Is gpt-5.1 the best code model ever?

I just love 5.1 codex. Asked him to do something, it spend 2 minutes and 20 cents, but forgot to respond with the code. Did the summary, saying it did x and y, but forgot to provide the changed code. This is peak trillion dollar industry performance.

Still waiting for Gemini 3...

My favorite series "Best AI model ever", episode 3781.

GPT: "you probably have something like this in your code..." Me: "I gave you the code, how can this be hypothetical?.." Claude: "change line 105 to this.."

The reason ai can’t fully replace humans is because humans can’t fully and accurately voice their requests in a timely manner and ChatGPT takes the request in the most literal way possible. It’s like a genie “careful what you wish for” 😅 no employer has the time or the patience to sort it out with a robot genie

2025 is almost over, and people still love using benchmarks when it's exceedingly obvious they all fine tune for these benches

The `echo 'noop' > /dev/null'` made me choke on my water laughing

"So I had access to GPT 6 for 8 months now"

Honestly Claude is still the best coding model. I work with Typescript mainly but use a lot of relatively esoteric languages like Solidity and Noir but Claude performs best on my benchmark. Further more it’s much better partner for architecture planning. But I’m very excited for Gemini 3

theyre really challenging the "This is the worst it will ever be" narrative