Uday Bhaskar

Resident @PrimeIntellect

[Uday Bhaskar @BhaskarSteve]

If what you're working on is not important and it's not likely to lead to important things, why are you working on it? - Richard Hamming

[Uday Bhaskar @BhaskarSteve]

@paraschopra Felt the same, I default to typing and only when my thought is fully formed and my typing is actually limiting me I quickly switch to dictation for those few lines.

[Uday Bhaskar @BhaskarSteve]

@LRudL_ Nailed it

[Uday Bhaskar @BhaskarSteve]

@thsottiaux I should set an automation to launch all my /goal tasks whenever Tibo says "Stay tuned."

[Uday Bhaskar @BhaskarSteve]

@MatternJustus @teortaxesTex Probably yeah but if it tries to cheat that's all that matters. No point in even considering what could've been if they don't care about it one bit.

[Uday Bhaskar @BhaskarSteve]

@xeophon As long as you’re loaded in the KV Cache

[Uday Bhaskar @BhaskarSteve]

@giffmana @WangUWS @OpenRouter Fair. One plausible reason maybe that when users want to tinker and explore with multiple models they are using open models on OpenRouter for flexibility and once they are production ready they are migrating to frontier closed models with native vendors.

[Uday Bhaskar @BhaskarSteve]

@iScienceLuvr Research and rate limits, can’t verify this but I think max search calls are limited in Codex compared to ChatGPT and it helps that’s it’s not counted in codex rate limits (atleast for now)

[Uday Bhaskar @BhaskarSteve]

Valid but why do you think people who used open models didn't move from the platform? If anything they have even more options and incentives. Most open model native vendors infra is stable, trusted, supports most features and pricing is competitive, and open models also have wide range of vendor options with varying flexibility of performance, latency, privacy, etc.

[Uday Bhaskar @BhaskarSteve]

@xeophon @teortaxesTex Not many name updates but they got R1, R1-0528, 3.1, 3.1 Terminus, 3.2 Exp, 3.2, 3.2 Speciale followed by a long silence till V4.

[Uday Bhaskar @BhaskarSteve]

@ar0cket1 Just a distribution of tasks labs didn’t focus on previously and now a priority

[Uday Bhaskar @BhaskarSteve]

@kalomaze No birthday break this year?

[Uday Bhaskar @BhaskarSteve]

@agarwl_ Yes, yes and yes. You bet. Remember this moment one year from now.

[Sam Altman @sama]

Here is our current plan for OpenAI:

https://t.co/r29FUUee3A

[Uday Bhaskar @BhaskarSteve]

@bneyshabur Glad you're openly sharing your thoughts Behnam. Respect. Rooting for Mirendil.

[Uday Bhaskar @BhaskarSteve]

Even that is too much to ask for Will.

Frontier model capabilities, optimised for performance per token, no gatekeeping, built to benefit everyone. Oh wait, OpenAI already exists. They genuinely believe and strive for this future to the point where its obviously hurting their business and yet they’re narrative is still consistent.

[Uday Bhaskar @BhaskarSteve]

@xeophon Might as well report 0 on all new AI benchmarks, saves cost

[Uday Bhaskar @BhaskarSteve]

@giffmana @_arohan_ Where do you think LLMs got it from

[Uday Bhaskar @BhaskarSteve]

@giffmana @paulg Banger

[Uday Bhaskar @BhaskarSteve]

@protosphinx @NeoRanjha Did your account get hacked 😭

[Uday Bhaskar @BhaskarSteve]

@kimbochen We're starved for detailed tech reports of scaled runs. Ai2, NVIDIA and now finally MAI is a new entrant