djc404

Same. Both impressive, but Sonnet has swipe and sound for me and Kimi doesn’t. Makes a big difference to the experience.

Bling bling ppl

Right you are Ken!

I think it is because the EU is requiring broader access to the deeper layers that Apple doesn’t want to give.

EDIT: by broader I mean all apps the user installs on their phone.

I’m struggling to build my own evaluation bench for local models against my own (scientific coding) use cases, and realising it’s quite hard. Good coding has many dimensions and it varies depending on the need. Nothing seems to collapse cleanly to a few numbers.

Anyone played with this and have examples I can steal from?

Probably depends on where the plane is coming from. Source: me who just returned from Europe last week, no spraying. And have traveled internationally several times in the last few years and also no spraying.

You can also use .claude/CLAUDE.md which will be found. I symlink that to my AGENTS.md which keeps my top level dir clean

Do you have any sense how using it with codex compares to OpenCode?

It’s always a bit tricky picking the right harness (when you have options). Sometimes the differences are subtle but meaningful. But who has the time to run everything twice and compare all the time!

Haha. I clicked the link and they wouldn’t show me the article until I disabled my ad blocker.

I think in many (most?) cases the best approach is to build a tool that defines exactly what the LLM model can/can’t do based on the requirements and then have it use that. Mostly about minimising the choices the model can make. Of course things can always go off the rails if you’re not careful or watching/validating!

This has been my experience as well, at least for the last few weeks. Codex 5.5 is the better planner and coder across big projects, but Opus is fine, though my Claude 5 hour window lasts ~2x longer than Codex. So I’ll sometimes use an orchestrator/worker skill to spread the load.