Say the ChatGPT app would provide the functionality to the system and I'd allow a scary popup saying "these guys will own you, sure?".. I guess they are going all in into Gemini instead.

But I don't want Gemini..

Technically it does retrieve the entire history and reevaulate it since the LLM is stateless. Just more ergonomic for the developer.

And prompt caching helps cut the costs down when a conversation drags on.

Tesla sw is miles better than the previous two. It is responsive and laid out well. When you get used to it, it is intuitive. Android was not that bad but the visual design was much worse and it was laggy.

The MB software can die in hell. It is the worst piece of shit ever been built by humans. And it is running on hopes and dreams instead of a capable processor. Note that this was an E class so not an entry model. (Even then, there is no excuse)

At that rate one needs to abolish all modern technology and go tribal. Cause I’m certain my toothbrush runs JavaScript.

Which I would not even try and test though if Anthropic did not ban my account. The shadiest thing I did was to use it with opencode for a while I think. Never installed claw or used CC tokens somewhere else.

This is a weird company doing weird shit.

For debugging I installed Bazzite (Linux gaming distro) assuming compatibility would be shit but I can at least test native linux builds of some games to see if there is a hardware issue. The thing runs perfectly. I've been playing propert windows games on Proton with higher / more consistent FPS. It is kind of funny at this point. Granted I do not play any competitive / multiplayer games.

I guess Valve did a great job on the Steam Deck sw.

Might work out fine on codex.

This caught my attention today as it was in dispute and there were a lot of comments (moslty angry). I like browsing polymarket (never placed a bet myself but it is fun to see the prices). Always thought this would kill someone someday..

GPT 5.4 on codex cli has been much more reliable for me lately. I used to have opus write and codex review, I now to the opposite (I actually have codex write and both review in parallel).

So on the latest models for my use case gpt > opus but these change all the time.

Edit: also the harness is shit. Claude code has been slow, weird and a resource hog. Refuses to read now standardized .agents dirs so I need symlink gymnastics. Hides as much info as it can… Codex cli is working much better lately.