🤖 An AI that never loses money??
·
📖4 min read
🧠 Anthropic's Sonnet 5 Is Smart but Slightly Feral
- Anthropic dropped Sonnet 5, bringing Opus-level agent skills to the cheaper tier - it runs a web browser, does long tasks, and codes like a beast. But its cybersecurity scores dropped because they, uh, didn't train it for that.
- Why it matters: Powerful AI agents for less money is great!! Just maybe don't hand it the keys to your firewall. (Fans still want the bigger Fable model tho.)
💰 AWS Will Literally Move Into Your Office for $1B
- AWS is spending $1 BILLION to insert its own engineers directly inside customer teams, helping companies build custom AI agents in days, not months.
- Why it matters: Big companies are notoriously bad at actually using AI, so Amazon is basically doing your homework for you. It's giving "hostage negotiation but for cloud contracts."
🚀 Etched Wants to Fight Nvidia and Brought $800M
- AI chip startup Etched emerged from stealth with $800 million in funding and a wild $1 billion in customer contracts - plus a working chip that runs AI models much faster and cheaper.
- Why it matters: Nvidia desperately needs a rival, and Etched is actually shipping silicon at scale. David vs. Goliath, except David has a fab.
🎨 Google Drops Nano Banana 2 (Yes, That's the Name)
- Google launched Nano Banana 2 Lite for cheap, super-fast images and Gemini Omni Flash for top-tier video editing. You can now link the two - make an image, turn it into a video, all in one step.
- Why it matters: Cheaper, faster creative AI is huge for devs building apps at scale. Also Nano Banana 2 sounds like a Mario Kart item and I'm here for it.
⚡ OpenAI Found a "Compute Multiplier" and a Keyboard
- OpenAI found a secret "compute multiplier" that halves their running costs - perfect timing as they launch their new Jalapeño chip. They also teased a hardware keyboard for Codex, dropping July 15th with Work Louder.
- Why it matters: Cheaper compute means better prices for users. Also OpenAI is naming chips after peppers now, so we're fully in spicy-hardware era.
🔬 Anthropic Also Casually Enters the Science Lab
- Anthropic launched Claude Science, an AI workspace that helps scientists review papers and crunch data in one place. They're also starting their own drug discovery program to cure neglected diseases.
- Why it matters: AI is moving beyond "write my emails" to actual scientific breakthroughs. Two major launches in one day?? Bestie, get some sleep.
🐦 X Opens the Gates to AI Agents
- X (formerly Twitter) launched a hosted MCP server, letting AI agents like Claude, Grok, and Cursor easily read posts and take actions on your account.
- Why it matters: It makes it stupidly easy for devs to plug AI agents into real-time social media data. What could possibly go wrong on X, of all places??
📈 Ex-DeepMind Nerds "Solved" the Stock Market
- Three former DeepMind researchers built an AI that supposedly never loses money trading stocks and crypto. Their startup, EquiLibre Technologies, is now valued at $500 million.
- Why it matters: They took the AI that beat humans at poker and pointed it at Wall Street. "Never loses money" is a phrase that has ended many careers, but okay 👀
📱 AI Agents Are Now Living in Your Pocket
- A startup called Acti launched an AI agent keyboard that does tasks right inside your apps. Meanwhile, the viral open-source agent OpenClaw just landed on iOS and Android.
- Why it matters: No more app-switching to use AI - smart agents live right in your pocket now. Truly no escape from the bots. We love to see it (?).
🔒 Proton's Privacy Chatbot Levels Up
- Privacy-focused Proton upgraded its Lumo AI chatbot to version 2.0 - it now handles images, remembers past chats, and runs 76% faster, all while keeping your data locked down.
- Why it matters: Turns out you don't have to sell your soul (or data) for good AI. Top-tier AI with zero data snooping - wild concept, right??