Resource Caps Are Business Strategy, Not Design Flaw
The whole "built with agents in mind" thing is weird. No one owns you infinite resources just because you are using an agent. Deciding to limit how much resources a user can use at a time is smart business, not a design flaw. Claude has token limits. Does it make it "not designed for agents"?
Hire Debuggers, Not Just Perfect Coders
In code interviews, I'd prefer a candidate who has a few bugs and shows his ability to reason about test failures and resolve them, to someone who just writes flawless code. AI can write good code for interview questions. I...
AI and ChatGPT Co‑author My API Code
I used Codex with ChatGPT 5.3 to implement an API. If I want to describe who wrote the code, I should say:
AI Writes Code More Elegant than My Own
Weird feeling where Codex writes more elegant code than I would have. Not because the elegant code is more work, simply because I didn't think of implementing in that specific way. But I appreciate it when I see it. If it was...
Introducing the Gwenchmarks Manifesto: Learn Benchmark Mastery
Folks asked me "what's your plan for gwenchmarks"? At first, it was a joke. But... teaching people how to plan, execute and read benchmarks is a good goal. So I wrote The Gwenchmarks Manifesto as a start. Still a bit...
Automatic Tenant Isolation Built Into Nile by Default
This is a common problem and one of our biggest motivations in building Nile - to isolate tenants automatically and by default.
AI Agents Speed Up Database Benchmarking for Patch Reviews
I discovered a new favorite use of AI Agents. Get ☕️, its a bit long: If you follow the postgres-hackers list, you know this pattern: - Someone submits a patch - Someone else raises performance concerns with the patch The rational thing to do...