Code 51 @code51

Joined October 2009

Tweets

337
Followers

131
Following

3K
Likes

2K

Code 51 @code51

5 months ago

Why the heck did they see Demis as the ultimate villain in AI?

0 0 0 15 0

Code 51 @code51

5 months ago

#gpt41 GPT 4.1 multilingual performance is through the roof. I wonder how they performed against #cohere benchmarks. Solid

0 0 0 51 0

After all the dust has settled, it becomes obvious what Gemini 2.5 Pro achieved. It’s hard to not get derailed with benchmarks and ELO. It seems they’ve found a systemic way to keep incremental progress. It’s all OAI vs Google now

0 0 1 43 0

Code 51 @code51

6 months ago

> be me, google > watch all cool sci-fi > invent transformers, attention lol > hire Ilya, leaves for OAI > scared of AI risk, hit the brakes > GPT, normies freak out > internal screaming > Bard, suck big > Gemini, suck bigger but 1M > frantic keyboard sounds > own AI programming

0 0 0 58 0

Code 51 @code51

6 months ago

Isn’t this whole Studio Ghibli trend (mass ghiblification) a huge f* you to artists by the public? As long as you’re memeable, you don’t have any rights on your art it seems.

0 0 0 38 0

Code 51 @code51

7 months ago

OpenAI is nothing without its (core) people maybe?

0 0 1 33 0

Code 51 @code51

10 months ago

AI is the second coming. @GeminiApp #Gemini2

0 0 0 33 0

Download Image

Code 51 @code51

a year ago

really wild that triceratops are evolutionarily far closer to chickens than rhinos

0 0 1 123 0

Code 51 @code51

a year ago

OpenAI is nothing without its contracts

0 0 0 148 0

Code 51 @code51

a year ago

Where the hell is Gemini Ultra? MMLU 90 vs #gpt4o MMLU of 88? 5 months of red-teaming? #geminiultra

0 0 0 212 0

Code 51 @code51

a year ago

#GPT4O Students are set up for a great despair once they see GPT-4o math performance. (Spoiler: It's bad) We have big problems in LLM benchmark design.