Erik Hanchett

Erik Hanchett teaches Vue reactivity with a shopping cart. A reader adds a product, the cart’s data changes, and Vue updates what the shopper sees. Erik is a full-stack developer, senior developer advocate at AWS working on Kiro, and creator of the YouTube channel Program With Erik. For more than a decade, he has taught developers through applications they can run and change themselves.

From GeoCities to Program With Erik

Erik built websites on GeoCities in high school, paused college, and worked in tech support before returning to study computer science. At the University of Nevada, Reno, he earned a bachelor’s degree in computer science and an MBA in entrepreneurship. A personal-branding class in the MBA program required him to keep a blog and make videos. Speaking to a camera felt awkward at first. Once he got comfortable, he found he could demonstrate a programming technique faster than he could write out the tutorial. By 2020, he was teaching on YouTube alongside his work as a senior software engineer at Cerity, an insurance technology company in Reno.

Erik’s public teaching led to both of his books. Packt approached him while he was blogging about Ember, and Manning later approached him while he was making Vue videos. His 2016 Ember.js Cookbook lets a developer turn straight to a recipe for the problem at hand, such as inspecting an application or testing a route. For Vue.js in Action, published in 2018, Erik and Benjamin Listwon built the lessons around a single online store. Readers develop the same application from product listings and a cart through checkout and an administrative interface. Erik’s companion code pairs the book’s excerpts with runnable chapter examples, the earliest of which fit on a single HTML page.

Erik also maintains Amazing Developers on YouTube, a community-contributed directory of other programming teachers. Learners can find someone who codes live, look for game development tutorials, or browse channels by the language they teach in.

The cloud services behind a frontend

On the AWS Amplify team, Erik taught frontend developers how cloud services fit into a web application by starting with familiar controls. His Amplify Authenticator tutorial shows how to choose which registration fields a user must fill in and the order in which they appear. In a 2024 Amplify Gen 2 tutorial, he added picture uploads to a Next.js to-do app. Readers could follow a file from the upload control to object storage and define the backend in TypeScript, the same language they used for the app.

At Kiro, Erik and Jay Raval used a hosted note-taking service to explain how a coding assistant can reach tools and data through the Model Context Protocol (MCP). Each developer could connect an assistant to the service without running a local copy of the server. Erik also tried Kiro on his own frontend code. When he moved an older application to Vite+ to bring its development tools into one toolchain, he used Kiro to finish the migration.

What to build first, and how to check it

As a software intern, Erik once dropped his assigned work when a visiting vice president floated an idea. His manager explained that he could write the idea down and discuss where it belonged on the schedule. Erik recognizes that eagerness in coding assistants. He’s comfortable asking an assistant to fix a small bug directly. For a complex feature, he prefers to review the requirements and design while changing direction still means editing a Markdown file.

For Using Spec-Driven Development for Production Workflows, one of his two presentations on the July 2026 AI Engineer World’s Fair online track, Erik asked Kiro to reorder the work on a movie catalog so the first four tasks would deliver a searchable grid he could try. The genre filter gave him a precise requirement to test. Whatever movies the catalog contains, the menu should list each genre once, in sorted order. A property-based test checks that rule across many generated catalogs.

Whether a patch fixes the issue it claims to fix is harder to express as a rule. Erik built a pull-request review app in Nuxt that gives Jev the evidence from a pull request and asks six questions in one request, including whether the patch addresses the reported issue and whether it keeps credentials out of browser-visible configuration. Jev returns a probability for each answer, so the developer can set a threshold for each decision and leave checks with definite answers in ordinary code.

What an agent carries between turns

In his other World’s Fair online-track presentation, Your Agent Is Wasting Tokens and You Don't Know It, Erik explained a cost that counting tool calls can miss. A tool can run just once, yet its entire result can travel back to the model on every later turn. Using Strands Agents, he showed how to store a large result separately and pass the model a summary. Trimming conversation history introduces a different problem. An early design decision can disappear with the message that explained it. Erik recommends carrying a short summary of those decisions alongside the recent messages.

At AI Engineer CODE 2026, Erik is scheduled to compare work that suits a terminal or IDE with jobs a persistent Kiro Crew agent can continue after the developer logs off. His session, Agents That Don't Clock Out, will also cover how he organizes skills and custom agents for reuse across jobs.