AI Agents

There’s a dirty secret in the AI agent world: most teams running Claude Opus are burning money for bragging rights. Don’t get me wrong — Opus 4.6 is a beast. It tops SWE-bench at 80.9%, handles 200K context windows without breaking a sweat, and orchestrates multi-tool workflows like a conductor with perfect pitch. But at $15 per million tokens (blended), it’s the filet mignon of language models. And most of us are building tacos. ...

The smartest chatbot in the world is still just a chatbot. Anthropic knows this — and it’s spending aggressively to change it. On Wednesday, Anthropic acquired Vercept, a Seattle-based startup that built AI agents capable of remotely operating a full MacBook. It’s the company’s second major acqui-hire in three months, following its December purchase of Bun (the engine behind Claude Code). The message is clear: Anthropic isn’t building a better conversation partner. It’s building a digital employee. ...

Not a Goldman Sachs research paper. Not a Fortune 500 earnings miss. A speculative thought experiment on a blogging platform — written by a firm most people had never heard of — just erased billions in market cap. DoorDash dropped 7%. American Express fell 7.2%, its worst day since April. IBM cratered 13% — worst since 2000. MongoDB and AppLovin slid 8%. The Dow shed 823 points. All because people read a scary story on the internet and two famous contrarians hit “share.” ...

Every week, a new blog post asks: MCP or RAG? RAG or Agents? Agents or MCP? The question is wrong. These aren’t competing technologies. They’re layers — and the teams shipping the most capable AI systems in 2026 are stacking all three. Here’s what each layer actually does and why you probably need more than one. MCP: One Plug to Rule Them All LLMs are brilliant and trapped. They can reason about your calendar but can’t read it. They can plan a database migration but can’t execute it. They need hands — and before MCP, giving them hands was a mess. ...

Alibaba just dropped a model with 397 billion parameters that only uses 17 billion of them. And it might be the most important AI release of 2026 so far. Qwen 3.5 landed on the eve of Chinese New Year — 60% cheaper to run than its predecessor, 8x faster at decoding, and supposedly beating GPT-5.2, Claude Opus 4.5, and Gemini 3 Pro on multiple benchmarks. Bold claims. But even if you discount the scoreboard, the engineering underneath tells a story worth paying attention to. ...

A solo developer vibe-codes an AI assistant. It goes viral. Anthropic sends a cease-and-desist over the name. He rebrands — twice. Within a month, OpenAI hires him to “drive the next generation of personal agents.” Peter Steinberger’s journey from hobbyist tinkerer to OpenAI employee is the most 2026 story imaginable. But beneath the speed-run narrative, his move tells us something important about where the entire AI industry is heading. From Side Project to Industry Shaker OpenClaw — originally called Clawdbot, then Moltbot — is an AI assistant that actually does things. Not just answers questions. It manages calendars, books flights, sends messages, and even interacts with other AI assistants on what became a bizarre and delightful AI social network. ...

A solo developer vibe-codes an AI assistant. It goes viral. Anthropic threatens to sue over the name. He rebrands twice. Within a month, he’s hired by OpenAI to build the future of personal AI agents. That’s not a hypothetical. That’s Peter Steinberger’s actual February. From Side Project to OpenAI in 30 Days Steinberger was already a known name — he founded PSPDFKit and ran it for 13 years. But OpenClaw was different. It started as a tinkering project called Clawdbot: an AI assistant that could actually do things. Not just chat. Manage calendars, book flights, send messages, even talk to other AI assistants on what became a genuinely weird and wonderful AI social network. ...