Hey Steve!
Meet Steve: the first self-improving AI to master Minecraft without human help.
It's a step to whole civilizations growing in virtual worlds, and a leap towards game characters that come to life.
Steve out-progresses 97% of players; 26x more than attempts by NVIDIA, DeepMind and OpenAI.
Characters can now develop new abilities and evolve through experience. Soon, in-game towns, civilizations, and whole ecosystems will increasingly look and feel like our own, or like a great story, and cheaply:
Once Steve’s capabilities exist, a System One model like Jev can take over, and it’s 92% cheaper. Soon, hundreds will be able to run locally alongside any game.
Steve completes ~96% of the game’s advancements, 100/105. It also defeats the Dragon and returns with the Egg on any seed. (1/4)
NOTE: Minecraft has been a challenging, unsolved AI benchmark for a decade, and in that time OpenAI (VPT, 2022), DeepMind (DreamerV3, 2023) and NVIDIA (Voyager, 2023) got their agents about as far as a diamond pickaxe, which is where we derive the 26x figure.
ROGΞR retweeted
Ads seem to work well. Time to scale
P.S. This is a run rate with hardware (not just subscriptions)
🎵🎵🚀
Meet the first Sound World Model from Sonilo.
Generated video can look incredible, but without the right sound, it still feels distant.
Sonilo takes a video and creates the music and sound effects it needs, timed to the scene, motion, mood, and environment.
And in head-to-head benchmarks, Sonilo outperforms leading models across both music and sound effects.
With Sound Effects 1.0, we’re bringing music and sound effects together into one sound layer for videos.
Music carries emotion.
Sound effects create presence.
Together, they make worlds feel real, immersive, and alive!
Sound Effects 1.0 is live.
#SoundWorldModel #Sonilo #AIAudio
ROGΞR retweeted
Sound Effects 1.0 from @Sonilo_music is now available on fal.
Generate realistic sound effects from video or text.
Upload a video to get a fully synced audio track, or describe the exact sound you need with a text prompt. Generated in seconds.
Most friends and colleagues who've known me for less than 20 years don't know this: I studied Electrical Engineering in undergraduate and built robots during my master's. But hardware wasn't where the growth was in 2006–2020 — so like many of my generation, I pivoted to software and models only.
Then at Snap between 2015 and 2017, I got a front-row seat to one of the boldest hardware bets in consumer tech. I didn't work on Spectacles directly, but I still remember the 2016 launch — bright yellow Snapbot vending machines dropping out of nowhere in Venice Beach, Big Sur, even the Grand Canyon, with people lining up for hours to buy camera glasses from a robot vending machine. The last time I saw hardware generate that kind of excitement was the launch of the original iPhone📱. That was creativity in the company's genes. Business-wise? A harder story. $3.5B invested over 10 years, still no clear path to a breakout product — and real pressure on the company's fundamentals.
But here's the lesson I keep coming back to: building software has become forever easier, but hardware remains hard.
Over the last two years at Embedding VC, we've been spending more of our time on physical AI — learning from our investments in early-stage companies like @dexcamera , @omidotme and Anyware Robotics, and later-stage ones like @GroqInc and @SpaceX . Hard and Deep tech takes 10–15 years. Patience isn't optional — taking on the hardest products to build is the strategy.
Going forward, we'll double down on physical AI — not by drawing a line between hardware, software, or models, but by backing companies and products that will produce orders of magnitude more high-value tokens than today's AI applications.
ROGΞR retweeted
48h after making the @withmarbleapp Curriculum open source: 900K views, 2.3K stars and 429 forks on @github. Feels like this was needed.
Following the amazing reaction to the Marble Curriculum yesterday, we've decided to make it open source 🛰️👇
Everything a child learns in primary school. 1,590 concepts. 3,221 connections across 8 subjects, from Math and Science to Computing and Life Skills. Anchored in the US and UK curriculums, standard by standard (NGSS, Common Core, DfE).
What you will find in the repo: every concept as structured JSON with its age band and the evidence a child must show to master it. Every prerequisite link marked hard or soft, with a written rationale. It's a true DAG you can compute learning paths on. Open license, you can build whatever you want with it.
Now is a unique time in history to be building in education. Getting AI and kids education right is likely one of the hardest and most important problems to crack over the next decade and we need as many smart and creative minds behind it.
We think a common solid basis, accessible to all and that can be built upon, is critical to move fast. That's why we're making this curriculum open source.
It's not perfect but we know it's a robust basis, and we believe that sharing it openly is the fastest way to progress in this field. If you're building in education, share this around you and tell us in comments if you find this useful and if you want to contribute.
We'll keep working and investing on it @withmarbleapp. Credit goes to @guillaume_boni for building this. I just made it look pretty.
Links below 👇
ROGΞR retweeted
Following the amazing reaction to the Marble Curriculum yesterday, we've decided to make it open source 🛰️👇
Everything a child learns in primary school. 1,590 concepts. 3,221 connections across 8 subjects, from Math and Science to Computing and Life Skills. Anchored in the US and UK curriculums, standard by standard (NGSS, Common Core, DfE).
What you will find in the repo: every concept as structured JSON with its age band and the evidence a child must show to master it. Every prerequisite link marked hard or soft, with a written rationale. It's a true DAG you can compute learning paths on. Open license, you can build whatever you want with it.
Now is a unique time in history to be building in education. Getting AI and kids education right is likely one of the hardest and most important problems to crack over the next decade and we need as many smart and creative minds behind it.
We think a common solid basis, accessible to all and that can be built upon, is critical to move fast. That's why we're making this curriculum open source.
It's not perfect but we know it's a robust basis, and we believe that sharing it openly is the fastest way to progress in this field. If you're building in education, share this around you and tell us in comments if you find this useful and if you want to contribute.
We'll keep working and investing on it @withmarbleapp. Credit goes to @guillaume_boni for building this. I just made it look pretty.
Links below 👇
US gov just barred foreign nationals from Anthropic’s strongest models — Fable 5 & Mythos 5 — on national-security grounds. Pulled for everyone overnight.
The investor question that stuck with me: if you can’t reliably access the best model, does your value just compound slower than those who can?
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
$643M for a 20-person team. ~$32M per person. 💰
The Nebius–Eigen AI deal proves it: in AI, the team is the asset.
It's why we incubated Base to Base Recruiting at Embedding Labs — connecting founders with senior, scarce talent through our network. First case study 👇
How do you assemble a small team of PhDs, build another AI startup, and get acquired for $643M in less than a year?
I traced Eigen AI's ten-month journey one hire at a time.
It was not a startup. It was a strike team — every person mapped to a specific piece of the inference efficiency stack.
The real asset was not pedigree. It was compressed trust: shared labs, shared codebases, and people who had already built the tools others depend on.
The question worth asking is not "Can you hire top PhDs?" It is "Can you find the bottleneck valuable enough and assemble the small group who already owns it?"
Full research: basetobase.jobs/research/eig…
Amazing 🧪
Cancer is cured by AI.
GitLab founder Sid Sijbrandij was diagnosed with stage 4 spinal cancer. Every trial rejected him.
His doctors had nothing left to offer.
So he stopped being a patient.
He built an AI research team. Fed them 25TB of his own medical data genomics, scans, treatment history, everything.
The system found a treatment his entire oncology team had missed.
Then engineered 19 custom vaccines from his own DNA.
Relapse-free since 2025.
Then he uploaded the entire blueprint. Free. For every person sitting in that same room, hearing the same verdict, with nobody left to call.
Medicine runs on averages. AI runs on you.
More skiing wisdom for startup life as we wrapped up the 2025 - 2026 ski season.
Defense vs Offense:
In moguls, beginners default to survival mode — braking, side-slipping, traversing. It feels safer. But defensive skiing actually makes you go faster sideways and kills your ability to turn.
Same in startups. When things get rough — bad quarter, market shift, team drama — the instinct is to play defense. Cut everything. Freeze. Wait it out.
But defense has side effects. You lose momentum, miss the next turn, and the terrain starts dictating your line.
The best skiers stay in a fighting stance — especially on the hardest terrain. Active, engaged, always initiating. Best founders do the same. 🎿
I spent my prime 4 years at. The building memories and creative experiences were great, but I leave with disappointment — in the lost momentum, and in a culture and organizational structure that hasn't evolved to support the company's ability to continue scaling. Unfocused bets, slow execution, and missed opportunities in monetization and AI cost Snap its edge.
Snap has produced incredible talent — many became top founders. Some left because they couldn't fulfill their ambitions after the company lost its moment, but many brilliant people still remain who believe in what Snap can be.
The social network is a great asset. With close to 500M DAUs and a deeply engaged young user base, the foundation is strong. As savesnapnow.com makes clear, Snap can be a double AI winner — through better efficiency, monetization, and AI deployment. The potential to 7x is real.
This is Snap's crucible moment. Time to make Snap glorious again.
Some thoughts on Snap: savesnapnow.com/
One piece of startup advice I give comes from ski coaching ⛷️: when the slope gets steep, your instinct is to lean back. But that’s exactly how you lose control. Commit forward. Trust your edges. That’s where the turns happen.
ROGΞR retweeted
Congrats @luoj! 🚀 @ShiplightAI addresses a big pain point in software development. Reliable AI-driven testing will undoubtedly improve release cycles and quality. Can’t wait to see how it performs in real-world applications!
ROGΞR retweeted
this is actually insane
> be tech guy in australia
> adopt cancer riddled rescue dog, months to live
> not_going_to_give_you_up.mp4
> pay $3,000 to sequence her tumor DNA
> feed it to ChatGPT and AlphaFold
> zero background in biology
> identify mutated proteins, match them to drug targets
> design a custom mRNA cancer vaccine from scratch
> genomics professor is “gobsmacked” that some puppy lover did this on his own
> need ethics approval to administer it
> red tape takes longer than designing the vaccine
> 3 months, finally approved
> drive 10 hours to get rosie her first injection
> tumor halves
> coat gets glossy again
> dog is alive and happy
> professor: “if we can do this for a dog, why aren’t we rolling this out to humans?”
one man with a chatbot, and $3,000 just outperformed the entire pharmaceutical discovery pipeline.
we are going to cure so many diseases.
I dont think people realize how good things are going to get
Congrats on the launch, team 🚀！
🚀 Today we’re coming out of stealth and unveiling @ShiplightAI, the most reliable AI-native testing platform that scales E2E test coverage with near-zero maintenance.
For years, software teams have accepted a painful tradeoff: Move fast and risk breaking production, or slow down under the weight of brittle tests and endless QA maintenance.
AI-assisted coding makes this gap even wider. Modern teams ship faster than ever. Products change daily. Yet testing is still stuck in a world of scripts, manual upkeep, and fragile workflows that collapse the moment your UI changes, making testing the new bottleneck.
Feng and I built @ShiplightAI to make testing a force multiplier for modern software teams. As AI accelerates coding, autonomous, reliable testing closes the loop from code to confident releases.
Our goal is simple but ambitious:
Make test coverage effortless, and teams can ship with confidence anytime.
This mission is personal for us. Throughout our careers, we’ve always been obsessed with building high-quality software that lasts.
We’re already trusted by some of the most innovative, fast-moving teams. Seeing Shiplight earn their trust and transform their release cycles is the strongest signal that we’re building something real.
We’re grateful to be backed by Pear VC (Shravan Reddy, Mar Hershenson) and Embedding VC (Roger Luo), along with incredible angels like Oliver Jung.
We’re just getting started, and we can’t wait to partner with more teams who are building fast and care deeply about quality.
👉 Book a demo: shiplight.ai/demo
#AI #SoftwareEngineering #QAAutomation #ShiplightAI #AgenticAI