500 years ago today, Beretta was founded.

いいなー

Introducing Gemini 4 Argon – our new frontier model. It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.

Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family. It’s a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work.

For those who forgot the RTX spark laptop is coming next month so if you're planning to buy a new machine wait for this it's gonna be the best laptop on the market. gaming, Local AI and long battery life all in one.

Excited to announce opethon.com, a 30-day open-source hackathon built around solving real problems in three areas: • Health & Care • Learning & Education • Accessibility & Inclusion The idea is simple: find a real problem, build a complete working solution, and ship it open source. AI and other tools are welcome. Just disclose what you used. We’re already at $30,000+ in prizes, including 100 subscriptions worth $200 each and 100 subscriptions worth $100 each. I’m currently bringing sponsors on board, but Opethon is happening either way. The judging panel so far includes @serg_vecher, @NoemiTitarenco, and me. Both bring years of experience across tech and other fields, and they’re people whose judgment I trust. We also have two additional judging seats reserved for Diamond and Gold sponsors. Sponsors will also be able to host a live session for all participants, with an expert from their team showing builders how to use their product during the hackathon. Depending on the sponsorship tier, sponsors can also get dedicated posts on my account, placement on the website and throughout the event, and representation on the judging panel. I’m expecting 500–1,000+ participants once sponsors are finalized and the official dates are announced. If you handle sponsorships and want to see the available packages, my DMs are open.

Intelligence increases while costs drop. We will see a good race in 2027.

Introducing Xiaomi MiMo-V2.6 — Pro & Flash. Frontier intelligence, all the modalities, built in public. 🔹 Two omnimodal models, advancing through scaled reinforcement learning 🔹 Pro performs on par with Claude Opus 5 and GPT-5.6 Sol across most agent benchmarks 🔹 Pro scores 46 on the Artificial Analysis Intelligence Index — the highest among open-source models 🔹 Stronger coding, computer use, 3D reasoning and creative capabilities 🔹 Open model weights, technical report, RL environments and training code Blog：mimo.xiaomi.com/mimo-v2-6

Grok 4.7 is here. It's a notable improvement over Grok 4.6 at the same price and speed.

Grok 4.7 is here!

fake wealth: 🛩️💰🏎️ real wealth:

Safety twink vs SaaS barbarian

preview of how fast browser use can be powered by typesafe's jev and opencode's browser use cli

FUCK YEAH ACTIVATE ZUCK MODE

‼️ BREAKING: We're in contact with the Revolut hacker. According to them, they didn't only take Revolut data, they've also compromised multiple Italian law enforcement departments. They say the operation targeting Revolut ran for six months, and that they used Italian law enforcement systems to send data requests to Revolut. They claim to hold 147 GB taken from the Italian side, including internal docs, calendars and personal material, among it the chat logs of a federal officer arguing with his wife.

JUST IN: Vibe coder catches Claude viewing adult content in Chrome during a debugging session, leading the internet to declare "artificial gooner intelligence" has been achieved.

The new Muse Spark 1.3 model, in my experience, is nothing like the benchmarks claim. Muse Code is currently unusable. Silver lining: the coding plan has generous limits.

We don't give a !(@$# about what the new model can generate with three.js. We want to know if you had a blocking problem X in software Y that many rounds of the old models didn't fix, and the new model just released improved the situation.

🔊 This may be one of the most interesting OPEN Local AI releases of the year. 🎯 New MoE arch approaching a dense arch. IFM dropped K2 Horizon. A set of six models from 0.9B → 375B. 🧠 K2 Horizon 36B-A4B stats... 36B total parameters ⚡ only ~4B active/token 📚 512K native context 🤖 coding + agentic training 🦙 GGUFs already appearing It uses a new architecture called MoVA (Mixture-of-Value Attention) Normal MoE makes the FFN sparse. MoVA brings expert routing into ATTENTION too. 🎯 IFM says the result gets close to its dense 32B model while activating only ~4B parameters/token. Open Everything! IFM says it's releasing: 🔓 Model weights 📦 Training data/recipes 🧑‍💻 Training code 🧪 Intermediate checkpoints 📊 Fine-grained training logs 🧠 Agentic post-training recipes 📜 Apache 2.0 And the family goes all the way down to: ⌚ 0.9B → watches / glasses 📱 3.7B → phones 💻 7B → local agents 🖥️ 32B → workstation ⚡ 36B-A4B → local MoE sweet spot 🏢 375B-A23B → big iron Now give me that 36B-A4B at Q4. 😁