I don't know anyone who uses LLMs who is not occasionally weirded out by what they can do. If you are not, you should be. They are weird. Wolfram had a rather startling (at least at the time) theory after using ChatGPT. Understanding whether he is right is important.

Definitely going to try Knip npmjs.com/package/knip … Claude Code leaves a lot of obsolete files around, hard to stay on top of.

Codex CLI hype, final verdict: The codex cli hype is real, I just tried it. GPT-5 (high) in codex is great: - It stays on track much longer than opus - Never "gives up" on your task even if takes a while - Much longer context window - Not arguing & "you're absolutely right"ing

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

If dopamine = pleasure... Then why does chocolate still taste good when it’s blocked? Because dopamine isn’t about liking. It’s about wanting and learning. Let’s rewire how we think about the world's most popular neurochemical 🧵

You don’t need api access necessarily just cli. Gemini, cursor-agent -m gpt-5, codex, at least will run headless. I regularly ask CC to ask Gemini, cursor-agent, or Codex for review and almost always get value … and especially when stuck!

A list of things I think Claude Code could do to win back people switching to Codex CLI: - open source Claude Code - reduce sycophancy/make it less verbose (or add option for that) - more transparency about how/why the model degrades - fix tui flashing bug! PLEASE - improve

You can’t optimize fear. You can only face it. Understand it. Dissolve it.

I agree I have had good luck using migrations with Prisma, and one pretty bad situation getting too far in a project letting Claude Code directly modify schema and data. You have to be explicit to have CC use migrations. Some tips below from Ian Nuttall and others.

Claude Code keeps adding "2024" to web searches even though we're in 2025, getting stale results. Built a pre-tool hook that automatically updates search queries with current year: npx claude-code-templates@latest --hook=pre-tool/update-search-year --yes Fixes the search

Codex 0.35.0 is out in pre-release with a new OpenAI GPT-5 Codex model that’s faster, has improved code quality, more steerable and better for complex tasks. You can also run codex resume now to pick up a previous session

Codex CLI pro tip Add this updated function to bashrc/zshrc & run cdx instead of codex to get: - cdx update to update codex - gpt-5 codex model - web search enabled - full auto mode (read, edit, run commands) - summaries of model thinking (function in image alt text)

I’m noticing Claude Code improvements from recent degradations, which is good. These CC issues led to me giving Codex a serious go and finding it worthy. CC is still more polished from a UX perspective, Opus is a strong coding model, but Codex/gpt-5 is on a strong arc and

Interesting how small of a dose of melatonin (spoiler: .3mg)

Updated TLDR (after hammering Sonnet 4.5 for 24h); GPT-5-Codex is good and fast Sonnet 4.5 is great and fast GPT-5-High is great and slow Opus 4.1 is no longer needed

Yes, I do this too! Keeping temporary design docs and plans in GitHub issues helps auto reduce the superfluous files and keeps a record that fits in a natural workflow.

"Sleep is not something you do, it is something that happens" "The enemy of sleep is effort" youtube.com/watch?v=AQF_eo…

Copying and pasting from Claude Code can have issues (like indentions). One strategy I use regularly is ask CC to copy the contents (prompt, url, etc) to my clipboard using pbcopy. Works well for Mac, not sure what the equivalent is for other OS.