What is sol doing here
Video tampering can mislead every frontier model we tested into harmful actions, including a 100% attempt rate on stabbing a baby doll, producing toxic gas, and pouring water into a toaster.
We're presenting at CoRL Late Breaking Results next week.
🧵 You Can't See Everything: Perceptual Attacks on Embodied AI
Yang Guo retweeted
At the end of Q3, Alan was at ~€880m of signed ARR. ~1.2m members. 3 continents. And we’re on track to cross €1 billion of signed ARR by the end of the year.
Q3 was a quarter of foundations. Not the flashy kind. The kind that makes the next 10 years possible.
1. We entered a third continent. With the acquisition of Tanel, we now cover ~70,000 people in Senegal and Côte d’Ivoire, under a new brand: Afiyah by Alan. And we received pre-license approval in Saudi Arabia.
2. We launched the biggest brand campaign in our history. Health Can’t Wait, with Kylian Mbappé. One conviction: better health starts with prevention.
3. Mo became the front door. Our AI health companion now answers medical, insurance and support questions in one place, and guides members to the right care. A new onboarding doubled the share of members who activate it.
4. Prevention stopped being a slogan. Alan Precision is live: 500+ members have done it, 17,000+ are on the waitlist. Prévenir, the occupational health service, now covers 35,000+ employees.
5. AI runs through everything. Our AI assistant resolves more than half of care contacts in France. Our email agent fully resolved 1 in 5 emails in September, with no human in the loop. Hopper, our in-house AI platform, is used by 80% of Alaners and writes 25%+ of the code we ship.
Healthcare systems are built to treat. We’re building one designed to prevent.
Q4 is where foundations turn into momentum. 1,000+ Alaners are on it.
This is true. The gap between what is possible and what is being done is still widening
a really interesting new type of PE / VC / opco hybrid is going to emerge over the next 5 years that basically buys services companies and automates 98% of the work. It will be a massive industry.
Think extremely well capitalised AI rollups. They'll become multi-sector because it doesn't really matter what service you're rolling up, the systems you need to build are very similar.
This is basically what firms like thrive and general catalyst are already doing.
Midwit PE firms that were just financially engineering upside don't have the in-house skillset to do this. You need a combination of people good at talking to instos and raising capital AND people who know how to ship excellent product very fast with AI.
i think there will be a golden period of a couple of years where you can run these services companies at extremely high margin and profit, but the risk is that then the world will start to catch up and margins in every services business will erode unless they have a compelling moat eg. network effects.
the upside is that the consumer will get a GREAT deal, with tons of services becoming much better, much faster AND much cheaper, all simultaneously
6
819
All labs have realized StarCraft is how u train intelligence lol. Same lessons hedge funds learned in early/mid 2010s
Rumors I’ve been hearing. I’ll start with OpenAI, then Anthropic.
OpenAI: The internal model’s reasoning jump comes from more aggressive RL, including training in games and simulated environments. Strategies learned there transfer to research mathematics and other fields. Navier–Stokes and ETH show the scale of the improvement. The clue is in the Synthetic RL hiring: self-play, simulators, and synthetic feedback.
Anthropic: Same direction with its internal Claude model. It’s trained on long, difficult tasks in demanding environments, and the reasoning skills transfer to other fields. Its ETH result shows how far the improvement has gone. The earlier Universes listing and the RL hiring around mathematical reasoning align with that.
Musk’s Digital Optimus hiring explicitly calls for measuring transfer from games into general computer use. That’s why his game-playing AI push matters: games develop abilities that work far beyond the games themselves.
The interesting part is where those abilities show up. Each lab’s internal model applies reasoning developed through interactive training to mathematical problems that had resisted decades of work.
8
1,166
Yang Guo retweeted
🗞️ if your team is too busy doing their 'normal job' to experiment with AI, you're preparing them to be replaced
ghuntley.com/replaced/
In this post, I guess this's the last time I'll say it. AI use is no longer optional if you wish to be employed. What follows is advice for current employees and hiring managers on how to hire in 2026 to source AI-first candidates.
This
“If we truly believe that AI is changing how we work, then ICs should get paid more than managers, because an IC can drive more impact than a manager could in the pre-AI system.” — @ElenaVerna
This is absolutely true. If you’re not willing to make org changes, impact is usually very small
I am gonna say something risky:
Most of the FDEs I've seen are solving BS problems.
1. Existing enterprise workflows were designed around people, not agents.
2. Real AI gains require rebuilding the org chart.
3. Most execs have no incentive to do that, and most CEOs aren't in founder mode.
4. So FDEs automate the status quo, because that's what's easy to sell. But the status quo is stupid and will be obsolete.
5. If you FDE on critical workflows, 1) enterprises are really this dumb to hand off trade secrets? 2) if you know that, shouldn't you start a competing firm?
If you FDE on non-essential AND soon-obselete workflows, did you learn anything meaningful and repeatable?
6. If you are a smart young FDE, why spend your best years automating dying workflows and managing people who fear AI will take their jobs? Shouldn't you start an agent-native firm from scratch in that industry instead?
Analogy would be, instead of teaching everyone to be AI builders within a tech company, you as FDE build agents that auto-ingest meeting notes for PM, QA, front-end, back-end, devops, UX and help each of them write PRD, Figma, and meeting notes faster for their sync meeting, but the right solution should be to shrink roles, hire full-stack AI builders, and kill meetings. You gain lots of experience with your PRD/Figma/standup meeting sync agents but it's a useless skill.
I am not a FDE expert so I might oversimplify things. I am genuinely curious about this topic and welcome any critiques and comments.
4
740
I love that the way we measure intelligence now is through Pokemon and StarCraft - this is real peak intelligence
Over the next year successful cos will do 2 things:
- Commingle token and headcount budgets
- Incentivize a small/medium preference to tokens, because tokens deflates while headcount inflates
The skill will be in how the incentivization is designed.
Internally we’ve seen that this starts very nonintuitive but with the right culture/incentive design ppl move to this very quickly
1
17
1,363
TLDR; commoditized
In 4 months, the concept of "SWE" will be completely obsolete. I think there are a few things that will happen
Coding ability will for the first time exceed that of any human on the planet
Such that any human opinion, besides the initial prompt, will just make it worse. There will be no value in engineering discipline anymore
The difference between L8 and L3 is not going to be engineering knowledge
Maybe it won't be in 4 months, maybe it'll be 6 months, but there will be a point where your engineering experience will be worth 0
So the question is just: what are the things that will still matter?
Cost of intelligence will still lag capability, so despite us eventually having access to a magic button that can solve any engineering problem, it's still finite
So what you'll need to evaluate is people's ability to stack rank the most important business tasks and allocate tokens efficiently
The ones that just turn Astra / Fable to max and let it rip, these are the new L3s.
The ones that know where to apply Jev and Opus 5.5 to maximize business outcomes are the new L8s
4
918
🤔
Checking in here on the AI apocalypse:
- incredible advances across math, science and medicine at unfathomable rates
- data centers creating skilled labor jobs - effectively rebooting the middle class and US manufacturing
- electricity needs driving massive investment to modernize and reinforce our aging electrical grid
- AI unlocking GDP growth by revitalizing sector after sector
- driving exceptional innovation at all layers of the stack and economy
- upskilling millions of workers with superpowers
- pushing geopolitical advantage to the west
Reads like an apocalypse
5
1,052






















