# No More Slop – swyx

Shawn "swyx" Wang

AI Engineer Code 2025

[Talk page](https://ai.engineer/talks/IoiHI7p12Ao-no-more-slop-swyx) · [Video](https://www.youtube.com/watch?v=IoiHI7p12Ao)

## No More Slop

Cheap generation makes output abundant and judgment scarce. AI engineering needs curation, codebase understanding and accountability to keep that abundance useful.

### How do you grow without losing taste?

[0:24](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=24s)

How do you grow a community, an industry and a conference while preserving the quality that made people care about them? That is the practical problem behind Shawn “swyx” Wang’s declaration of war on slop at the [AI Engineer Code Summit](https://ai.engineer/code/2025). Each expansion creates more possibilities—and more work deciding what deserves attention.

The first AI Engineer Summit backed the emergence of the AI engineer as a role. The second extended that focus to leadership; the third concentrated on model labs. The World’s Fair widened the audience to AI product managers and designers. Code Summit then concentrated on a theme: coding. Curation becomes the means of preserving taste through that growth, expressed in who attends, who speaks and what the event chooses to cover.

### Judge the output, whoever produced it

[1:52](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=112s)

Slop made [Oxford’s 2024 Word of the Year shortlist](https://corp.oup.com/feature/reflecting-on-a-special-word-of-the-year-launch-event/), but [brain rot won](https://corp.oup.com/news/brain-rot-named-oxford-word-of-the-year-2024/). The definition on swyx’s slide associates slop with artificial intelligence and describes it as low quality, inauthentic or inaccurate. He accepts those quality criteria and challenges the restriction to AI: humans can produce all three failures too. **Slop is a property of the output, not a reliable label for its producer.**

![Slide titled “We are in a War against Slop” showing Oxford’s definition with red boxes around “artificial intelligence” and “low quality, inauthentic, or inaccurate.”](https://ai.engineer/api/talk-essay-media/98d08a293e15b876b759601b78f0f860543149327165b590dd1baa361b74f315)

Oxford’s definition of slop, with two phrases highlighted.

The contrasting internet term is *kino*: work that feels creative, accomplished or worth experiencing. Either a human or an AI can contribute to either outcome. That makes the useful question less about whether AI was involved and more about the judgment exercised in making and selecting the result.

### The same tools can produce very different work

[2:42](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=162s)

Crediting Paul Rambles for the idea, swyx starts with **Sora**. He describes his own videos featuring himself as Sam Altman as boring; more creative users make cats playing didgeridoos. Access to the generator does not supply the idea. He then makes the same point through Netflix’s *K-Pop Demon Hunters* and *The Electric State*: a shared studio does not imply a shared level of quality.

A brief visual comparison extends the joke to different models, without a spoken explanation of their merits. The next comparison adds time: something fresh can become slop as a trend repeats and loses its novelty. A further dig at *Game of Thrones* makes deterioration within a single body of work part of the same pattern. Tools, studios and an earlier success cannot guarantee that the next output will deserve attention.

### A promising category is not enough

[3:45](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=225s)

The distinction also applies to products built around the same startup idea. For his first AI Engineer Summit keynote, swyx used Tome, an AI presentation company. He reports returning to the original deck and finding it gone after the company closed. The example brings the discussion from the appeal of generated output to the experience of relying on the product that holds it. He follows it with contrasting takes on vibe coding: a common category still leaves room for very different execution.

Even two exponential charts can deserve different reactions. The slide places a METR chart under “Evidence” and an AI 2027 chart under “Fan Fiction.” These are swyx’s editorial judgments about the presentations, not conclusions established by the curves’ shapes. His invitation is to investigate why apparently similar stories of accelerating progress feel different. **An exponential shape does not tell you what supports the claim.**

![Slide titled “The same AI chart = Kino vs Slop,” with a METR chart labeled “Evidence” beside an AI 2027 chart labeled “Fan Fiction.”](https://ai.engineer/api/talk-essay-media/9decf0a768724dfa9dddababb1cedbb7f2b3e6b7613ae8640e39740205cbd93e)

Two AI progress charts framed as “Evidence” and “Fan Fiction.”

### Generation gets cheaper; judgment still costs

[4:31](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=271s)

Brandolini’s law provides the analogy: refuting bullshit requires an order of magnitude more energy than producing it. Generative AI intensifies the imbalance by making production cheaper. Swyx claims generated-token costs are falling by 100–1,000× each year; he does not specify models, workloads or a measurement period for that annual rate. The relevant pressure is the growing gap between the ease of producing material and the effort required to evaluate it.

His proposed law of anti-slop substitutes taste for energy: fighting slop requires an order of magnitude more taste than producing it. This is a rhetorical rule about the burden of judgment, rather than a measured ratio. If low-effort output is abundant, people who care about quality have to do more than generate competing output. They must elevate what gets made, selected and shared.

### Use AI to improve the selection, then demand good code

[5:12](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=312s)

AI can help with that work. Swyx describes his side project AI News as a newsletter that tells readers not to read it when nothing is happening. Its useful editorial decision is sometimes to save the reader’s attention. He then points to a skill prompt for **Claude** that explicitly acknowledges slop and instructs the model to avoid it. Swyx reports a significant improvement in the displayed left-to-right comparison, without giving evaluation criteria or a numerical score. The practical move is to make quality expectations part of the instructions.

Code raises the stakes. Swyx illustrates the amplification of technical debt with two engineers producing the debt of 50, then cites private-data exposure affecting millions of users. He places these examples in the year of the talk, but does not identify the incidents or provide measurement details. The concern extends beyond unattractive output: cheap code can create expensive maintenance and security consequences.

He then challenges unnamed claims of models working autonomously for 30–60 hours because they omit whether the resulting code was good. Runtime measures how long a process continued; it does not answer whether its output was useful. His demand, modeled on the language of taxation and representation, is **no autonomy without accountability**.

### Match attention to the work

[6:25](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=385s)

The semi-async valley of death names a problem in how people collaborate with coding agents. Swyx’s distinction is between difficult work that benefits from sustained human attention and commoditized work that can move further into the background.

| Work | Collaboration mode |
|---|---|
| Hardest problems | Keep human attention closely coupled to the machine |
| Commoditized tasks | Allow more asynchronous execution |

The choice is about where attention contributes to solving the problem. Making every task more asynchronous would discard the close collaboration he wants to preserve for the hardest work.

Producing code also creates a need to understand it. Swyx presents [Codemaps](https://cognition.com/blog/codemaps) as a way to use AI to scale codebase understanding, and directs attendees to Cognition for a fuller demonstration. That supplies another target for AI assistance: help people comprehend the growing system they are responsible for, alongside helping them change it.

### Operate the application, manage the context

[7:08](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=428s)

Computer use extends assistance into the applications where work happens. Looking back to [Anthropic’s computer-use launch](https://www.anthropic.com/news/3-5-models-and-computer-use) roughly a year earlier, swyx describes agents becoming capable of operating complex applications, including an IDE. He reports using Devin to automate updates to the conference website. The accompanying slide frames computer use as AI testing and shows a Cognition post beside an annotated conversation. This connects autonomous action to the broader task of checking and maintaining what people will actually use.

![Slide titled “You can use AI testing to fight code slop,” showing a Cognition post with a desktop game image beside conversation screenshots marked with red boxes and an arrow.](https://ai.engineer/api/talk-essay-media/9288efb21200c62c9a3fcefac18a9b6ecdd8a9c45656fe00a80c0dbdac5c564c)

AI testing illustrated with a Devin post and an annotated conversation.

The next control is context management. Swyx identifies using subagents to fight **context rot** as a lesson from the previous conference day: delegation can help keep an agent’s working context focused instead of accumulating everything in one conversation. He closes this sequence with a principle he attributes to Greg Brockman: humans design clear module boundaries, and AI writes the code between them. Together, these ideas preserve deliberate structure around an expanding amount of generated work.

### Refuse the incentives that produce slop

[7:59](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=479s)

The closing call-and-response turns these ideas into a response to pressure: “No more slop.” First comes a boss demanding more lines of code by the end of the quarter. The demand rewards volume without establishing value. Next comes pressure to ship an insufficiently tested release that could embarrass the company. The asymmetric burden returns here: creating and releasing the work can be much easier than dealing with its consequences.

Finally, the pressure comes from Twitter’s engagement incentives: publish bait, even when attracting attention pushes toward misleading the public. The final slide pairs an engagement-bait example with a corrective community note. The refrain remains the same. Taste has to survive the moment when a manager, a release deadline or an algorithm rewards abandoning it.

![Slide reading “Algo wants Engagement bait?” beside a social post containing a video preview and a community note with a sentence outlined in red.](https://ai.engineer/api/talk-essay-media/ad719163d5e40e84c628778d94660bddd4d9dcc3ee718b1dd8dddfe7d820882f)

An engagement-bait example with a highlighted corrective note.

## Resources

- [Oxford's 2024 Word of the Year announcement](https://corp.oup.com/news/brain-rot-named-oxford-word-of-the-year-2024/) — Oxford explains its selection of brain rot as the 2024 Word of the Year.
- [Stanford AI Index 2025](https://hai.stanford.edu/ai-index/2025-ai-index-report) — Research highlights covering AI capabilities, adoption, and falling inference costs at a fixed performance level.
- [METR's task-completion time-horizon research](https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/) — Explains how human task duration and agent success rates can measure autonomous software-task capability.
- [Windsurf Codemaps](https://cognition.com/blog/codemaps) — Introduces structured codebase maps with navigation to source lines and contextual explanations.
- [SWE-grep and Fast Context](https://cognition.com/blog/swe-grep) — A swyx-coauthored explanation of specialized retrieval subagents and the latency tradeoffs of codebase search.
- [Improving frontend design through Skills](https://claude.com/blog/improving-frontend-design-through-skills) — Anthropic's guidance on prompting Claude toward distinctive frontend design and away from generic visual patterns.
- [Anthropic's computer-use launch](https://www.anthropic.com/news/3-5-models-and-computer-use) — The historical announcement of Claude's computer-use capability alongside updated Claude 3.5 models.

## Slides

### Every AIE has a Secret

Each AI Engineer event had a focus, from AI engineers and leadership to model labs and design. For the 2025 Code Summit, swyx identifies curation as the focus.

[0:48](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=48s) [A list of AI Engineer events and their themes, ending with curation for the 2025 Code Summit.](https://assets.ai.engineer/sites/aiecode2025/sha256/9fdb4b47be622c925e90166d756d3daa5895f165e896327b773fb58ba637bf91.webp)

Every AIE has a Secret

- 2023 Summit: AI Engineer
- 2024 World’s Fair: AI Leadership
- 2025 Summit: Model Labs
- 2025 World’s Fair: AI PMs & AI Design Engineers
- 2025 Code Summit: Curation

Embedded social posts: [illegible]
AI Engineer Code Summit

AIE / CODE
PRESENTED BY
Google DeepMind
THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

### Past Event Stats

Four sold-out events report between 29k+ and 150k+ remote livestream attendees. Swyx asks how to grow the community while maintaining quality through curation.

[1:24](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=84s) [Four columns compare attendance, livestream reach and YouTube views across AI Engineer events.](https://assets.ai.engineer/sites/aiecode2025/sha256/fee6a2dabb1015b87818a45027abc5c3d152f3c5423c9f01b25e3abc146b8671.webp)

Past Event Stats
All events sold out (attendance + expo sponsorships)

October 8–10, 2023
San Francisco
AI Engineer Summit
5,000+ Applicants
525 Curated Attendees
29k+ Remote Livestream Attendees
399k+ YouTube Views
October 8, 2023 - February 7, 2024

June 23–25, 2024
San Francisco
AI Engineer World’s Fair
150+ Speakers
2,000+ Attendees
50k+ Remote Livestream Attendees
800k+ YouTube Views
June 20, 2024 - Dec 31, 2024

February 19–21, 2025
New York
AI Engineer Summit
5,000+ Applicants
800 Curated Attendees
150k+ Remote Livestream Attendees
1.5mm YouTube Views
May 28 - August 28, 2025

June 3–5, 2025
San Francisco
AI Engineer World’s Fair
200+ Speakers
3,300+ Attendees
150k+ Remote Livestream Attendees
4.5mm YouTube Views
May 28 - August 28, 2025

AIE / CODE
PRESENTED BY Google DeepMind
THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

### We are in a War against Slop

Oxford defines slop as AI-generated content distributed online indiscriminately or intrusively and characterized as low quality, inauthentic, or inaccurate. swyx agrees with those qualities but says they don’t require AI.

[2:05](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=125s) [A slide presents Oxford’s definition of slop as low-quality, inauthentic or inaccurate AI-generated content distributed online.](https://assets.ai.engineer/sites/aiecode2025/sha256/79bf812ce14a2e0ec7d1492d8f6f36559c1179cfd0f1ed5ed39b41d4f9cc22dc.webp)

We are in a War against Slop

slop

(n.) Art, writing, or other content generated using artificial intelligence, shared and distributed online in an indiscriminate or intrusive way, and characterized as being of low quality, inauthentic, or inaccurate.

Interface text:
corp.oup.com/word-of-the-year/
OXFORD
Products & Services
New Thinking
Our 2024 shortlist
Our a[illegible]

AIE / CODE
PRESENTED BY
Google DeepMind
THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

### Oxford’s Definition of Slop: Highlighted

Swyx agrees with Oxford’s quality criteria but disputes tying slop to AI: humans can produce low-quality, inauthentic, or inaccurate content too.

[2:15](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=135s) [The Oxford definition of slop with red boxes around the artificial intelligence attribution and the low-quality, inauthentic or inaccurate criteria.](https://assets.ai.engineer/sites/aiecode2025/sha256/b59a21a94ec32f0b2b329eef76772bc44c23c6ee72cc2ce0aaf312f0c55c658e.webp)

We are in a War against Slop

Our 2024 shortlist

slop

(n.) Art, writing, or other content generated using artificial intelligence, shared and distributed online in an indiscriminate or intrusive way, and characterized as being of low quality, inauthentic, or inaccurate.

Interface text:
corp.oup.com/word-of-the-year/
OXFORD
Products & Services
New Thinking
Our a
AIE / CODE
PRESENTED BY Google DeepMind
THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

### Any human or AI can be an agent of slop

Swyx challenges Oxford’s AI-specific definition: humans can also produce work that is low quality, inauthentic, or inaccurate.

[2:30](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=150s) [A slide states that any human or AI can be an agent of slop.](https://assets.ai.engineer/sites/aiecode2025/sha256/a627bff3624f993c97330a4a905792959c1c0b3e431aae5085d81e7389167d28.webp)

Any Human or AI can be an agent of Slop

Interface text:
AIE / CODE
PRESENTED BY
Google DeepMind
THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

### Kino and slop: video examples

swyx contrasts a creative cat video he calls kino with his own podcast-style Sora video, which he calls slop.

[2:35](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=155s) [A cat video is labeled Kino, and a podcast-style video is labeled Slop.](https://assets.ai.engineer/sites/aiecode2025/sha256/9851fd0dcf1f2a4efe104aeae7114065180e39f15e50820826bf3e1284ba4b16.webp)

Kino

“paula”
@paularambles
there’s ai slop and there’s ai kino. this is ai kino.

Slop

THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

AIE / CODE
PRESENTED BY
Google DeepMind

Interface text:
1:27
From ねこやん
Rate proposed Community Notes
12:56 PM · Nov 13, 2025 · 4.8M Views
725 · 5.7K · 52K · 18K
1.00
Sora
4 · 1

### Kino and slop: film examples

Swyx puts Netflix’s K-Pop Demon Hunters under “Kino” and The Electric State under “Slop” to show that the same studio can produce both.

[2:59](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=179s) [K-Pop Demon Hunters is labeled Kino, and The Electric State is labeled Slop.](https://assets.ai.engineer/sites/aiecode2025/sha256/64b58fa46735e59927d421d07f3f7dba031a992d52bc77d887e4557d91f0ea71.webp)

Kino
KPOP DEMON HUNTERS

Slop
MILLIE BOBBY BROWN
CHRIS PRATT
RAGE WITH THE MACHINES
[illegible] RUSSO BROTHERS FILM
THE ELECTRIC STATE
ONLY ON NETFLIX | MARCH 14
[illegible]

Interface text:
AIE / CODE
PRESENTED BY
Google DeepMind
THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

### Kino ideas can become slop trends

A Studio Ghibli-style photo conversion sits beside another anime image. swyx argues that an idea that feels fresh at the start of a trend can degrade into slop.

[3:15](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=195s) [A social media post comparing an anime-style image with its source photo, beside another anime-style image.](https://assets.ai.engineer/sites/aiecode2025/sha256/9a63681c4aac9a7f178e0e37f80fa34819e705010287aa18c2efec12a6198768.webp)

Kino idea can degrade into Slop trend

Grant Slatton
@GrantSlatton

tremendous alpha right now in sending your wife photos of yall converted to studio ghibli anime

4:26 PM · Mar 25, 2025 · 53M Views
2.4K   5.8K   43K   15K

AIE / CODE
PRESENTED BY
Google DeepMind

THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

### The same team can produce kino and slop

A Game of Thrones poster sits beside a scene with a cup circled in the shot. swyx uses the contrast to argue that the same team can produce both kino and slop, and that humans generate slop too.

[3:29](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=209s) [A Game of Thrones promotional image beside a scene with a cup circled.](https://assets.ai.engineer/sites/aiecode2025/sha256/1dbe92caf1fed34d135b808cb7ae9aa94ee7960c6e7e02f43b6026646477f399.webp)

The same team = Kino AND Slop

HBO ORIGINAL
GAME OF THRONES

THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

Interface text:
AIE / CODE
PRESENTED BY
Google DeepMind

### An asymmetric war against slop

Swyx uses Brandolini’s law to frame the fight against slop: refuting bullshit takes an order of magnitude more energy than producing it.

[4:36](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=276s) [A screenshot explaining Brandolini’s law highlights the greater effort required to refute misinformation.](https://assets.ai.engineer/sites/aiecode2025/sha256/22047edd0a14ca316b7389da40ca592e8dcaac60870a4dd6284b96afb2cb1c6b.webp)

We are in an Asymmetric War against Slop

Brandolini's law

Brandolini's law (or the bullshit asymmetry principle) is an Internet adage coined in 2013 by Italian programmer Alberto Brandolini. It compares the considerable effort of debunking misinformation to the relative ease of creating it in the first place. The adage states:

The amount of energy needed to refute bullshit is an order of magnitude bigger than that needed to produce it.[1][2]

Interface text:
19 languages
Article  Talk  Read  Edit  View history  Tools
From Wikipedia, the free encyclopedia
AI Engineer
Engineering the future of AI

### Taste needed to fight slop

Adapting Brandolini’s law, swyx argues that fighting slop takes an order of magnitude more taste than producing it.

[4:56](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=296s) [A slide says the taste needed to fight slop is an order of magnitude greater than that needed to produce it.](https://assets.ai.engineer/sites/aiecode2025/sha256/12a3f81f6957805918c39bb47e62d4099ad8e77dc588c7b092589650a593a06c.webp)

The amount of taste needed to fight slop is an order of magnitude bigger than that needed to produce it.

AI Engineer
Engineering the future of AI

### AI News issue archive

Swyx’s AI News archive marks several days “not much happened today.” He says the newsletter tells readers not to read it when nothing is going on.

[5:15](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=315s) [An AI News archive highlights several entries labeled as having little news.](https://assets.ai.engineer/sites/aiecode2025/sha256/a63c1ae5ba31a19ad8adfa6dbe924a72f202d9690eaacd3bc96ec025292c2dab.webp)

[cropped]use AI to fight slop

Quick Navigation
2025
November (14)
October (23)
September (22)
August (21)
July (21)
June (21)
May (22)
April (23)
March (20)
February (20)
January (20)
2024
December (22)
November (21)
October (23)
September (19)
August (24)
July (21)
June (20)
May (25)
April (21)
March (22)
February (18)
January (25)
2023
December (24)

November
Nov 20 — Nano Banana Pro (Gemini Image Pro) solves text-in-images, infographic generation, 2-4k resolution, and Google Search grounding
Nov 19 — OpenAI fires back: GPT 5.1 Codex (API) and GPT 5.1 Pro (ChatGPT)
Nov 18 — Gemini 3 Pro — new GDM frontier model 6, Gemini 3 Deep Think, and Antigravity IDE
Nov 17 — xAI Grok 4.1: #1 in Text Arena, #1 in EQ-bench, and better Creative Writing
Nov 14 — not much happened today
Nov 13 — minor updates to GPT 5.1 and SIMA 2
Nov 12 — GPT 5.1 in ChatGPT: No evals, but adaptive thinking and instruction following
Nov 11 — not much happened today
Nov 10 — not much happened today
Nov 07 — Terminal-Bench 2.0 and Harbor
Nov 06 — Kimi K2 Thinking: 1T-A32B params, SOTA HLE, BrowseComp, TauBench && Soumith leaves Pytorch
Nov 05 — not much happened today
Nov 04 — not much happened today

Interface text:
news.smol.ai/issues
AINews
subscribe / issues / tags / Search (Cmd+K)
Filter regex (eg [illegible]): [illegible]
Title  Desc  Tags

### Prompt to fight slop

An aesthetics prompt calls out AI slop and asks for distinctive typography, color and motion. Swyx says the dashboard improves from left to right.

[5:27](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=327s) [A slide combines a frontend design article, an aesthetics prompt, and two dashboard examples.](https://assets.ai.engineer/sites/aiecode2025/sha256/71323a53f0f0627f902d31d5115b683f09280114640c3b6f967fef6f27d4fb93.webp)

You can prompt to fight slop

Improving frontend design through Skills
Best practices for building richer, more customized frontend design with Claude and Skills.

Category: Coding
Product: Claude apps
Date: November 12, 2025
Reading time: 5 min
Share: Copy link

You might notice that when you ask an LLM to build a landing page without guidance, it will almost always conform to Inter fonts, purple gradients on white backgrounds, and minimal animations.

DISTILLED_AESTHETICS_PROMPT = """
You tend to converge toward generic, "on distribution" outputs. In frontend design, this creates what users call the "AI slop" aesthetic. Avoid this: make creative, distinctive frontends that surprise and delight. Focus on:

Typography: Choose fonts that are beautiful, unique, and interesting. Avoid generic [illegible]
Color & Theme: Commit to a cohesive aesthetic. Use CSS variables for consistency. [illegible]
Motion: Use animations for effects and micro-interactions. Prioritize CSS-only [illegible]

Admin Panel
Total Users: 12
Active Users: 10
Admins: 3
New This Month: 4
User Management
[illegible]

Users
2,847
1,243
12
+387
[illegible]

Interface text:
claude.com/blog/improving-frontend-design-through-skills
Claude
Meet Claude
Platform
Solutions
Pricing
Learn
Contact sales
Try Claude
claude-cookbooks / coding / prompting_for_frontend_aesthetics.ipynb
Preview
Code
Blame
Raw
CLAUDE.md file.
AI Engineer
Engineering the future of AI

### What about code slop?

A joke that two engineers can create the tech debt of at least 50 sits beside a report of exposed Tea user IDs, illustrating the risks swyx calls code slop.

[5:41](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=341s) [Slide lists database deletion, technical debt and security breaches as risks of code slop.](https://assets.ai.engineer/sites/aiecode2025/sha256/6025f01f1555b53e2cb541163bcaee25f85a8dd4620503443d6641248674dfd0.webp)

What about code slop?

1. Deleting production databases
2. Creating tech debt
3. Security breaches...

I Am Devloper
@iamdevloper

vibe coding, where 2 engineers can now create the tech debt of at least 50 engineers

Vibe Coding Leads to Major Security Breach: Tea App Exposes 1.6M Users' IDs

Ognen Plavevski
AI Solutions Architect | Skopje | Generative AI | LLMs | MLOps | Deep Learning | Chatbot Development | AI Strategy | Digital...
July 26, 2025

The Tea app, a women's dating safety platform that recently topped the App Store charts with 1.6M+ users, has suffered a significant data breach exposing sensitive personal information.

What happened:
• An unsecured Firebase database containing users' selfies and driver's licenses was publicly accessible
• Personal data was discovered and shared on 4chan forums
• The exposed database has since been secured, but damage may already be done

Key concerns:
• Users submitted government IDs and photos for verification to join the women-only platform
• The app was designed to help women share safety information about potential dates
• This breach undermines the very trust and safety the platform promised to provide

The bigger picture: This breach perfectly illustrates why "vibe coding" and rapid development without proper security foundations is dangerous. In our rush to ship fast and "move fast and break things," we're literally breaking people's lives and privacy.

### Autonomy without accountability

Swyx questions whether longer autonomous runs produce good code, pairing Anthropic’s 30-hour claim with a chart showing agent runs from 2 to 200 minutes.

[6:01](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=361s) [Slide pairs an article about long-running AI tasks with a chart of agent run durations.](https://assets.ai.engineer/sites/aiecode2025/sha256/a7e9e3ca30b7a9b85d1d14de66dbf1d1a1bbcdbc8e3992572d350db964f2683b.webp)

Autonomy without accountability

SHALL I COMPARE THEE TO A SUMMER’S DATA?
Anthropic says its new AI model “maintained focus” for 30 hours on multistep tasks
New Claude Sonnet 4.5 beats OpenAI and Google on coding tests.
BENJ EDWARDS - SEP 29, 2025 3:10 PM

The Claude Sonnet 4.5 logo. Credit: Anthropic

On Monday, Anthropic released Claude Sonnet 4.5, a new AI language model the company calls its “most capable model to date,” with improved coding and computer use capabilities. The company also revealed Claude Code 2.0, a command-line AI agent for developers, and the Claude Agent SDK, which is a tool developers can use to build their own AI coding agents.

Anthropic says it has witnessed Sonnet 4.5 working continuously on the same project “for more than 30 hours on complex, multi-step tasks,” though the company did not provide specific details about the tasks. In the past, agentic models have been known to typically lose coherence over long periods of time as errors accumulate and context windows (a type of short-term memory for the model) fill up. In the past, Anthropic has mentioned that previous Claude 4.0 models have played Pokémon for over 24 hours or refactored [text cut off]

Amjad Masad @amasad · Sep 10
Longer Autonomous Runs.

Agent 3 is 10x more autonomous than V2, capable of handling much more complex builds by detecting and fixing errors on its own.

You can track the progress of your build with Live Monitoring on your phone, freeing you up to focus on other creative work.

Autonomous Agent Run
2 min — Agent V1
20 min — Agent V2
200 min — Agent 3

Interface text:
ars TECHNICA
AI · BIZ & IT · CARS · CULTURE · GAMING · HEALTH · POLICY · SCIENCE · SECURITY · SPACE · TECH · FORUM · SUBSCRIBE
29 · 72 · 589 · 703K
AI Engineer
Engineering the future of AI

### Using AI to fight code slop: fast context

SWE-grep aims to speed context retrieval. The chart places interactive work and long background tasks on either side of a productivity dip where there is not enough to delegate and it is not fun to wait.

[6:22](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=382s) [A SWE-grep article appears beside a graph of productivity against agent autonomy.](https://assets.ai.engineer/sites/aiecode2025/sha256/236d42c3d359a64cb14e688c9f6b0b94e6684b7736a5f0f3306ebc8d95f73ae5.webp)

You can use AI to fight code slop

Introducing SWE-grep and SWE-grep-mini: RL for Multi-Turn, Fast Context Retrieval
October 16, 2025
By Ben Pan, Carlo Baronio, Albert Tam, Pietro Marsella, Mokshit Jain, Daniel Chiu, Swyx, Silas Alberti

IN THIS ARTICLE:
Where to try Fast Context
Motivation
RL for training fast, parallel agentic code search models
Data and Evals
Fast Context as the first step to Fast Agents

TL;DR: We trained SWE-grep and SWE-grep-mini, fast agentic models specialized in highly parallel context retrieval. They match the retrieval capabilities of frontier coding models, while taking an order of magnitude less time. Available now in Windsurf’s new Fast Context subagent!

Modern coding agents face a fundamental tradeoff between speed and intelligence. Frontier models can solve complex tasks, but it can take minutes of searching before they edit a single file, breaking your flow. In Windsurf and Devin, we observed that our agent trajectories were often spending >60% of their first turn just retrieving context.

Context retrieval has been historically done in 2 ways:
• Embedding Search (RAG): once the upfront work of indexing the codebase is done, queries are fast. But:

The Semi-Async Valley of Death
Y-axis: Experienced / Observed Productivity
X-axis: Agent Autonomy
X-axis ticks: 100ms, 10s, 10m, 1h, 10h
Deep Work / Focus / Hardest Problems
Parallel / Background / Commodified
Make faster for flow and interactivity holding quality constant
Trade time for higher intelligence
Not enough to delegate
Not fun to wait

Interface text:
Cognition
DEVIN +
BLOG
CONTACT US
CAREERS ↗
TRY DEVIN ↗
AI Engineer
Engineering the future of AI

### Using AI to fight code slop: Codemaps

Codemaps use AI to scale codebase understanding. The chart contrasts a plateau in manual understanding with understanding that rises alongside model intelligence.

[6:44](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=404s) [A Codemaps article appears beside a graph relating coding ability to codebase understanding.](https://assets.ai.engineer/sites/aiecode2025/sha256/ea8b654d228505df656b121de7c53335e87a9179ab3a44360dd64c36f9817a58.webp)

You can use AI to fight code slop

Windsurf Codemaps: Understand Code, Before You Vibe It
November 4, 2025
By The Cognition Team

IN THIS ARTICLE:
Why Codemaps?
Our solution: Just-in-Time mapping for any problem
Fight back against Vibeslop
What’s Next

“Your code is your understanding of the problem you’re exploring. So it’s only when you have your code in your head that you really understand the problem.” — Paul Graham

Software development only becomes engineering with understanding. Your ability to reason through your most challenging coding tasks is constrained by your mental model of how things work — in other words, how quickly and how well you onboard to any codebase for solving any problem. However most AI vibe coding tools are aimed at relieving you of that burden by reading → thinking → writing the code for you, increasing the separation from you and your code. This is fine for low value, commodity tasks, but absolutely unacceptable for the hard, sensitive, and high value work that defines real engineering.

We all need more AI that turns your brain ON, not OFF.

Your coding ability is constrained by your codebase understanding

Y-axis: Complexity of Codebase / Tasks you can Prompt
X-axis: Time
Your max coding ability
Codemaps Understanding scales w/ model intelligence
Code understanding uplift
Limits of Manual Understanding
Vibe coding uplift

Interface text:
Cognition
DEVIN +
BLOG
CONTACT US
CAREERS ↗
TRY DEVIN ↗
AI Engineer
Engineering the future of AI

### No More Slop

Swyx puts his message on a dark slide and asks the audience to say “No more slop” together.

[8:07](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=487s) [The words “No More Slop” on a dark background.](https://assets.ai.engineer/sites/aiecode2025/sha256/1785c54dac77f35d5b48233b3d04bfc32a045414746c0c6d9c084111eda55fd1.webp)

No More Slop.

Interface text:
AI Engineer
Engineering the future of AI

### AI LOC Mandate?

Brian Armstrong says AI generates ~40% of Coinbase’s daily code and targets over 50% by October. Swyx asks the audience to answer pressure for more lines of code with “No more slop.”

[8:13](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=493s) [A question about an AI lines-of-code mandate beside a post and rising chart.](https://assets.ai.engineer/sites/aiecode2025/sha256/e1878e597927fcc94c80ce1a9a59f3bdc303214167bbf240fbd559b7b925d0c1.webp)

AI LOC Mandate?

Brian Armstrong
@brian_armstrong

~40% of daily code written at Coinbase is AI-generated. I want to get it to >50% by October.

Obviously it needs to be reviewed and understood, and not all areas of the business can use AI-generated code. But we should be using it responsibly as much as we possibly can.

Coinbase % AI-Generated LOC - T6W
Y-axis ticks: 0.00%, 10.00%, 20.00%, 30.00%, 40.00%, 50.00%
Timeline: 4/1/2025, 5/1/2025, 6/1/2025, 7/1/2025, 8/1/2025

Interface text:
2:58 PM · Sep 3, 2025 · 1.3M Views
1K · 976 · 6.3K · 1.1K
AI Engineer
Engineering the future of AI

### Algo Wants Engagement Bait?

Swyx uses a Wan 2.2 post to illustrate engagement bait: the post claims streaming, while a proposed note says a five-second clip takes roughly a minute to generate on state-of-the-art hardware.

[8:41](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=521s) [A social post about AI video beside a note disputing its real-time streaming claim.](https://assets.ai.engineer/sites/aiecode2025/sha256/e736b4cbcbac84c2a626ae30e843beee51ba23078912b58d21c8f1ecc5d573b2.webp)

Algo wants
Engagement bait?

Lord Bebo
@MyLordBebo

🇨🇳 We’re screwed ... it becomes indistinguishable from reality

Ali’s Wan 2.2 lets you stream without showing your face. It maps your voice and motion onto another face.

Notes suggesting context to be shown with the post

Needs more ratings · Oct 31 · View details
Not shown on X

No it cannot stream in real-time. Generation of a 5 second clip takes roughly 1 minute using state-of-the-art hardware.
https://github.com/Wan-Video/Wan2.2

Is this note helpful? Yes  Somewhat  No

THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

AIE / CODE
PRESENTED BY
Google DeepMind

Interface text:
Follow

### No More Slop

Swyx closes with a call for no more slop, after criticizing the Twitter algorithm’s pressure to post engagement bait and mislead the public.

[8:47](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=527s) [The words “No More Slop” on a dark background.](https://assets.ai.engineer/sites/aiecode2025/sha256/aacff0b5df49f16de79a758ec63b333255082fb0b2f924ca00c97a5552813afa.webp)

No More Slop.

Interface text:
AIE / CODE
PRESENTED BY
Google DeepMind
THE WAR ON SLOP
SWYX / Editor
LATENT SPACE

### AI Engineer Code Summit

[9:03](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=543s) [AI Engineer Code Summit title card with the event website below.](https://assets.ai.engineer/sites/aiecode2025/sha256/6da1c793dd5cbb677a83eda2f53af42277db0f258e41e400b87ac61b616f52b7.webp)

AI Engineer
Code Summit

https://ai.engineer

## Full transcript

[0:11](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=11s) [upbeat electronic music] Morning. How's everyone doing?

[0:23](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=23s) Good.

[0:24](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=24s) Good? I'm gonna need a lot of energy for this talk, so please back me up. I'm very nervous. [laughing] Uh, but we'll get through this. I'm declaring war on slop today.

[0:33](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=33s) Let's talk about this. Uh, every AIE has a secret. I, I've told this to, uh, some folks that are personal friends, and I'll just show you, show the secret now.

[0:43](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=43s) The first summit we had the secret, which was we knew that the AI engineer was gonna be a thing. Second summit, we extended it to leadership. Third summit, we realized that basically we always needed to concentrate on model labs, and that's why you see, um, all, all the, all the top tier labs here today.

[0:59](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=59s) Um, World's Fair, we started expanding the TAM of, of what AI engineering, uh, is, is, is affiliated with, with AI PMs and AI designers. And with the Code Summit, as Jed just talked about, we really started to focus on curation and focusing in on a theme.

[1:14](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=74s) And if there's one theme that really matters this year is, it's coding. But I'm not here to talk about coding. The rest of the day you're gonna hear about coding, so just, just indulge me five minutes to talk about slop.

[1:25](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=85s) Um, we, we've done really well, right? Like, so, so, like, and I think, like, slop is sort of associated with quantity and quality, and I think, like, that's something that I'm really trying to think about as well.

[1:34](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=94s) Like, how do we grow this community, grow this industry, and grow this event, uh, with the, the same kind of taste and high quality that you've come to expect?

[1:43](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=103s) And this is something that, you know, I hope, ho- hopefully you guys can see that we care a lot about, uh, in curating all of you coming here and all of the speakers that you're about to see.

[1:52](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=112s) Um, we're in a war against slop. Um, this is actually, uh, the Oxford English Dictionary. Uh, this is a g- a candidate for the 2024 Word of the Year.

[2:01](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=121s) It lost to brain rot. [laughs] [laughing] But, um, but slop is, slop is pretty good, and I think it's, it's probably gone even more of an issue this year than last year.

[2:11](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=131s) Maybe it will win this year. I have an issue with Oxford, though, because they

[2:17](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=137s) did us dirty by saying slop is co- is generated using artificial intelligence. The other part I, I agree with. It's low quality, inauthentic, or in- inaccurate. But it doesn't take AI to be low quality, inauthentic, or inaccurate.

[2:31](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=151s) Any human or AI can be an agent of slop, right? You've seen this yourself. I'm gonna indulge me with a few examples. By the way, I got this, uh, if you're not familiar with like sort of internet slang, the opposite of slop is kino.

[2:42](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=162s) Um, and I got this idea from Paul Rambles, who, uh, ha- ha- like, you know, when I do Sora videos, I do really boring Sora videos with me as Sam Altman.

[2:52](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=172s) When other people who are actually creative and good at their job do Sora videos, they do cats playing didgeridoos. Uh, slop can be produced by the same studio. There's K-Pop Demon Hunters by Netflix, and there's Electric State by Netflix. [laughing]

[3:08](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=188s) Slop can be produced in terms of different models. No comment. [laughing]

[3:15](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=195s) Slop can be-- Slop can de-- Uh, something that's kino can degenerate into slop, right? If you're s- early on a trend and you're, and y- you're, you're starting that and it's, and it's fresh and new, that's great.

[3:25](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=205s) Um, if you're re- if you recognize the other image, you're too online. [laughs]

[3:31](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=211s) Okay, not enough people recognize that image. [laughing] Um,

[3:35](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=215s) s- go do your homework. Um, yeah, and obviously I'm just gonna throw in a, a dig at Game of Thrones because this is the same, same thing, right? Like, slop is everywhere.

[3:42](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=222s) It's generated by humans and AI. You get it. [coughs]

[3:45](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=225s) Okay, um, the same startup idea can be kino versus slop. When I first presented the, my first keynote at AI Engineer Summit, we actually used Tome, uh, which is sort of like an AI slides company.

[3:55](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=235s) Uh, I loaded it-- I loaded up the same slide deck, and it was gone, uh, recently because, uh, it was, it was actually, uh, sort of, uh, closed as a company.

[4:04](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=244s) Um, there's, there's, you know, different takes on vibe coding, and I think one of them, uh, is much better than the other. And I think like this is, these are just like the tensions that we have to navigate.

[4:12](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=252s) Um, one of our speakers later on, Exo, uh, METR, um, I think it's really interesting that both of them are exponential charts, but one of them feels more kino and the other is more slops.

[4:21](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=261s) And I think, like, I would really like to have people investigate why. Um, so let, let me, let me just skip through. Basically, we're in, in an asymmetric war on slop.

[4:31](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=271s) Um, I think, uh, the closest law that I found that matches this is Brandolini's law, which actually states the amount of energy needed to refute bullshit is an order of magnitude bigger than needed to produce it, right? [laughing]

[4:43](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=283s) So we need to co- coin an, uh, appropriate law as well, um, because, uh, you know, like the, the, the cost of generated tokens is, is dropping by a hundred to a thousand times every single year.

[4:54](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=294s) Um, so this is, I guess, swyx's law of anti-slop. The amount of taste needed to fight slop is an order of magnitude bigger than that needed to produce it, right?

[5:02](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=302s) There's so much low taste out there. We need to elevate, uh, what's out there in the world because that's what we stand for as humans. Um, I think there's, there's a positive message.

[5:12](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=312s) You can use AI to fight slop. Um, uh, I, I'm proud, I'm proud to, um, run as a side project AI News, which is the only newsletter that tells you not to read it when there's nothing going on. [laughs]

[5:22](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=322s) Thank you. [laughs] Oh. [clapping] Appreciate that. Um, you can also prompt to fight slop. The next speaker is, um, B- Mahasin Berry. Um, I found this in the, in the sort of prompts, uh, in the, in the skill set that they, that they put, that they put where they actually a- acknowledge slop and tell Claude not to produce slop,

[5:39](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=339s) and it, it actually improves significantly from left to right. Um, what about code slop, right? We're here about code. Um, creating tech debt where you can sort of-- two engineers can create a tech debt of 50 engineers or, you know, on a more serious note, you can start exposing private data of millions of users, and this, this

[5:56](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=356s) all happened this year. Everything I'm mentioning all happened this year. I'm kind of using this keynote as a f- as a way of recapping. Um, and it, and, you know, just to be spicy a little bit, even people who are saying things like, "Oh, my model can go up to thirty to sixty hours, uh, a- autonomously," well,

[6:11](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=371s) it feels a bit sloppy because you're also not saying, well, was the code good or not? You're just saying how long it went. So in the same way that you have no taxation without representation, you don't want autonomy without a-accountability.

[6:25](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=385s) Um, something I've been working on more recently is that using AI to fight code slop as well. So, uh, this is from the... A, a, bunch of people quoted this yesterday, of the semi-async valley of death, where you can sort of keep human attention and mind meld with the machine in order to work on the hardest problems,

[6:40](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=400s) whereas the stuff that's commoditized, you can make it more async. So you can check out more on that details. What seems to be less appreciated is my, is the other work on Codemaps, uh, which I, which I've more recently done, where we actually use AI to scale codebase understanding, which is also a fight, a way to

[6:55](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=415s) fight slop. And you can, uh, talk to the Cognition folks, uh, downstairs, um, um, who, who can, can show you more in detail. Um, the, the last thing I always want to shout out as well is, is this trend of, uh, computer use.

[7:08](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=428s) I think computer use kind of debuted this time last year, uh, with Anthropic, but, uh, it's, it's getting really, really good now, guys. It can really autonomously operate the most complex apps, including an IDE.

[7:19](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=439s) So that, I think that's really, uh, exciting, and you should probably use that to fight slop. We use it for the website, and here's a example of us using Devin to automate the sort of the, the, the sort of website ups- website updates.

[7:30](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=450s) Um, and finally, uh, something I learned from this conference yesterday is that you can also use subagents to fight context rot. Um, and I think that is one of the biggest themes of, of, uh, that I'm observing as well, if you want to take away something from this conference.

[7:43](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=463s) Um, and I... Al- also, one of the biggest highlights of the year for us as AIE, and myself personally, was chatting with Greg Brockman, who always, uh, preaches the concept of modularity, where you can sort of keep b- clear boundaries on what is human designed and let the AI code everything in between.

[7:59](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=479s) Um, so these are all ideas, but I just have this one message that I want to compre- compress down to you today that I want you to say with me.

[8:07](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=487s) No more slop. Yeah? [cheering] [clapping] Your boss tells you, "I want more lines of code in, by the end of the quarter." What do you say to that? Say it with me.

[8:20](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=500s) No more slop. No more slop. [laughs] You're fighting an asymmetric war. This is how bad it is, right? You have an insufficiently tested release that, that is potentially embarrassing to your company.

[8:31](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=511s) What do you say to people who really wanna push it? No more slop. No more slop. Exactly. [laughs]

[8:38](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=518s) Uh, your, your Twitter algorithm wants engagement bait, uh, and is sort of, you know, force, uh, and, uh, telling you to, to, to lie basically to the, to the broad public.

[8:47](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=527s) What do you say to that? No more slop. Exactly. That's it. Uh, I hope you have a great conference, and let's, [laughs] let's hear it for, uh, not having any more slop.

[8:55](https://www.youtube.com/watch?v=IoiHI7p12Ao&t=535s) Thank you. [clapping] Thank you. [upbeat music]
