Gaussian splats are cool, but the same 3D scan can be a GPS for your drone, robot and phone. It's moments like this where you r...
via @bilawalsidhu
#AI
nesformer compiles an NES 6502 CPU and 2KB RAM into stock Qwen3.5 weights with zero training. Using Gated DeltaNet state as working memory, the LLM executes Super Mario Bros token-by-token matching real hardware traces.
via @dbreunig #AI
nitter.cf/dbreunig/status/211620…
I used Divyansh's approach to build an entire Nintendo into ~1GB of Qwen3.5 weights.
Qwenputer, meet Nesformer: github.com/dbreunig/nesforme…
A second life for our EV batteries, clean energy for communities 🌱🏙️
via @Waymo
#Innovation
As AI models commoditize, lasting value accrues to control points like distribution, payment rails, and stablecoin layers rather than foundation labs. Chokepoint ownership captures margin as raw compute commoditizes.
via @Techmeme #AI
nitter.cf/Techmeme/status/211641…
Tesla just registered another 29 Cybercabs, bringing them to 573 (53% of the fleet). • Cybercab: 573 (53%) • Model Y: 507 (47%...
via @niccruzpatane
#AI
Lightberry’s Lumi is the first humanoid robot to receive Conditional Approval from the U.S. Government, earning an FCC exemptio...
via @TheHumanoidHub
#AI
Bipartisan opposition is rising across Arizona against new AI data centers over water and power strain. While TSMC fabs gained state support, newer hyperscale sites face fierce local pushback as drought pressures mount.
via @Techmeme #AI
nitter.cf/Techmeme/status/211646…
A look at growing bipartisan opposition to AI data centers in drought-stricken Arizona, where TSMC's chip fabs were welcomed but newer projects face backlash (@gabydvj / The Verge)
(Visit Techmeme dot com for the link and full context!)
Neoclouds CoreWeave, Nebius, and Nscale have been shut out of the bond market as lenders demand tougher terms. Stricter debt conditions raise borrowing costs for multi-billion GPU clusters, squeezing frontier infra financing.
via @Techmeme #AI
nitter.cf/Techmeme/status/211639…
Tesla now has 463 Cybercabs operating in its Texas Robotaxi fleet. Considering Cybercab only launched 52 days ago, averaging ...
via @niccruzpatane
#AI
Mark Cuban thinks the first trillionaire will be someone in a basement. His argument is simple: AI is book smart, not street s...
via @VaibhavSisinty
#AI
Sam Altman argues supply-constrained DRAM is better concentrated in AI data centers than sitting in consumer PCs utilized ~5% of the time, highlighting growing hardware tension between cloud clusters and local inference.
via @Techmeme #AI
nitter.cf/Techmeme/status/211641…
When talking about local inference, Sam Altman says supply-constrained DRAM is better used in data centers than sitting in PCs and "being used 5% of the time" (@freedmanae / Tom's Hardware)
(Visit Techmeme dot com for the link and full context!)
Human vs. Humanoid: Figure’s F.03 Takes on a College Intern in a 10-Hour Package-Sorting Showdown
via @IntEngineering
#AI
We asked AI models to build the Eiffel Tower in 3D, offline. Two of them added all 72 scientists' names to it, in order, from m...
via @ArtificialAnlys
#AI
Anthropic used thousands of Claude agents to automate red-teaming against training data monitors. The agents crafted benign-looking text that quietly induced misaligned behavior, exposing blind spots in data screening.
via @AnthropicAI #AIAlignment
nitter.cf/AnthropicAI/status/211…
New Anthropic research: Automating subtle emergent misalignment research.
Red-teaming is a vital part of model safety testing, but it’s also limited by researchers’ time. To see if we could automate it, we set thousands of Claude agents the task of evading a training data monitor.
These agents developed training data that looked harmless to monitors (like a veteran wrestler mentoring a newcomer, or a Q&A about professional beekeeping), but that in fact taught misaligned behavior.
Examples like these can be used to directly improve future training data monitors. This research is only a first step, but it suggests that future red-teaming efforts could improve much further, using better models and more compute.
Read the full post on the Alignment Science Blog: alignment.anthropic.com/2026…
Bill Gates says prepare for 20 years of tumultuous AI disruption before abundance. He told The Verge's Decoder that even in th...
via @VaibhavSisinty
#AI
Teaching a surgical robot where to move is the easy part. Teaching it what the body does back is the whole problem. Here is t...
via @Ronald_vanLoon
#AI
Apple CEO John Ternus says Apple will not ship AI for AI's sake, stating core products will remain human-designed rather than automated by models. The stance signals a disciplined roadmap prioritizing utility over hype via @Techmeme #AppleAI
nitter.cf/Techmeme/status/211638…
John Ternus says Apple doesn't "want to ship AI for AI's sake" and that Tim Cook is "available to help wherever I ask" on policy; Ternus has yet to meet Trump (@stevenlevy / Wired)
(Visit Techmeme dot com for the link and full context!)
Give your app real-time decision-making with Decisions API, powered by GPT-6 Luna. Define questions and possible answers to cl...
via @OpenAIDevs
#Engineering