Daily Brief

A good exec search takes 6 to 12 months. Ali Ghodsi on why people settle, why the signals were always there, and why backdoor references beat front doors.

Feed a model one hour of video and roughly a million visual tokens go in, yet the loss lands on about two percent of them, since the only ground truth is a transcript or a few labeled frames. Armen Aghajanyan calls that a humongous waste, and predicting every pixel treats a background pixel with the same weight as a gripper tip or a contact point. Perceptron's answer is a perceptive objective that learns which percepts will matter, rather than hardcoding the gripper. The second problem is context bloat from cameras that never switch off. Patch averaging buys ten times compression, but his fix

Well under one percent of the Common Crawl corpus that frontier models train on is in an Indian language, even as frontier labs call India a fast growing market. Krishna Prasad Srinivasan says the knowledge exists but was never digitized, and Sarvam's answer is a three billion parameter vision language model, small enough for one GPU, that he says beats document AI models a hundred times larger. The model is unusual twice over. Its backbone is a state space model rather than a transformer, because a page can run to ten thousand visual tokens and quadratic attention gets expensive, while an SSM

Full episode: https://www.youtube.com/watch?v=6AgOfiZOWiY Me on twitter: https://x.com/dwarkesh_sp

Ali Ghodsi built a connector in two days. His team was taking three quarters. What he learned about where the time actually goes, and how they got to seven connectors a quarter.

A generalist robot that succeeds 80 to 90% of the time makes a great video but a poor product. Jason Ma, co-founder and CTO of Dyna Robotics, shows how Dyna got its model, Dyna-1, to a 99.4% success rate folding restaurant napkins for 24 hours straight, including recovering after it pulled over the whole stack. The key is a reward model that watches the robot and scores its progress. When the score dips, the robot is making a mistake, so the team can collect targeted recovery data and fine-tune again in a human-in-the-loop active-learning cycle. Ma also covers Dyna's pre-training data pyramid

The models are ready. The organizations aren't. Ali Ghodsi on why diffusing AI means re-engineering the processes of every company on the planet, and why that takes a decade at least.

Suchet Bargoti opens with a live demo instead of slides. From a laptop on conference Wi-Fi, he launches a docked drone over San Mateo, starts a second one on a power line in Colorado, has one autonomously track a car, and then sends the whole fleet back to dock. Skydio, the largest US drone manufacturer, now has thousands of these docks deployed with utilities, police and construction companies, and about 16 million people live within two miles of one. Bargoti, Skydio's Director of Inspection and Mapping, explains the shift to "drones as infrastructure." He shows a drone spotting a utility po

The first thing it did was breathe. Cyrus Clarke connected an OpenClaw agent to a 900-pin shape display at the MIT Media Lab and, instead of giving it tasks, asked it to discover who it is. It raised and lowered the whole grid like breathing, reached for its own edges, and spelled "HI CYRUS" in pins. The video of that first day reached 15 million views, with reactions that swung from awe to calls for him to stop. Clarke explains why he chose a body with no face, no limbs and no instruction manual, and what he took from object-oriented ontology, the original Greek sense of aesthetics as sensor

Live from STATION F, the AI Engineer Paris 2026 Main Stage continues with a full day of technical talks from the teams building and scaling production AI systems. Watch sessions covering model customization, inference infrastructure, voice and realtime AI, coding agents, software factories, and the emerging AI economy. Day 2 main-stage lineup: • Jakob Pörschmann (Black Forest Labs) — From Pixels to Robots • Dorian Lods (ElevenLabs) — Generating Identity at Scale • Harry Mellor (Hugging Face) — How a Transformers Model Loads in vLLM • Arielle Le Bail (Stripe) — Inside the AI Economy: What St

Full episode: https://www.youtube.com/watch?v=6AgOfiZOWiY Me on twitter: https://x.com/dwarkesh_sp

What happens when AI stops merely *reading* DNA and starts learning how to *write* it? Eric Nguyen, CEO and co-founder of Radical Numerics and a key figure behind Evo, joins us to explain the leap from early genome language models to Omni—a system that can outperform specialized models on disease-variant prediction, reason across biological sequences, and even begin to “show its work” when designing better molecules. The surprising part is *where those capabilities seem to be coming from*: in several cases, the model learns useful biological structure it was never explicitly taught. But the s

近日某起海外犯罪案件引起不少人的关注，有朋友非常生气，认为国外法院判的太轻，所以问了一个问题，如果这些人服刑之后回到国内，我国的法院还可以对他们再审理吗？