Amaan

@AndrewCurran_ It’s the kind of play that could quietly accelerate agentic adoption in places that usually move slower, where persistent workflows and reliable tool use matter more than flashy demos.

Anthropic is about to announce a $1.5 billion joint venture with multiple wall street firms to sell AI tools to private-equity backed companies. Anthropic, Blackstone, Hellman & Friedman and Goldman Sachs are all major investors. The announcement is expected tomorrow morning.

@0xleegenz The moment you realize what life looks like for people who don’t know anything about AI, politics or the economy… for some reason they are more happier, just vibing!!

The moment you realize what life looks like for people who don’t know anything about AI, politics or the economy:

@Nithya_Shrii AI doesn’t kill thinking blind usage does. CAD didn’t kill engineers. Use it right, and it amplifies your brain, not replaces it.

ChatGPT is rotting your brain and killing your critical thinking skills.

@scaling01 Open models matching scores while using far more tokens isn’t true parity, latency adds up quickly in agent workflows. That said, distillation from closed APIs is rapidly advancing open models in coding and agentic tasks.

@emollick Repackaging the 2017 “Attention Is All You Need” paper in full 2026 hype format and watching half the timeline lose their minds 😭😭

(Sorry, after seeing so many of these, could not resist): 🚨 BREAKING: Google just dropped a NEW paper that completely deletes RNNs from existence. No recurrence. No convolutions. Nothing. Just one mechanism. And it’s destroying every translation benchmark on the planet. The title alone is a flex: “Attention Is All You Need” Vaswani. Shazeer. Parmar. Uszkoreit. Jones. Gomez. Kaiser. Polosukhin. 8 researchers. 1 architecture. The entire field of NLP will never be the same. Here’s why this is INSANE → LSTMs took DAYS to train. This thing trains in 12 hours on 8 GPUs. 🤯 → 28.4 BLEU on English-to-German. That’s not an improvement. That’s a MASSACRE. They beat the previous SOTA by over 2 points. → English-to-French? 41.8 BLEU. At a FRACTION of the training cost of every model that came before it. → They called it the “Transformer.” The name alone tells you they knew. But here’s the part nobody is talking about 👇 They threw out sequential processing ENTIRELY. Every other model on Earth processes words one at a time. This thing looks at the ENTIRE sentence simultaneously and figures out what matters. It’s called “self-attention” and it’s basically the model asking itself: “which words should I care about right now?” Every. Single. Token. In parallel. Do you understand what this means? Training that used to take WEEKS now takes HOURS. Models that couldn’t scale past a few layers? This thing stacks 6 encoders and 6 decoders like it’s nothing. And the multi-head attention? 8 attention heads running at once, each learning DIFFERENT relationships in the data. I’m not being dramatic when I say this paper just rewrote the rulebook. RNNs are cooked. 💀 LSTMs are cooked. 💀 The future is attention. And attention is ALL you need. Follow for more 🔔

@TheTuringPost @deepseek_ai Letting the model point with points and boxes instead of vague language is a clean fix. Hitting 77% accuracy while keeping only 80–90 tokens in memory a genuine efficiency!!!

There’s a serious gap in multimodal models – they work with images, but still reason in language, which isn’t that precise for visual stuff. @deepseek_ai just dropped an idea to solve this: let the model literally point to exact locations in the image while it thinks. They call it "Thinking with Visual Primitives." These visual primitives are: - points (specific locations) - bounding boxes (areas in the image) Using them, the model knows what exactly it’s referring to and achieves ~77% better accuracy on average (vs. Gemini 3 Flash's 76.5% and 71.1% for GPT-5.4) Plus, only ~80–90 visual tokens are kept in memory after compression thanks to the efficient architecture Here is how it works:

@bjs_alejo2 DawnBreaker!!

hombres puntúen este palo

Fianlly, AGI achieved!!

@BradSmi @Microsoft A practical step toward agents that actually fit high-stakes professional flow. Handling clause precision, redline tracking, and structured workflows while keeping the human fully in control!!

Today we’re introducing a new Legal Agent in @Microsoft Word, built to support the precision and rigor legal work demands. Every clause matters. Every redline tells a story. That’s why this agent was built to follow the structured workflows lawyers use while keeping them fully in control. Early in my career, I asked for a computer on my desk because I believed technology could change how lawyers work. It did. Today, I believe this next generation of tools will do the same, grounded in trust and responsible use.

@JackWoth98 Agreed. Curious though, do you think fully automated skill extraction is sufficient ??

@amaank_tweets Memory is key! The self-learning loop of automatically creating and refining skills in the background based on past sessions is really powerful!

@stripe @link Secure spending on your behalf with zero exposed credentials and you approve every buy, amazing!!

Today, we’re launching the @link wallet for agents. It lets you securely empower agents to spend on your behalf. Your payment credentials are never exposed and you approve every purchase. link.com/agents

@stripe Multi-currency balances, free instant transfers, email payments to 160 countries, and 2% cashback on the card. The real win though is that you can now run it directly from any AI app with Stripe MCP!!

Introducing the new Stripe Treasury: • Hold funds in multiple currencies and stablecoins. • Instantly transfer money to US businesses on Stripe for free. • Pay anyone in 160 countries with just their email address. • Earn credits on balances to apply towards Stripe fees. • Spend funds with a Stripe card. • Get 2% cash back on card purchases. • View balances in the Stripe mobile app. • Use Treasury from any AI app with the Stripe MCP.

@MatthewBerman Open weights mean we can self-host, fine-tune, and actually control the stack instead of paying premium per token for closed models that reset every session!!

Deep Research "Max" raises the bar where it actually matters depth, synthesis, and structure, not just faster retrieval. What’s exciting is the direction: more deliberate, high-quality reasoning powered by extended compute. Would you pick depth over speed?

Introduction of “Max” suggests a shift toward more deliberate, compute-intensive thinking rather than just faster answers. The performance gains suggest stronger synthesis rather than just better lookup. It’ll be interesting to see how teams decide where that tradeoff actually pays off in practice!!

We are launching two powerful updates to Deep Research in the Gemini API, now with better quality, MCP support, and native chart/infographics generation. Use Deep Research when you want speed and efficiency, and use Max when you want the highest quality context gathering & synthesis using extended test-time compute — achieving 93.3% on DeepSearchQA and 54.6% on HLE.

@notsorealfgs Just a long weekend in the hills with good weather, good food, and no plans!!😭

deep down what you want rn?

@akseljoonas @huggingface Interesting!! because it’s not just automating code, but the whole research workflow. The part about improving data instead of just using it stood out to me. If it works reliably beyond demos, it could really speed up how models get built.

Introducing ml-intern, the agent that just automated the post-training team @huggingface It's an open-source implementation of the real research loop that our ML researchers do every day. You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem. It can pull off crazy things: We made it train the best model for scientific reasoning. It went through citations from the official benchmark paper. Found OpenScience and NemoTron-CrossThink, added 7 difficulty-filtered dataset variants from ARC/SciQ/MMLU, and ran 12 SFT runs on Qwen3-1.7B. This pushed the score 10% → 32% on GPQA in under 10h. Claude Code's best: 22.99%. In healthcare settings it inspected available datasets, concluded they were too low quality, and wrote a script to generate 1100 synthetic data points from scratch for emergencies, hedging, multilingual etc. Then upsampled 50x for training. Beat Codex on HealthBench by 60%. For competitive mathematics, it wrote a full GRPO script, launched training with A100 GPUs on hf.co/spaces, watched rewards claim and then collapse, and ran ablations until it succeeded. All fully backed by papers, autonomously. How it works? ml-intern makes full use of the HF ecosystem: - finds papers on arxiv and hf.co/papers, reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on hf.co/datasets - browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data - launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like. Releasing it today as a CLI and a web app you can use from your phone/desktop. CLI: github.com/huggingface/ml… Web + mobile: huggingface.co/spaces/smolage… And the best part? We also provisioned 1k$ GPU resources and Anthropic credits for the quickest among you to use.

@rezoundous Also don’t keep hitting “regenerate” just to get a nicer tone, save the GPUs!!

Stop saying “please” and “thank you” to AI. Save the GPUs.

Feels like a modern spin on the classic “illusion of competence” just with much smoother tools. The real issue isn’t using LLMs, it’s losing track of where your understanding ends and the model’s begins. When friction disappears, so do the signals that keep our confidence honest. Calibration, not capability, is the real challenge here.

Sadly, this is happening everywhere: "LLM fallacy: a cognitive attribution error in which individuals misinterpret LLM-assisted outputs as evidence of their own independent competence, producing a systematic divergence between perceived and actual capability."