what'd we get the $250 anthropic credit for?

ah, assumed that was about API; seems it's something else https://x.com/ClaudeDevs/status/2102871550974427462

This is not good.. twice now GPT 6 SOL has just all on its own to apply code corrections without permission. I have a standing rule to show all diffs before it applies anything. Twice now when it applied a change, it randomly decided to tweak a totally different file in a completely different folder because one of its use cases that it showed me was hallucinated and when I wanted to apply the fix it felt the need to make sure that the user case did in fact show what it dreamed up so it wrote and applied code to ensure the result. Unbeleivable...

I guess even llms.. If you cant make it.. fake it.. DOH!

Reminds me of horror stories where people would say things like "I just want to start over.. this is not working out." and the ai would delete their entire git or database and reinitialize it to "start over". I never beleived those stories until today....

We have a new leader! Opus 5.5 takes the lead!

I've been impressed with 5.5

Seriously? ChatGPT Pro max for 500?



Modified by: PitViper on 2026-09-24 07:32:14 PM

Hmm.. this to me screams they are bleeding bad...

Time to switch to claude

but will that be like max100x? Because they have a max5x and a max20x like anthropic, right?

Modified by: Mysticdrew on 2026-09-24 07:37:59 PM

I read that right now.. claude 20x has 3x the tokens of openai 20x. And they are still going to push the 500 plan. Insane.

Codex just updated today.. and introduced more bugs.. I'm seeing more and more code slop everywhere..

Might need to look at retrying to host my own LLM.

The qwen3.8 27b is actually really good at coding. Its the one I use on my own local server.

Modified by: PitViper on 2026-09-25 10:55:01 AM