binal retweetledi
binal
1.4K posts
binal
@binalkp91
ai @ sequoia capital global equities
Katılım Haziran 2013
1.8K Takip Edilen506 Takipçiler
GIF
Punchbowl News@PunchbowlNews
Senate Commerce Committee Chair Ted Cruz on Tuesday acknowledged the desire “to protect against catastrophic risk” from the most advanced artificial intelligence systems, while warning against regulatory overreach. @BenBrodyDC has the story: punchbowl.news/article/tech/c…
ZXX
binal retweetledi
Security is an economic decision.
For a fixed cost, within @XBOW, which model has the best odds of crafting an exploit?
GPT-5.5 > Mythos > Opus 4.6 on real OSS web vulns.
Curves below.
English
So there's still nearly 83% of the market that has nothing to do with semiconductors.
Kevin Gordon@KevRGordon
The Semiconductor industry is now a record 17.4% of the S&P 500's market cap
English
binal retweetledi
[2/5] In arxiv.org/abs/2604.17121, @ShoaibASiddiqui, @savvyRL, and I explain why feedforward transformer architectures are fundamentally ill suited to complex state-tracking tasks.
English
I have apparently accidentally reimplemented Erlang supervisors and it's pretty nice!
English
@celestepoasts maybe I’m numb to magnitude at this point but it doesn’t seem like that much versus all the GW deals being signed?
English
binal retweetledi
GPT-5.5 Pro has ranked 1st on VoxelBench
It scores 100+ Elo points higher than GPT-5.5!
English
A long, long time ago, I was about to receive an offer letter and was asked what would I do if it came with a $10m guarantee. My answer:
“I would hire an army of Cimmerian mercenaries, conquer your fund, see the employees driven before me, and hear the lamentations of the women.”
I didn’t get that job.
English
@Miles_Brundage They said it's an "early checkpoint" in the post which leads me to believe just 5.5 though hard to say to your point.
English
If you are surprised by the GPT-5.5 being good at cyber thing, you have Big AI Lead Delusion. There are none
(sidenote, I'm not 100% clear if this is GPT-5.5 or GPT-5.5 Cyber. Naming conventions are so chaotic + there is ~no info on the latter that it is hard to say)
English
binal retweetledi
v2 for xtr-warp-rs is out, adding sharding support to the indices
The entire search pipeline has been rewritten around efficient transfers and new kernels that enable parallelization and scheduling optimizations, all while staying true to the WARP formula
Details below 👇
English
@tunahorse21 o1 pro when they let you paste in as many tokens as you wanted, i always wonder how much i cost OpenAI those first few months
English
asked it to generate an image of itself answering how many r's are in strawberry
ChatGPT@ChatGPTapp
at long last
English
@Miles_Brundage i think it's reasonable that codex is both growing wildly and openai has enough compute headroom (for now?) to keep resetting rate limits. sounds like they have pathways to keep scaling compute too
x.com/thsottiaux/sta…
Tibo@thsottiaux
Looking at the traffic dashboard for Codex just now, it would be scary if we didn't have a lot more compute coming online in the coming weeks. All according to plan fortunately.
English
Ah that would make sense (though I'm still not sure what to make of something being on fire given the credit reset thing) x.com/binalkp91/stat…
binal@binalkp91
@Miles_Brundage I took “they” to be Anthropic
English
Isn't this obviously false since they keep resetting the limits like every week? x.com/jimcramer/stat…
Jim Cramer@jimcramer
The bottom line: they are short compute and Codex is on fire...
English
binal retweetledi
1/8 Introducing Recurrent Transformer (RT). At 300M params, RT improves validation CE over standard Transformers. The best RT model is only 6 layers, but wider at 2048 — beating deeper 12- and 24-layer Transformers by trading depth for width.
English
binal retweetledi
What does a good language model look like internally in geometry?
We find a simple but surprising signal:
👉 the more spread out its hidden representations are, the better it predicts
(even for semantically similar contexts)
ICLR 2026 arxiv.org/pdf/2506.24106
Presenting now👇
English