Shopify's Claude Code Warning: AGENTS.md Support
Shopify CEO Tobi Lütke may ban Claude Code over AGENTS.md support. See how fragmented agent system prompts and skills affect Codex teams.
Read articleNotes from real projects
I build with coding agents and write down what holds up.
Hands-on AI coding notes from Codex, Claude Code, Cursor, and new LLMs: the workflows I keep using and the mistakes I won't repeat.
Start here
The posts I'd start with
The Story Behind Codex Reset, According to Tibo
Tibo Sottiaux says Codex Reset began as usage compensation after broken builds. He still controls the reset button.
Read article
Ox Alpha Rebuilt My 3D Globe Dashboard on OpenRouter
I ran Ox Alpha in OpenCode on my 3D globe test. One prompt, a working Three.js page, and a live demo at hqman.me/demo/ox-alpha-3d-globe.
Read articleNew notes
Latest articles
GPT-6 Astra Rebuilt This 3D Globe Dashboard
I ran GPT-6 Astra on a static Three.js logistics dashboard. One prompt, a collapsed activity drawer, and a live demo at hqman.me/demo/gpt-6-astra-3d-globe.
Read articleKeep Using OpenAI Models in Cursor After November 12
Use CLIProxyAPI and Cloudflare Tunnel to keep OpenAI models in Cursor after the proposed November 12, 2026 cutoff, with setup and security caveats.
Read articleCodex Usage Reset August 30, 2026: Tibo on Longer Limits
The August 30, 2026 Codex usage reset comes with token fixes from OpenAI. Tibo says the same limits may last 10% to 50% longer.
Read articlePick a topic
Browse the archive
On my workbench
What tools and workflows do I test?
AI coding
Setups and workflows I've reused in Codex, Claude Code, and Cursor. I include enough detail to run them on your own project.
Coding agents
I use MCP and skills to manage context when several agents touch the same codebase. I write about the signs of drift and how I get a run back on track.
LLM reviews
I run Claude, GPT, Kimi, GLM, Gemini, and Grok on code I know. I share the output and cost, including the parts where I had to take over.
How I test
How do I test each tool and workflow?
Each guide follows the same loop, so you can repeat it on your own codebase.
- Pick a real task from a project I already ship, not a toy demo.
- Run the tool end to end and record the exact commands, settings, and model.
- Check the output against tests, a manual review, and the official docs.
- Note the cost, the failure modes, and the point where I had to take over.
- Write the workflow so you can repeat it on your own codebase.
What you get
What can you expect from each guide?
Every post is grounded in real use. Here is what each type delivers and how I verify it.
| Content type | What you get | How I verify it |
|---|---|---|
| AI coding workflow | A repeatable setup with the exact commands | Run on a live project, then checked with tests |
| Coding agent notes | Context control and review tactics | Tracked across real multi-agent runs |
| LLM review | Output, cost, and the limits I hit | Same task on code I already know well |
Named sources
How do I verify claims with official documentation?
When I describe what a tool is for, I check it against the official documentation and link the source.
"Codex is OpenAI's coding agent for software development."
"Claude Code is an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools."
"Cursor is a coding agent for building ambitious software."
Common questions
Frequently asked questions
What is AI Kai about?
AI Kai is a technical publication by Kai Wang covering practical AI coding workflows, coding agents (Codex, Claude Code, Cursor), and hands-on LLM evaluations with reproducible setups.
How are the coding agent workflows tested?
Every guide is tested end to end on real production projects and verified against unit tests, manual code reviews, and official documentation.
How can I reproduce these setups in my codebase?
Each article provides the exact commands, configuration files, prompt architectures (like AGENTS.md), and failure modes so you can replicate them directly.
A little context
Who is behind AI Kai?
I build AI products and use coding agents for the work behind them. Most posts start with a problem I hit on my own machine or in a production codebase. I include the steps and numbers so you can check the result for yourself.