Today we’re releasing two benchmarks for robot learning from egocentric video.
MicroPlanBench: can a model predict the next action? GPT-6 Astra leads but still misses 1 in 3. GPT-6.1 Sol is essentially tied at about 1/5 the cost.
MicroAnnotationBench: can a model label long clips of real work, with exact timing? Our nanoannot-1 scores 54.8, ahead of every frontier model we tested. The best, Gemini 4 Argon, scores 49.6 at 10x the cost.
One takeaway: knowing what happens isn’t knowing when. The GPT models top next-action prediction, but on minute-long clips they cut 50%+ more segments than human annotators and miss most boundaries. Timing is its own skill.
Links below.
super interesting video by @omarwasm on @FarzaTV's journey from shutting down Buildspace and going off the grid to building @heyclicky
https://t.co/6qzLi3gtKc
Would you shut down a $100M company without selling it?
@FarzaTV did. His reasoning at the time:
“I just thought the thing was worthless. Why would anyone want this pile of garbage? ...I think the thing is garbage all the time. I’m always trying to push it forward.”
When you spend five years looking at everything wrong with your company, you stop being able to see what's right with it.
Sometimes, nobody's harder to convince of your company's worth than you. That's not where the story ends, though. He's building again.
Full conversation with @FarzaTV on YouTube and Spotify.
Replit is officially in London. 🇬🇧
As we open our first international office here today, we’re honored to mark this important milestone with @MayorofLondon, Sadiq Khan and his team, His Majesty’s Government, The Lord Mayor’s Appeal and London & Partners.
London will be the hub for our European operations as we expand our work with small businesses and enterprises, helping more people turn ideas into working products with AI. The city's deep talent, diverse skills base, universities and base for leading AI companies made it ideal for us.
We are also partnering with TLMA which supports young Londoners not in education, employment or training (NEET) on a pilot for upskilling.
This is just the beginning. We’re here to build in London and build with London!
Recently finished reading the third door by @AlexBanayan - highly recommend it.
Have gifted copies to my friends.
Shows you how each successful person has done non-conventional actions to reach their position e.g. Spielberg sneaking onto sets until he got his first role