Vibe coding & coding agents
AI reading list
Vibe coding starts with describing what you want a model to build. The people here explore different levels of involvement: quick prototypes, repeated workflows and engineering that keeps tests and code review in the loop. Choose a task, open their own work and compare what each approach asks of you.
People to read
Open the first material, then continue with the author on X. People appear once, grouped by their main reading use and ordered by handle within each group.
10 people
No people match that search. Try another name, subject or reading task.
Builders & products
Start with an app or demo and the person’s own account of how it was made.
-
Replit's founder explains the move from an editor to creating software through natural language. His own essays connect the product vision to concrete infrastructure choices, including parallel agents and reversing their changes.
Start with this Web material · amasad.me
Read the Replit pivot in section one, then the filesystem example in section four: earlier infrastructure supported new agent workflows.
This work in the author’s profile
Published . Material checked . Authorship source.
-
Pieter shares the working setup of a solo developer building products with AI: agents on remote machines, switching devices and interactive app previews. His project archive connects those notes to products and the Vibe Jam competition he organizes.
Start with this Web material · levels.io
A year of coding on a VPS with Claude Code
His remote setup explains how one coding session stays available across devices, and what he would change for a team.
This work in the author’s profile
Published . Material checked . Authorship source.
-
The creator of OpenClaw documents building his own apps and command-line tools with agents. His workflow posts connect prompts, screenshots, iteration and refactoring to real projects, with the October guide and December update showing how his approach evolved.
Start with this Web material · steipete.me
Just Talk To It - the no-bs Way of Agentic Engineering
A detailed account of steering agents while building apps: discuss options, show a screenshot, keep changes understandable and maintain the code. Read it as Steinberger's experience developing his own projects.
This work in the author’s profile
Published . Material checked . Authorship source.
Agent workflows
Compare ways to supply context, plan tasks and collaborate with coding agents.
-
The creator of Claude Code explains how the tool fits into a developer's workflow and shares his own setup. Read him for project instructions, reusable commands, planning and a way for the agent to verify its changes through tests or a browser.
Start with this Video · latent.space
Claude Code: Anthropic's Agent in Your Terminal
In this interview with Cat Wu, Cherny explains terminal access, simple project memory and reusable commands. The transcript makes it an accessible first read on why a coding agent is designed this way.
This work in the author’s profile
Published . Material checked .
-
Read Ralph for an agent loop with specifications and feedback from builds or tests. Then explore Jiti, a running application that gains functions through conversation. Huntley gives you two concrete approaches to repeated development with an agent.
Start with this Web material · ghuntley.com
Ralph Wiggum as a "software engineer"
A compiler experiment explains the loop: one task at a time, specifications, a fresh context and feedback from builds or tests.
This work in the author’s profile
Published . Material checked .
-
Riley makes AI app ideas concrete by connecting a service's input and output to a particular user workflow. His writing also asks what an app contributes when competitors can use the same model.
Start with this Web material · newsletter.vibecodeapp.com
AI Wrapper Startups: What They Are & Why They Matter
Start with his definition of an app built around another company's AI, then compare niche knowledge, distribution and API dependence.
This work in the author’s profile
Published . Material checked .
-
Steve's Beads and Gas Town writing tackles forgotten tasks, session handoffs and agents working in parallel. Read him when a growing coding workload needs a persistent record and coordination beyond one chat.
Start with this Web material · steve-yegge.medium.com
Start with planning outside Beads and recording task dependencies. Then see how the next agent session can resume that work.
This work in the author’s profile
Published . Material checked . Authorship source.
Engineering & evaluation
Understand what the generated code does and how people evaluate the result.
-
Addy explains how to keep agent-written changes inspectable: start with a specification, work in small steps and check the running result. His engineering essays also cover the tools, context and evaluations around a model.
Start with this Web material · addyosmani.com
My LLM coding workflow going into 2026
Follow the specification, small-change and testing sequence. It gives you a concrete way to organize an AI coding session.
This work in the author’s profile
Published . Material checked .
-
Read Karpathy for the original description of building through conversation, then explore autoresearch to see an agent change code, run an experiment and measure the result. A useful route from the phrase vibe coding to the mechanics of a repeatable agent task.
Start with this X post · x.com
Vibe coding: the original description
Karpathy describes asking a model for changes while the code recedes from view. Read the original example before comparing other agent workflows.
This work in the author’s profile
Published . Material checked . Authorship source.
-
Willison explains when a quick AI-built experiment is useful and what you need to understand before maintaining or sharing it. Follow his essays with the Agentic Engineering Patterns guide for concrete ways to test generated code and ask for a source-based walkthrough.
Start with this Web material · simonwillison.net
Not all AI-assisted programming is vibe coding (but vibe coding rocks)
A clear introduction to the original meaning of vibe coding and the responsibilities that appear when other people depend on an app. Useful before deciding how much code review and testing your own project needs.
This work in the author’s profile
Published . Material checked . Authorship source.
A few bearings for your reading
These guides explain the people, language and questions you may meet along the way.
From a vibe-coded prototype to a checked change
Use a small search-filter bug to connect a requirement, code walkthrough, failing case and focused fix, with work from Simon Willison and Addy Osmani.
An AI agent demo checklist: tools, state and success
Assess an AI agent demo through its tools, state, access and checked outcome. Read a published database-error recovery and plan the next useful test.
How to evaluate an AI application on your own task
Define a useful success test, trace a published failure and compare fixes through Hamel Husain, Jason Liu, Eugene Yan and Chip Huyen.
How this list is selected
We selected people through their own writing, software, demonstrations and explanations of how they work with AI. Some build products or sell tools; others focus on engineering practices and model limitations. Each card explains a reading use and opens a specific original work. Vibe coding and reviewed agent-assisted engineering cover different levels of responsibility; the selection lets you compare them.