@pebble_bed
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States App Store
Account-level information from X, not a live location or the device used for a specific post.
Early stage VC by and for technologists. High conviction, high concentration.
San Francisco, CA
Joined March 2022
- Tweets144
- Following111
- Followers1.5K
- Likes266
The clankers have the keyboard. Rather than mourn the transition, we find it exhilarating.
Announcing After Coding, a new Pebblebed podcast with @keithmadams. Episode 1 with @josh_wills is live.
link >>
Are you an ML researcher looking to meet other researchers, exchange ideas, and hear what people are working on?
Come join us for our monthly ML Suds & Science gathering!
Share what you’re working on, learn about others’ research, or just come hang out and chat with fellow ML folks.
September 22nd, Tuesday, 7 pm.
luma.com/yen3sv4z
Krea is cooking 👨🏻🍳
yesterday, we showcased two major updates coming to Krea: Krea 3 and Krea Agents.
Krea 3 is our new foundation model, with amazing editing capabilities that build on the aesthetic diversity and personalization we developed for Krea 2.
in parallel, we’ve dedicated the past three months to building the world’s best agents for creative work.
Krea Agents is a whole new platform built on top of Krea, with delightful interfaces optimized for agentic workflows.
and of course, this platform will support seamless integration with any of our customers’ existing workflows or agents through MCP or API.
What would you feel like if, in a few years from now, a robot could bring you into a meditative state?
Last weekend, we invited people to experience this future.
We know machines can process data and automate physical labor. But what happens when we ask a machine to automate care?
If someone feels comforted by an AI therapist or a robot facilitating a meditative state, do we actually care that it's just code and motors?
Chelsea, the @taurobots humanoid, took a break from cleaning houses and led this session.
Thank you to @GundersonLaw for supporting this experience.
Our recurring gathering for ML researchers is back on August 26.
ML Suds & Science is an intimate gathering features three short talks covering optimizer choices, architecture decisions, and training runs, followed by dedicated time to connect.
Apply to attend or speak:
luma.com/7g9m6klk
You’ve seen them clean apartments, but what about letting them guide you through a musical experience?
@taurobots have been working hard to master their sound bath skills.
Watch them live on Aug 15
luma.com/gjo9ws1t
Our teammate @AnonymousTermin (Balaji Daggupati) spent years playing grandmasters.
Now, he channeled that into Nous, a new chess engine powered by diffusion instead of search.
By having the model imagine and revise its own moves, he’s built an AI that second-guesses itself the way a human would.
Read here >> pebblebed.com/blog/introduci…
Tau's humanoids are coming to clean your house 🫧
Today, we’re launching Tau’s humanoid cleaning service in San Francisco at $30 per hour.
Access is initially invite-only as we scale operations. If you don’t have an invite yet, join the waitlist at tau-robotics.com.
All footage is shown at 1× speed. Each humanoid is jointly controlled by a human operator and AI.
Literally enabling the blind to read again. An incredible accomplishment for @maxhodak_ and the @ScienceCorp_ team!
If ML is what you do all day, come find your people.
Tuesday, July 21st at 7pm: the third ML Suds & Science at Pebblebed. Four short talks on research people are working on right now, then open conversation.
luma.com/9w02xctf
When @jamesse_ and @benmc27 told us they'd ship hundreds of thousands of buildings — robots and everything — we believed them. Proud to have backed Build's seed.
Read @tammiesiew article about @builddotinc >> pebblebed.com/blog/build
In an age of abundant cognition, the surplus flows to the binding constraint: land.
Proud to support @builddotinc that is building the systems that manage it!
The world's data centers can't move fast enough. We build the agents to fix that.
Today Build launches with $8.5M from @IndexVentures, @Pebblebed (co-founders of OpenAI and Meta AI Research) and angels including the CFO of OpenAI and CTO of Blackstone.
It's time to Build.
Pebblebed retweeted
so much fun hosting friends with our friends at @pebble_bed for an @aiDotEngineer crossover!! sg in sf showing up! 🇸🇬🤝🇺🇸
Memory is the part of the agent stack everyone underestimates until they're burning most of their tokens rebuilding context the model should already have. Good thing @cognee_ has been heads down on it. Congrats on 1.0!
Introducing Cognee v1.0: a major breakthrough in agentic intelligence.
It is 145% better than Opus 4.8 and GPT 5.5 at long context memory retrieval.
Cognee allows a 100 BILLION token context window 100,000x more than Claude. It's:
- 6.9x cheaper than GPT 5.5 and Opus 4.8
- Cold starts in 350ms & searches in 260ms
Why this matters:
Today agents forget important context, redo tasks, waste tokens, and slow down as workflows get more complex.
Cognee solves this.
It’s not a place to build agents. It connects to the agents you’ve already built, across any platform, and makes them significantly cheaper, faster, and more accurate.
Here's how it works:
Readers added context they thought people might want to know
Cognee's website and GitHub do not provide benchmarks or evidence supporting claims of 145% better long-context memory retrieval than GPT-5.5/Opus 4.8, a 100 billion token context window, or 6.9x cheaper costs. cognee.ai github.com/topoteretes/co…