Ramaneumann

⛴️⚡

[Emad @EMostaque]

If Claude is good enough for Nobel Prize winners it is good enough for you

https://t.co/P5pM4MGSqQ

[evan conrad @evanjconrad]

best possible lee kuan yew quote: “the american culture is we start from scratch, and we beat you”

[Joshua Meier @joshim5]

Today we are announcing our collaboration with Pfizer to put Chai's frontier AI—including our latest model, Chai-3—directly into the hands of one of the world's leading pharmaceutical teams.

[Lisan al Gaib @scaling01]

Claude Mythos speeds up training code of small AI models by 52x

Humans need 4-8 hours to reach a 4x speedup

[Richard Hanania @RichardHanania]

India has sent 96 people to America who started billion dollar companies. No one else is even close.

There's only about 5 million Indians in America. Almost one in 50,000 of them is a unicorn founder!

What a holy, special, beautiful people.

I will always fight for them.

[Adam Karvonen @a_karvonen]

Big development - Anthropic is now advocating to build verification mechanisms to enable the option to pause AI development.

[himanshu @himanshustwts]

quite a piece of a release where they used ~ 20 models in different stages of training

- used 4 models from qwen family
- used 6 models from deepseek family
- used 2 models from minimax family
- used 4 models nvidia's stack
- used 2 models from glm famiy
- used gpt-oss-120B + gpt-5.2

[kasra @kasratweets]

bringing back the best pieces of media on this

[rohit @krishnanrohit]

I literally laughed out loud reading this

[Arnav Gupta @championswimmer]

See the system is never at fault. Either you are at fault, or your lawyer is at fault, or Nehru was at fault 60 years ago, it one of these things.

But the system is never at fault

[Y Combinator @ycombinator]

"You need to make yourself a big target for luck, and the way to do that is to be curious. Try lots of things, meet lots of people, read lots of books, ask lots of questions."

— Paul Graham, How to Do Great Work

https://t.co/AyIk5fnKZL

[Gajesh @gajesh]

saturday morning, gautham gives internal access to all eigen labs employees.

we started off doing some small parameter optimisations.

sunday 5AM and we had our first big breakthrough with my harness + opus 4.8 + workflows.

we have been up all night just breaking this and we just cracked this.

24 hours later: we’re giving access to everyone and we have got 82% closer to google’s confidential algorithm.

we started the challenge with ~51% there and initially gautham wrote it fully from scratch.

all with collaboration, composition of work and ultimately: open innovation.

can’t wait for everyone else to play with this.

let’s innovate and more such challenges coming up.

[we’re now at 5.5billy]

[ADONIS @adonispara]

Nietzsche describing the mindset of high performers

[gabriel @gabriel1]

half of the best engineers i know say shit like "yeah i spent 2.5 years at a startup, we were doing insurance for greek frogs out of college and i made $80k, but now i make $5m working on spaceships"

[kalomaze @kalomaze]

i want to not overclaim but afaic this is the single worst security vulnerability in social media history, for any platform, period
i don't think people are giving this the weight it deserves. it's not generically bad. the scope was ~any non-2FA accounts. as in, billions hackable

[Aditya 🦀 @ItsAditya_xyz]

I did a little bit digging. Found out that your collg is already hacked by someone from turkey.

The page is SEO hijacked https://t.co/OyHRaoy7QT and shows links for turkish escorts lol

[Farza 🇵🇰🇺🇸 @FarzaTV]

Watch me control my computer with just my voice. This is the future of operating systems.

No hands.

GPT-Realtime 2.0 is very, very underrated.

Demo:

[ib @Indian_Bronson]

250 years of continuously extinguishing the bottom ~2% of men who could not behave produced the modern Englishman and Anglo-American , who from the early 1700s to early 2000s almost entirely created the modern world.

[Ankit Gupta @agupta]

I agree with @JTLonsdale on these suggestions.

I’ll add one more. aggressively expand the most successful public-private partnership in American history: the research university system.

We should 10x the budgets of the NIH and NSF and fund massive dataset buildouts through universities that can be used to power the next gen AI systems. Training on those datasets can be restricted to American entities for a number of years to bolster our industries.

and yes we can reform the parts of academia that need fixing in parallel. There is no world in which America wins in which the majority of new basic science is happening abroad.

[MiniMax (official) @MiniMax_AI]

Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities

- Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas
- MiniMax Sparse Attention scales context to 1M
- Natively Multimodal from Step Zero

API: https://t.co/fHRdSV7BwZ
Token Plan: https://t.co/BDCycxepZw
🚀New! MiniMax Code: https://t.co/GvB4YiB6Ul

Weights & Tech Report in ~10 Days