i love that the codex team releasing codex cloud, codex cli, codex ide, and codex review all in ~3 months is a self-contained proof of how effective the tools are

Claude Code: no evals [well known code agent company]: no evals [well known code agent company 2]: kinda halfassed evals [leading vibe coding company]: no evals [ceo of company selling you evals]: mmmmm yess all my top customers do evals, you should do evals [vc's in love

How it feels giving claude access to the prod db

Markdown files are all you need

gpt-5 is a super human at resolving merge conflicts. It can pick up a 2 months old PR, understand the underlying intentions, understand how the codebase changed since then, resolve all the conflict and revive the PR. It’s crazy.

This is also the conclusion I’m coming to - git repo - claude code - directory specific claude.md s - compounding self reinforcing knowledge (e.g ‘update the claude.md automatically based on the session) It’s amazing how far this gets you

The Codex CLI and IDE extension are evolving at truly a mind blowing pace. github.com/openai/codex/r…

Just saw the king of NYC himself while eating lunch at the park. Jerry Seinfeld. The legend. I know every episode by heart.. starstruck but I had to ask for a picture. "Hey! Do you mind if I get a quick selfie?" "No, sorry I'm with my family." Getting rejected by Jerry might

I think congrats again to OpenAI for cooking with GPT-5 Pro. This is the third time I've struggled on something complex/gnarly for an hour on and off with CC, then 5 Pro goes off for 10 minutes and comes back with code that works out of the box. I had CC read the 5 Pro version

ok, officially announcing the new AIE CODE summit today. back in NYC, the city so nice we're doing it twice*! CODE is the first modality to hit serious PMF after CHAT, and has become an existential battleground for all the model labs + is a key tool for digital transformation

I just ran into Curtis Sliwa on the subway If you don't know already, we have some other pretty horrific candidates for mayor in NYC. A Caviar Communist A Failed Despot An Ottoman Prince None of whom would be caught dead taking the subway without an entourage. Curtis

trying to switch over to OpenAI Codex can we please get the message queue that's the feature I most miss from Claude Code

one-shotted by openai codex cli

My biggest tip for prompting LLMs: If a human reading the prompt for the first time couldn't do the task properly without asking clarifying questions, then a LLM will fail as well. Flow: write your v1 prompt -> ask yourself "what is still unclear?" -> fix the prompt -> repeat!

claude code is a video game- addictive, easy to learn, hard to master, "just one more turn"

“‘maximally int-” yeah it’s grok

While gpt5 pro is really good, absolutely, the practice of dumping related code + a problem will consistently outperform agent harnesses in most of my testing. The reason is simple - focused context allows the model to reason only on the problem, rather than tool calls.

"Claude Code is not just for coding. It's an AI agent that can do whatever you want it to do." Here's my new tutorial with Alex Finn where he showed me exactly how to build a life OS with Claude Code in 25 minutes. He saves 2+ hours daily using: ✅ /researcher to analyze

This has been my experience too GPT-5 pretty much sucks for essays and notes and personal usage Very very good for tool calling and agentic reasoning

kitze Everything is just an abstraction order higher It's like when C++ people said Python wasn't coding Or Python people to Javascript devs It's all algorithmic tinkering