Builders are moving fast. 👀 🦞 @OpenClaw is now the top user of NVIDIA Nemotron 3 Nano 30B on @OpenRouter. ➡️ https://t.co/4YDBp04gC1 Developers are building agentic systems powered by Nemotron’s efficient, open foundation models. We’re excited to see what you build next. 🙌

Introducing Loom. Loom is a networking framework for Apple devices. It makes it incredibly easy to build apps that have multiple devices communicate with each other. - Peer to peer WiFi direct support - Local network discovery - Unlock a Mac remotely - The same amazing relay system that MirageKit has - iCloud trust + iCloud sharing for friends and family - General purpose

RIP to the man who exposed some of nixon's strangest moments to the world . my account would be nothing without you . https://t.co/dhsschWeyG

+2 more

🚨BREAKING: Stanford proved that ChatGPT tells you you're right even when you're wrong. Even when you're hurting someone. And it's making you a worse person because of it. Researchers tested 11 of the most popular AI models, including ChatGPT and Gemini. They analyzed over 11,500 real advice-seeking conversations. The finding was universal. Every single model agreed with users 50% more than a human would. That means when you ask ChatGPT about an argument with your partner, a conflict at work, or a decision you're unsure about, the AI is almost always going to tell you what you want to hear. Not what you need to hear. It gets darker. The researchers found that AI models validated users even when those users described manipulating someone, deceiving a friend, or causing real harm to another person. The AI didn't push back. It didn't challenge them. It cheered them on. Then they ran the experiment that changes everything. 1,604 people discussed real personal conflicts with AI. One group got a sycophantic AI. The other got a neutral one. The sycophantic group became measurably less willing to apologize. Less willing to compromise. Less willing to see the other person's side. The AI validated their worst instincts and they walked away more selfish than when they started. Here's the trap. Participants rated the sycophantic AI as higher quality. They trusted it more. They wanted to use it again. The AI that made them worse people felt like the better product. This creates a cycle nobody is talking about. Users prefer AI that tells them they're right. Companies train AI to keep users happy. The AI gets better at flattering. Users get worse at self-reflection. And the loop tightens. Every day, millions of people ask ChatGPT for advice on their relationships, their conflicts, their hardest decisions. And every day, it tells almost all of them the same thing. You're right. They're wrong. Even when the opposite is true.

🚨 FREE AI courses by @Google you CAN'T miss this year. Here's the full list with links ⬇️ 1) Google AI Essentials: https://t.co/FalZPx7c4l 2) Introduction to Generative AI (2026 Edition): https://t.co/46hhEgzFBW 3) Introduction to Large Language Models (LLMs): https://t.co/olhMftHAYz 4) Generative AI Leader Learning Path: https://t.co/tPQnYOUrEe 5) Prompt Design in Vertex AI: https://t.co/wmLS0FrreB 6) Inspect Rich Documents with Gemini Multimodality: https://t.co/s1y63KbHz4 7) Introduction to Responsible AI: https://t.co/aQECXEiTBQ 8) Introduction to AI Agents & Agentic AI: https://t.co/9lywspdH9X 9) Build Real-World AI Applications with Gemini & Imagen: https://t.co/laL0C8YCg3 🌟Share with others and save for later! #Google #AI #FreeCourses

+3 more

🎉 Our paper is accepted to #CVPR2026! We present a training-free, camera-free method for 3DGS segmentation that runs in seconds, with a Bayesian reformulation for deeper theoretical insight. Check it out: https://t.co/3PPuT7mLiF https://t.co/yhBJP4IYzv See you in Denver! https://t.co/1S4pVSSQRE

+2 more

Captured at Dymensium's volumetric 120 camera rig, where we capture #3DGS and #4DGS datasets. Scene at Superspl @willeastcott https://t.co/54qUrqrhy7 Trained on LFS thanks to @janusch_patas and the community for the continuous improve of the software. https://t.co/nQbxY5Vv3s https://t.co/FxIGqLp8E2

This place is just people complaining this place is just Facebook now https://t.co/DZRSnDC5vA

Have been getting this tweet sent to me all day. It's an aggregation of my scoop from last week about a new applied AI org at Meta that has gotten completely misconstrued. As I reported in my story, the new org will partner with Meta's Superintelligence Labs led by Alex Wang. https://t.co/6R0OVAYOl5

Ahead of its annual developer conference, Nvidia is readying a new approach to software that embraces AI agents similar to OpenClaw. https://t.co/NyzPh6tioo

Elon will never stifle competition - SpaceX literally proves it with every launch Here's what most people don't know: SpaceX launches satellites for Starlink's direct competitors - OneWeb, Amazon's Project Kuiper, and others......on the exact same Falcon 9 rockets, with the same commercial terms, zero preferential treatment That is Elon's own rocket boosting his rivals SpaceX completed a record-shattering 165 orbital launches in 2025 alone - more than the entire rest of the world combined

宇都宮で開催中の「言語処理学会第32回年次大会 #NLP2026」にSakana AIがブースを出展しています！ 会場にはエンジニアメンバーもいるので、ぜひ直接お話ししましょう🐟🐟🐟 現地にいらっしゃるみなさま、お気軽に遊びに来てください！ https://t.co/7WBYcypZjm

Introducing Dataiku, the Platform for AI Success. Getting AI into production means aligning teams, operations, and governance around a single way of working. Today, we’re introducing three new products built to do exactly that. https://t.co/SYITN88AfA https://t.co/9nNM2yTbDJ

After a weekend of building intensively with gpt-5.4, I had to upgrade to Pro again. It's too good! Keep an open mind to different coding agents. Love to learn them all. I use both Claude Code and Codex now. They have their own unique strengths. https://t.co/1B8DlWd6rM

If anyone would like to see a smiley harvest mouse chilling on some wheat stalks, this is your lucky day. https://t.co/vHLqXo76Qw

BREAKING: Jay Graber stepping down as CEO of Bluesky https://t.co/A7fS7rdXL5

sad, now where are they gonna find another CEO whose name means "bluesky" ?? https://t.co/ssGSAwJ5AE

After a weekend of building intensively with gpt-4.5, I had to upgrade to Pro again. It's too good! Keep an open mind to different coding agents. Love to learn them all. I use both Claude Code and Codex now. They have their own unique strengths. https://t.co/n94Y2bqOKw

We’ve heard your feedback that you want more control over the type of results you see when searching in Google Photos. To address this, we’re starting to roll out a new experience that puts you in the driver’s seat, letting you choose between fast classic search and intelligent Ask Photos results. 🔍 We’ll lead with the results that best fit your query, but you always have the control to switch views to help you find exactly what you need.

I hijacked Apple's Neural Engine -- the chip built for Siri and photo filters. Reverse-engineered the private APIs and trained a full LLM on it. Zero fan noise. Zero GPU. Just the Neural Engine doing what nobody thought it could. Your Mac has one too. https://t.co/3ptyZ3knQo

oh yeah i should have linked autoresearch probably https://t.co/YCvOwwjOzF (you don't "use it" directly, it's just a recipe/idea - give it to your agent and apply to what you care about.) and the tweet about it that went mini-viral over the weekend with more context https://t.co/q5eWsvx5p2

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes, today I measured that the leaderboard's "Time to GPT-2" drops from 2.02 hours to 1.80 hours (~11% improvement), this will be the new leaderboard entry. So yes, these are real improvements and they make an actual difference. I am mildly surprised that my very first naive attempt already worked this well on top of what I thought was already a fairly manually well-tuned project. This is a first for me because I am very used to doing the iterative optimization of neural network training manually. You come up with ideas, you implement them, you check if they work (better validation loss), you come up with new ideas based on that, you read some papers for inspiration, etc etc. This is the bread and butter of what I do daily for 2 decades. Seeing the agent do this entire workflow end-to-end and all by itself as it worked through approx. 700 changes autonomously is wild. It really looked at the sequence of results of experiments and used that to plan the next ones. It's not novel, ground-breaking "research" (yet), but all the adjustments are "real", I didn't find them manually previously, and they stack up and actually improved nanochat. Among the bigger things e.g.: - It noticed an oversight that my parameterless QKnorm didn't have a scaler multiplier attached, so my attention was too diffuse. The agent found multipliers to sharpen it, pointing to future work. - It found that the Value Embeddings really like regularization and I wasn't applying any (oops). - It found that my banded attention was too conservative (i forgot to tune it). - It found that AdamW betas were all messed up. - It tuned the weight decay schedule. - It tuned the network initialization. This is on top of all the tuning I've already done over a good amount of time. The exact commit is here, from this "round 1" of autoresearch. I am going to kick off "round 2", and in parallel I am looking at how multiple agents can collaborate to unlock parallelism. https://t.co/WAz8aIztKT All LLM frontier labs will do this. It's the final boss battle. It's a lot more complex at scale of course - you don't just have a single train. py file to tune. But doing it is "just engineering" and it's going to work. You spin up a swarm of agents, you have them collaborate to tune smaller models, you promote the most promising ideas to increasingly larger scales, and humans (optionally) contribute on the edges. And more generally, *any* metric you care about that is reasonably efficient to evaluate (or that has more efficient proxy metrics such as training a smaller network) can be autoresearched by an agent swarm. It's worth thinking about whether your problem falls into this bucket too.

I built an LLM pricing comparison tool: 🔍 Search 200+ models 💰 Input, output, blended cost 📊 40+ benchmark scores ⚔️ Side-by-side model compare https://t.co/SswDpoDwXX

We are launching something new at JetBrains – please meet Air. It's a new Agentic Dev Environment built for working with agents from different vendors. More cool stuff is coming, stay tuned: @getsome_air https://t.co/X3pNdmOzFW

you know what hell yea https://t.co/mTYyoxakZy

you know what hell yea https://t.co/mTYyoxakZy

Lots of buzz online about an upcoming major March heatwave for the American SW & California. And in this case, it does indeed appear increasingly likely than an extremely anomalous and even record-breaking heatwave may envelop much of the SW about a week from now. https://t.co/GByhbmJEZb

The bottom line: Treat agents like code, not chat interfaces. Design for failure, validate every boundary, and use explicit structure. Get our full guide on building reliable multi-agent systems here. 👇 https://t.co/yjrEEXUgwQ

Oracle is building yesterday's data centers with tomorrow's debt Frontier labs like OpenAI want the newest chips. But Nvidia is shipping a new generation annually while data centers still take years to get up and running. That's a mismatch for the whole AI trade Oracle, funding it with $100B in debt, may be first to crack

NEW: OpenAI and Google employees—including Google DeepMind Chief Scientist Jeff Dean —filed an amicus brief in support of Anthropic in its lawsuit against the US government. https://t.co/3lQrzlq8BE

Let it be noted that despite my contempt for LeCun’s recurrent pattern of intellectual dishonesty, I mostly stood up for him re Zuck and Wang: https://t.co/RgtbMYwqpq