Some thoughts on reflectivity ricardosqa.com/Reflectivity
2
2
20
5,980
context rich real world manipulation data is unbeatable
After 2+ years in the robotics data space, we are shutting @Eidon_AI down. The thesis was right. But the business is brutally hard. We close this chapter by open-sourcing everything we built and sharing lessons for anyone venturing into the space.
Article

Robotics Data is a Broken Business

A technical bottleneck does not necessarily imply a good business. We recently shut down @eidon_ai after 2+ years in the data space. We started in early 2024 with a gamified multimodal data collection

2
533
Ricardo Sequerra Amram retweeted
@F13hq is out of stealth. Welcome to the #P9family, @gpjanik & Ingimar! 👋 They’re building models that natively generate vector graphics, with precision, factuality, and utility built in from the start. AI graphics have earned a reputation for filling space rather than adding value. Gregory and Ingimar see a much bigger opportunity: graphics that people can edit, trust, and actually use in their work. They have a clear view of where today’s models fall short and real ambition to build something fundamentally better. Looking forward to being part of your journey, Team F13!
3
4
15
930
Looking forward to it 🔥
Here is a glimpse of the bare minimum for tomorrow.
1
1
8
965
Ricardo Sequerra Amram retweeted
Once again @attio is the fastest growing CRM in the world (off a baseline that gets higher and higher). When we started many people thought CRM was too hard and urged us to build a point solution, I'm glad we didn't. ramp.com/vendors/categories/…
11
15
82
4,640
Ricardo Sequerra Amram retweeted
Just launched another weather balloon that you can track at balloon.recastsystems.com @Recast is going to be open-sourcing our $25 balloon tracker and weather sensor. Radiosondes shouldn't be $250. We want everybody launching balloons.
a cheeky weather balloon launch to test the home grown gps and sensor module with homies is pretty much the ideal Sunday 👌; the Recast team is developing little weather balloon module to make it easier for hobbyists to launch and track balloons
20
8
199
16,347
This is the mindset, and I'm proud to work with many founders and teams who think like Torsten. They play hard mode. They don't complain. They just f***in do it.
"It’s underestimated how much of a difference aggression and ambition makes. I think 95% of a company's success is that.” @torstenreil on not letting anything stop you: “Everything else is excuses.” “Regulation doesn't matter. Yeah sure it takes longer to start a company in Germany. Yes you have to go to a notary. Who cares?” “It's not a problem.” “The regulations that we are subject to as a defense company with regards to the certification qualification of our systems, export law, and a whole bunch of other things… dealing with confidentiality, classified information, is a hundred times more than what a startup has to deal with in terms of bureaucracy and you can still deal with it.” “It still doesn't slow us down.” “Who gives a shit? Stop making excuses.” “I just don't like when people make excuses.” “You just have to have a few very aggressive companies that point out that everyone else is basically mailing it in, and I think a lot of people are."
2
17
2,174
Ricardo Sequerra Amram retweeted
Today we’re unveiling Odyssey-3, a big step forward for foundation world models. It can control robots, power humanoids, drive cars (on the roads of India!), train AIs, pilot drones, and even play video games. We can’t wait to see what intelligent systems it enables.
261
545
4,466
2,317,395
Ricardo Sequerra Amram retweeted
One of the killer applications of world models is to train agents themselves. If a world model can simulate infinite, intelligent worlds, we can push agents to new heights of intelligence and utility. Can't wait to see what's built with Odyssey-3 ❤️
Today we’re unveiling Odyssey-3, a big step forward for foundation world models. It can control robots, power humanoids, drive cars (on the roads of India!), train AIs, pilot drones, and even play video games. We can’t wait to see what intelligent systems it enables.
12
13
167
18,744
👀 watch this space
The fellas at @MorganStanley are officially actuator pilled: “To give investors a sense of scale for the opportunity (globally here, not just the US), we proxy the actuator opportunity as motors + bearings + reducers, sizing the 2030 global TAM at $94bn in our base case...”
1
1
12
3,617
Ricardo Sequerra Amram retweeted
Kaiko's Series B extends to $110 million, led by @SPGlobal alongside @Nasdaq, @BNPParibas, @cbventures, @RBC, @DRWTrading, Susquehanna, @Broadridge, @CantonNetork, @StellarOrg, and our existing shareholders: @anthemis, @PointNineCap, and @Revaia_Cap. This round represents a significant investment in data and data infrastructure from the institutions that lead capital markets and digital assets today. Read the full announcement 🔗 kaiko.com/news/sp-global-lea…
20
25
174
64,142
Ricardo Sequerra Amram retweeted
Replying to @ric0seq
I am working on that exact problem for my PhD - How should a country value scarce electricity-grid capacity when that capacity is an input into AI compute whose market value can be orders of magnitude larger than the electricity itself?
1
1
2
1,089
Portugal 🇵🇹 is sitting on one of the most valuable assets in Europe, and most Portuguese don't know about it. Surprise, surprise, it isn't the sun, beaches, or tourism. It's grid capacity on the Atlantic coast.
14
31
307
84,241
The fumble wasn't corruption (even if the Influencer scandal brought down Costa's regime). The real problem for me is that Portugal has no way to price grid capacity. Connections are first come, first served. In 2021, nobody knew what they were giving away, and even if they had, there was no legal way to charge for it.
1
14
5,154
For the next 3GW: auction grid capacity for large loads. Let the state pre-permit sites and sell them as powered land. Reserve 3-5% of every campus at cost for Portuguese universities and startups. Put the proceeds in a national sovereign fund. Create prosperity for the people.
2
34
4,418
Ricardo Sequerra Amram retweeted
Looking at Harvey's latest round announced yesterday, one might think "legal AI is done". Our take is that it's not. The performance of frontier models on legal tasks is still very far from where coding models are. Part of the solution is to specialize models. Meet @grubelai, a new research lab out of Tübingen and Munich, founded by Moritz Hardt (Director at the Max Planck Institute) and Reinhard Heckel (Professor at TU Munich) that builds dedicated datasets and new models leveraging test-time training to push the performance of legal models. Excited to lead grubel's preseed round to help get the lab off the ground together with a great line-up of angels including Jeff Dean (former Chief Scientist of Google), Ion Stoica (Databricks cofounder), Chris Re (Together AI cofounder), Harvey's cofounders Gabe Pereyra and Winston Weinberg, and our friends Alexander Rudolf Diehl & Sauraj Gambhir (Prior Labs cofounder). We just hired two great founding members of the technical staff: Vedant Nanda, ex-Mistral and Navid Rekabsaz, ex-Thomson Reuters. We're looking for two more AI researchers and a COO. Join us!
1
4
21
1,585
Ricardo Sequerra Amram retweeted
Welcome to the P9 family, grubel.ai! Officially out of stealth now. ⚖️ Every legal matter has its own facts, documents, and requirements. grubel is building AI that adapts to each one, automating how models learn from a matter and improve their work. We feel lucky to be along for the journey as Moritz Hardt, Reinhard Heckel & team push the boundaries of AI in legal and beyond. Welcome, grubel! 👋
6
5
19
3,068
Ricardo Sequerra Amram retweeted
The best-kept secret (yet) is that the @PointNineCap cafe is so cool that people come to hang there on Sunday (and that @qui_rins has a good eye). Ping @qui_rins if you want to hang w. us there. Anytime, it's open!
3
3
13
1,784