Guess what nvidias p/e is

Haladir (@Haladirofficial) is building operational superintelligence. Today's models fail under real-world constraints. Haladir combines formal solvers with LLMs to turn unreliable outputs into verifiable, optimal decisions, from post-training to deployment. Congrats on the

99% of RL fails because you don't wait long enough for it to grok

I quit XAI a week ago. It was an incredible time with incredible people. I will now sail to Australia with my pet penguin.

Did you know z3 is the most powerful piece of software ever written

Most AI agent failures are invisible until users complain. @sentrial_dev detects loops, hallucinations, and user frustrations in real time and helps engineering teams diagnose and fix issues fast. Production monitoring built for AI agents. Congrats on the launch @theneelshar

New York I am in you

this is great:

claude writes code, codex checked claude, amp checks that codex checked correctly, droid implements amp's checks, claude check's droid's code

struggling to see the forest through the trees

Future of Hotels Is Run by Lance. @TryLance now backed by @ycombinator

if you were actually good at research you would be at openai or jane street

Before the YC batch kicks off, we hosted 50 YC founders tonight and got a glimpse of what everyone’s building. Felt like people were less focused on what’s “hot” and more focused on building things that genuinely need to exist. Gonna be a fun few months, it was great meeting

Agi in certain domains is an oxymoron

Cudamalloc()

pls join the capi table 💀

SoftBank boutta be insolvent

I wonder if scaling in robotics is similar to scaling in biology. That is, scaling data like in LLMs doesn’t work that well, but scaling along the modality axis does. Perhaps the models need sufficient modalities in order for scaling to work.

Bing bop boom boom boom bop bam

So excited to partner with @jibranhutch and the team!