Kudos to OpenAI and Gemini for the free month. If you haven't seen they are giving a free month trial.

As someone who spent a decade doing annual corp policy training about what is and isn’t a bribe, even as an in the trenches engineer who spoke to no one outside the company work related. I often wonder, do those still happen? How do people sit through them now? Is the test still part of it?

But the whole thing really was [insert AI name here] fault is a good excuse right?

And Gemini CLI can't shake it off, it's stuck in a "my mistake" loop and can't undo what it did. Coach is pulling him aside, motioning to Claude to get in there.

That depends on how it’s used?

Yes! If your hesitant to give them access to your machine, you can work though a GitHub repo. You lose tokens with checkout and it getting back up to speed on the problem each turn, but it is there. Both Codex and Code have a web version that works through GitHub

And not so much ChatGPT, I should be more clear, I'm using Codex.

I've been doing this to some degree myself ad hoc with ChatGPT, it is pretty good at being a project manager, just using prompting techniques. It works pretty well. Having something tailored would be nice though.

Such a good book!

You can get Qwen 1.7B 4 bit on device pretty easy. Just need to juggle it cause the KV cache will eat ALL the ram. MLX doesn't have an elegant way to do that unless I missed it so I just start and stop it behind the scenes. The "player" never interacts with the LLM directly, it's behind the curtain.

That's hard to say 😀 I think so. macOS and iOS, building something I always wanted, others may like it. Uses an on device LLM. I "really" wanted to use Apple's Foundation Model, it's on device, but they have really put the guardrails on that thing. It refuses to generate text a lot, switched to Qwen

If a game is cool, yeah.

Finished with 3% left! Woo