Allen @MisalignedModel

OpenAI launched GPT-5-Codex today, making a point to emphasize they've "resolved infrastructure constraints" and can now evaluate on the complete SWE-Bench dataset GPT-5-Codex scores 74.5%, and has finally, legitimately caught up to Claude Opus 4.1 in SWE-Bench ty @deedydas

0 0 1 258 0

Download Image

Allen @MisalignedModel

a month ago

it finally happened to me! I found a Herman Miller for $8 at a garage sale. I can't find any black ones for sale, which might be a good thing? I'll be stoked with anything above $2000 since the insides are not in great shape

1 0 1 155 0

Download Image

Allen @MisalignedModel

a month ago

The reflections on the bonnet (!!!!!) This is crazy

Google AI Studio @GoogleAIStudio

a month ago

The reflections on the bonnet (!!!!!) This is crazy https://t.co/Y3jzurjUNA

295 1K 9K 1.0M 3K

Download Image

0 0 0 174 0

Download Image

Allen @MisalignedModel

a month ago

There are basically O(20) people globally who actually know, like deep in their bones intuition know, how Big O notation works

Anjney Midha @AnjneyMidha

a month ago

There are basically O(20) people globally who actually know, like deep in their bones intuition know, how Big O notation works

62 23 1K 179K 370

0 1 2 120 0

Allen @MisalignedModel

a month ago

but why'd they have to make it look so damn cool

boto @TsarinaBomba

a month ago

but why'd they have to make it look so damn cool

1K 1K 80K 10.7M 4K

Download Image

0 0 0 80 0

Allen @MisalignedModel

a month ago

duality of man

0 0 0 57 0

Download Image

Allen @MisalignedModel

a month ago

this man is too powerful for his own good

0 0 1 43 0

Download Image

Allen @MisalignedModel

a month ago

I am fairly certain that this soccer channel voice is completely AI generated It took me ~5 minutes to figure it out, AI voice is nearly past the uncanny valley youtu.be/7WhA8q66FXI

1 0 0 99 0

Chris Paik @cpaik

2 months ago

Cursor's Problem docs.google.com/document/d/1q3…

107 92 1K 342K 1K

Deedy @deedydas

2 months ago

Ridiculous that OpenAI claimed 74.9% on SWE-Bench just to prove they were above Opus 4.1’s 74.5%… By running it on 477 problems instead of the full 500. Their system card only says 74% too.