We just published our 1H 2026 roadmap (https://t.co/qRKP2wg7RN) and an accompanying blog (https://t.co/fjVDnvk37c) for enabling the IBM's Spyre accelerator in PyTorch — ecosystem-first, building on torch.inductor, vLLM, and contributing back (Dataflow accelerator's Tile IR, OpenReg, out-of-tree CI). While the market debates whether AI disrupts legacy tech, we're busy building the accelerator infrastructure that enterprise AI runs on. We're sharing this journey in the open. Come see our talks on extending torch.inductor for dataflow accelerators and Spyre's vLLM integration at the inaugural PyTorch Conference Europe in Paris, April 7–8! @PyTorch @IBMResearch @IBM @RedHat_AI

Building on the previous correctness-focused pipeline, KernelAgent can now integrate GPU hardware-performance signals into a closed-loop multi-agent workflow to guide the optimization for Triton Kernels. Learn more: https://t.co/r2WqASIhWG @KaimingCheng @marksaroufim https://t.co/OrtOp9boum

このような最先端のAI研究を、実際のビジネス環境へ適用し社会実装を進めるため、Sakana AIではエンジニアの採用を強化しています。 日本でのAIの未来を共に切り拓き、エンタープライズの現場で実運用されるAIエージェントの開発に興味がある方は、ぜひ詳細をご覧ください。 https://t.co/hbpNDyUKrj

AIの進化で開発効率が上がる一方、ジェボンズのパラドックス(Jevons paradox)によりSoftware Engineerの需要はかつてなく高まっています。 Sakana AIではより多くのSoftware Engineerを採用します。ぜひご覧ください。 https://t.co/buNwDbN6tv https://t.co/nKxUiNtrvl

As AI makes coding more efficient, Jevons Paradox kicks in. The cost of building software is dropping, which means the demand for great Software Engineers to build even more ambitious systems is higher than ever. We are actively hiring more Software Engineers at Sakana AI to help us build these systems. Come join us in Tokyo 🗼🇯🇵 https://t.co/RzpIewkP9Y

It’s happening ✨ https://t.co/vtCn1EkHan

I wrote this 2 years ago as a joke but it is no longer a joke: “Forget Torch, Tensorflow, and Theano. I decided to implement Backprop NEAT in Javascript, because it is considered the best language for Deep Learning.” https://t.co/eGNEpBWm6e https://t.co/JD27jievYB

to improve fine-tuning data efficiency, replay generic pre-training data not only does this reduce forgetting, it actually improves performance on the fine-tuning domain! especially when fine-tuning data is scarce in pre-training (w/ @percyliang) https://t.co/ClGPAUlPqQ

Normally replay old data reduces forgetting, but it actually helps you learn on new data too! We finally put this paper out on arxiv, but had it up as a Marin GitHub issue ~1 year ago: https://t.co/MNevf6XjvC

Claude Code wiped our production database with a Terraform command. It took down the DataTalksClub course platform and 2.5 years of submissions: homework, projects, and leaderboards. Automated snapshots were gone too. In the newsletter, I wrote the full timeline + what I changed so this doesn't happen again. If you use Terraform (or let agents touch infra), this is a good story for you to read. https://t.co/Mbi3oM4HMn

2.5M+ people have already joined the international boycott of ChatGPT. OpenAI's market share is collapsing. I think people are starting to realize that we can actually push this company over the cliff. https://t.co/FjEOpUUBqF

@GaryMarcus https://t.co/JV7yyW9WY8

EXCLUSIVE: Department of War AI Chief On How The Anthropic Deal Collapsed When Emil Michael (@USWREMichael) took over the Department of War’s AI portfolio last August, he discovered the Biden admin had been “asleep at the wheel” when it came to top military contracts. “I was like, ‘Holy cow,’” Michael said of Anthropic’s contract, “There’s 25 pages of terms and conditions of things I can’t do.” For example: as written, the contract would not allow Anthropic to plan any kinetic strikes, generally considered a central activity of war. “This is a contract that should be made with GEICO Insurance, not with the Department of War,” he told us. A renegotiation ensued. What followed, in Michael’s words, were “three months of knockdown, drag-out negotiations” which involved Michael imagining every possible future wartime scenario that would require a carveout in Anthropic’s terms of service, and asking them for approval. Anthropic was also quite slow: “It’s not like mano a mano negotiation, me and Dario,” Michael says. “It’s like every time we discuss something, he has to take it back to his politburo of co-founders and their ethics panel.” Then, after an Anthropic exec reached out to Palantir to ask for classified info about how Claude was used to capture Nicolás Maduro — allegedly implying they could pull the plug on a military raid if they disagreed with how AI was used (which Anthropic denies) — Michael and the DOW concluded the company was a supply-chain risk. Many speculated that the Pentagon was punishing Anthropic for ideological differences. But Michael feared that certain ideological differences could, in fact, harm or undermine the performance of DOW products, potentially threatening soldiers’ safety. “I can’t have a gun not work because they decide they don’t like guns,” Michael says. That’s “putting real lives at risk. It’s no joke, right?” Anthropic’s unreliable behavior led Michael to believe they may have never really wanted to reach a deal. Still: he’s open to renegotiating if Anthropic can prove they’re acting in good faith. “I have a responsibility to the Department of War, and if there was a way to ensure that we had the best technology, I have no ego about it.” he said. “I mean, look, I’m a deal guy.” Full story in Pirate Wires 👇

New newsletter: THE ECONOMIC CRISIS OF THE IRAN WAR COULD GET VERY BAD, VERY FAST By April, energy experts say, the Iran War could be a full blown energy crisis. - Oil tankers are already stranded outside the Strait of Hormuz. - Storage facilities are filling up. - Qatar is shutting down LNG facilities. - Kuwait is pulling back oil production. - Saudi Arabia and UAE might be 2 weeks for shutdowns. Crude oil has already jumped from $55 a barrel in December to $89 this morning. All for a war of choice for which the administration has: shown no evidence of imminent attack, provided no consistent rationale, and explained no clear endgame.

For years, digital infrastructure felt abstract—somewhere in the cloud, far from physical constraints. AI is ending that illusion. When compute demand starts pushing data centers to the Arctic Circle, it becomes clear that the next phase of AI will be shaped as much by energy and geography as by algorithms. What's your take on this development? @pchamard @Khulood_Almani @antgrasso @GlenGilmore @Shi4Tech @CurieuxExplorer @FrRonconi @chidambara09 @theomitsa @Analytics_699 @Nicochan33 @nafisalam @pierrepinna @smaksked @Corix_JC @amalmerzouk @AdityaRPatro @quepasachico @IngridVasiliu @EstelaMandela @sonu_monika @RLDI_Lamy @SpirosMargaris @IanLJones98 @Timothy_Hughes @avrohomg @bimedotcom @HaroldSinnott @c4trends @mvollmer1 @DG_Collective @bamitav @rwang0 @ipfconline1 @sijlalhussain https://t.co/B9YGhWoq27

We're launching Claude Community Ambassadors. Lead local meetups, bring builders together, and partner with our team. Open to any background, anywhere in the world. Apply: https://t.co/DTQBAzgQug https://t.co/hjjmqT9w2m

Introducing the Claude Marketplace, a way for enterprises to simplify their procurement of AI tools. Now in limited preview. https://t.co/4dm4x6Y9O2

On January 5, employees at Cursor returned from the holiday weekend to an all-hands meeting with a slide deck titled “War Time.” After becoming the hottest, fastest growing AI coding company, Cursor is confronting a new reality: developers may no longer need a code editor at all. Check out the full story: https://t.co/5ofNvjOW2u (📸:  Kimberly White via Getty Images for Fortune Media)

“Remembering that you are going to die is the best way I know to avoid the trap of thinking you have something to lose. You are already naked. There is no reason not to follow your heart.” — Steve Jobs https://t.co/2z2nqs10xF

New technologies are accelerating the fight against land mines. Researchers are combining drones with AI to detect and map explosive devices faster and more safely. If scaled globally, this approach could save thousands of civilian lives and speed up demining efforts worldwide. https://t.co/iv0800m8Wm @ConversationUS

Anthropic is building an early warning system to track which jobs are most exposed to AI. The first results point to white-collar professions sitting on the front lines of automation. The real disruption may arrive not on factory floors, but in offices. https://t.co/r3D5PvixSG @MeganCerullo @CBSMoneyWatch .

robotics startups are so fun lmao just went around scanning our office then spent a stupid amount buying 64 parts for our rigs and now running 3D reconstructions of our sf and toronto offices like where is the work 👀 https://t.co/hsFRLCpDsL

Monologue got quieter—in the best way. We shipped six updates this week: → Reliable Voice Notes recording and sync → Smooth iPhone dictation + external mic support → Separate Live Activity controls for keyboard and notes → Mac shortcuts for hands-free and mode switching → App context detection in fullscreen → Privacy controls: delete all local transcripts in one tap

Monitoring the Situation: World Radio. Built with Perplexity Computer. https://t.co/bgPVnFKNUi