior-engineers-struggle-to-build-ai-agents",
      "youtube-0248b9a525953988f244",
      "youtube-73b4dd769ce49567782c",
      "youtube-779ba950e9d92722872b",
      "youtube-864f145429d587598a52",
      "youtube-9b2aaaa07c15a78315e7",
      "youtube-9bd4b34b669298acf016",
      "youtube-b01c3c14dd8f8e90af4c",
      "youtube-ba21fa5f6c4457b4bcf7"
    ]
  },
  {
    "slug": "anthropic",
    "name": "Anthropic",
    "url": "https://ai.engineer/orgs/anthropic",
    "summary": "Anthropic builds **Claude**, a family of AI models and applications for coding, agents, and professional work. Its products serve individuals, developers, researchers, and enterprises: **Claude Code** is its coding agent, while Claude Science combines researchers’ tools and packages with access to computing resources and produces auditable artifacts. Developers can use its API to connect agents to external tools, with tool search and programmatic calling designed to handle thousands of tools in complex workflows.\n\nIn **September 2026**, Anthropic [began merging Cowork and chat into one Claude](https://claude.com/blog/cowork-is-now-claude), rolling out first on Pro and Max. Its [redesigned Projects](https://claude.com/blog/projects-redesigned) lets Claude coordinate parallel cloud sessions with shared memory and project context. That experience launched as a **limited beta for select Pro and Max Claude Code users**, with broader access and Team and Enterprise support to follow.\n\nFounded in **2021** by former OpenAI employees including CEO **Dario Amodei** and president **Daniela Amodei**, Anthropic is structured as a public benefit corporation. Its mechanistic interpretability research uses dictionary learning to identify concepts represented inside models; in 2024, researchers extracted millions of features from Claude Sonnet and manipulated them to test effects on behavior. Anthropic also introduced the **Model Context Protocol**, an open standard connecting AI applications to external systems, and donated it to the Linux Foundation’s Agentic AI Foundation in 2025.\n\nIn May 2026, the company reported that its **annualized revenue run rate exceeded $47 billion**, alongside enterprise adoption across industries. That month, Anthropic raised **$65 billion in Series H funding** at a **$965 billion post-money valuation**, including $15 billion in previously committed hyperscaler investments. The financing was intended to expand computing capacity, support safety and interpretability research, and scale its products and partnerships.",
    "speakerSlugs": [
      "barry-zhang",
      "mahesh-murag",
      "alexander-bricken",
      "joe-bayley",
      "gagan-bhat",
      "isabella-kai-he",
      "jamie-neuwirth",
      "zack-witten",
      "thariq-shihipar",
      "boris-cherny",
      "cat-wu",
      "lance-martin",
      "katelyn-lesse",
      "kevin-bai",
      "mihir",
      "john",
      "david-soria-parra",
      "karan-sampath",
      "ash-prabaker",
      "andrew-wilson",
      "theodora-chu",
      "john-welsh",
      "eugene-yan",
      "alex-albert",
      "christopher-lovejoy",
      "karina-nguyen",
      "angela-jiang",
      "sachin-malhotra",
      "priyanka-phatak",
      "gabe-cemaj",
      "mike-krieger"
    ],
    "talkSlugs": [
      "agent-skills",
      "anthropic-for-vps-of-ai",
      "anthropic-s-applied-ai-team-on-the-evolution-of-agentic-surfaces",
      "building-agents-with-mcp",
      "building-with-anthropic-s-claude-the-prompt-doctor-is-in",
      "claude-agent-sdk-workshop",
      "claude-code-the-evolution-of-agentic-coding",
      "claude-fable-claude-tag-and-anthropic-s-culture-cat-wu-thariq-shihipar-ft-simon-willison",
      "claude-for-long-horizon-tasks",
      "effective-ai-agents",
      "evolving-claude-apis-for-agents",
      "field-guide-to-fable",
      "forward-deployed-engineering-101",
      "frontier-feud",
      "future-of-mcp",
      "gateways-are-all-you-need",
      "how-to-build-agents-that-run-for-hours-without-losing-the-plot",
      "mcp-origins-and-requests-for-startups-theodora-chu-model-context-protocol-pm-anthropic",
      "remote-mcps-what-we-learned-from-shipping-john-welsh-anthropic",
      "using-llms-to-secure-source-code",
      "what-s-new-from-anthropic-and-what-s-next",
      "why-your-enterprise-tech-stack-isn-t-ready-for-ai-agents-and-what-to-build-instead",
      "writing-principles-for-task-tuned-prompt-engineering",
      "youtube-242b5fb2ac696c582eb9",
      "youtube-32474d2e87f6aba2c4ad",
      "youtube-502af1124ddc0a68f915",
      "youtube-6a8df9148b55ab1a57f3",
      "youtube-6e9bac1afcd6ff1d78de",
      "youtube-7be948df4cdb3236b72a",
      "youtube-d9470a6ec6f5a0a3b05f"
    ]
  }
]

talkSlug · title · startMs · endMs

[
  {
    "talkSlug": "agent-skills",
    "title": "Intro",
    "startMs": 0,
    "endMs": 181000
  },
  {
    "talkSlug": "agent-skills",
    "title": "Skills Defined",
    "startMs": 181000,
    "endMs": 300000
  }
]

talkSlug · title · jsonUrl · csvUrl

[
  {
    "talkSlug": "agent-skills",
    "title": "Don't Build Agents, Build Skills Instead – Barry Zhang & Mahesh Murag, Anthropic",
    "jsonUrl": "https://ai.engineer/api/data/transcript?slug=agent-skills&format=json",
    "csvUrl": "https://ai.engineer/api/data/transcript?slug=agent-skills&format=csv"
  },
  {
    "talkSlug": "ai-coding-workflow",
    "title": "AI Coding Workflow: From Product Idea to Tested Implementation",
    "jsonUrl": "https://ai.engineer/api/data/transcript?slug=ai-coding-workflow&format=json",
    "csvUrl": "https://ai.engineer/api/data/transcript?slug=ai-coding-workflow&format=csv"
  }
]