ChatGPT, Claude and Gemini all came out of America. Yet America couldn’t produce a free model that stayed in the race.
For close to two years, anyone hunting for the best AI they could download and run on their own hardware ended up with a Chinese model. DeepSeek, Qwen, Kimi, GLM, MiniMax, Xiaomi’s MiMo. They kept taking turns passing each other, while American labs locked their strongest work behind subscriptions.
… Then on October 5, a New York startup, already known in some circles as “the American DeepSeek,” finally released its response. The model is called Beam. — Read More
Recent Updates
Import AI 477: In which I tell the IMF the Singularity has begun
Here’s a speech I gave on Saturday the 17th of October at the IMF meeting in Bangkok. This speech is inspired by the megatrend readers have seen me cover this year of AI beginning to operate at the frontier of human science, sometimes autonomously. While we all may quibble over the details, I think it’s evident we’re in the early stages of the singularity, and I wanted to use my privileged position of giving the Per Jacobsson lecture to try and tell the assembled audience of economists, bankers, and policymakers, that they should expect significant, rapid change in the future. In this speech, I focused primarily on how humanity can share in the benefits of AI if we exercise proper foresight and restraint, and I’m going to be expanding on the stakes of the singularity and the risks that we’ll all face in coming issues.
As I said a few times in the speech – some of which was ad libbed and not captured here – I take my own responsibility here extremely seriously. I find myself both a narrator of what is happening in AI, a principle agent behind what is happening in AI, and a translator that goes between many communities. In all of this, I am seeking to be truthful and to be clear about my own culpability. This is a forever work-in-progress. — Read More
AI Agent Architecture: 7 Layers of Every AI System
𝗢𝗻𝗲 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 𝗗𝗶𝗮𝗴𝗿𝗮𝗺 𝗧𝗵𝗮𝘁 𝗘𝘅𝗽𝗹𝗮𝗶𝗻𝘀 𝗘𝘃𝗲𝗿𝘆 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁
Every AI agent you’ve ever used follows the same pattern.
ChatGPT. Claude. Copilot. Devin. Custom agents built with LangGraph or CrewAI. Even the autonomous multi-agent systems running inside enterprises right now.
Strip away the branding and the frameworks and they all share one architecture loop:
𝗣𝗲𝗿𝗰𝗲𝗶𝘃𝗲 → 𝗥𝗲𝗺𝗲𝗺𝗯𝗲𝗿 → 𝗧𝗵𝗶𝗻𝗸 → 𝗣𝗹𝗮𝗻 → 𝗔𝗰𝘁 → 𝗢𝗯𝘀𝗲𝗿𝘃𝗲 → 𝗟𝗼𝗼𝗽
— Read More
CAPTCHAs in the Agentic Era: Solvers That Learn from Every Encounter
Vision-language models (VLMs) can solve visual CAPTCHAs without task-specific training, but the agents built on them approach every challenge from scratch. For such an agent, the hundredth instance of a familiar puzzle costs as much time and compute as the first. Specialized detectors invert the trade-off, answering in milliseconds but only for categories they were trained on. Neither improves with exposure. We study what changes when a solver improves with use. Our system pairs a fine-tuned YOLOv8 detector with an open-weight VLM behind a confidence-based router, and runs entirely from screenshots and operating-system input events, with no browser automation or DOM access. It reaches 85.4% overall and 84.2% macro accuracy across 16 classes, exceeding either component alone. Every answer VLM produces also serves as a training label, so the detector absorbs categories it was never trained for, typically after one or two encounters and without human annotation. The same loop also repairs it. A CAPTCHA operator can perturb images against the publicly released detector and drive its accuracy to 0%, but the perturbations leave VLM untouched, and its labels let the detector recover. Under a year-long simulated arms race in which the CAPTCHA operator re-crafts its perturbations each month, the solver recovers every round, and a cheap ~70%-accurate open-weight teacher hardens it as effectively as a perfect oracle. Visual CAPTCHA defenses that assume a failing bot stays failing therefore understate how quickly an adaptive solver returns. — Read More
MCP Is Dead
Last week, Perplexity announced that they were dropping MCP and going back to plain old APIs and a command-line tool.
I love MCP. I used it frequently to connect my local systems to my agent frameworks. It makes LangGraph and Strand a joy. However, MCP is a hack, a naive protocol that should never leave your local network. It can’t be fixed. You cannot retrofit into good architecture.
Simplicity is its selling point, an API for people who don’t want to worry about APIs; a sort of USB for agents to universally connect to other applications. It is a great solution for exposing your local services to your local agents. It is fast to wire up and resilient to change. If you are trying to do something quickly and not for production, it’s perfect.
But the internet is not, and cannot be built on USB connections. — Read More
AI Agents (07): Agentic Harness (Frameworks)
If this is the first story you have come across in this series, don’t miss out on the entire AI Agents series — explaining in great detail the core of AI agents, their communication protocols, famous frameworks, agent scaling, monitoring and observability — and a lot more. We have a great multi-part Generative AI series as well; don’t forget to checkout that too!
Building scalable, secure, performant and dynamic AI agents from scratch is one of the most tedious modern software development tasks. Luckily, there are several frameworks out there which help us with automating the whole process by abstracting away the details.
— Agentic harness is the software-level infrastructure built around an LLM to equip it with the ability of taking action in the real world, while monitoring and logging everything it makes the agents do. —
— Read More
Gemini 4 Argon available on Google Cloud, built to help solve your business’ hardest problems
Last week we announced that Gemini Enterprise has evolved to make it even simpler to get work done, and introduced the Gemini agent, your new universal agent for work. It’s always on and ready to answer your questions. It will write code, complete longer and more complex tasks, or perform whatever you need to get the job done, all from a single prompt box. And model intelligence underpins it all.
To enable a new frontier of work, we are making our new frontier model, Gemini 4 Argon, available on Google Cloud. — Read More
Google Says It Just Cracked One of Quantum Computing’s Biggest Bottlenecks
Google’s Willow chip cut errors 40x on the scarce “magic states” quantum computers burn through, a step toward smaller, useful machines—and toward ones that could threaten today’s encryption.
Google says it has turned one of quantum computing’s most promising paper tricks into something that actually works on a chip.
In a paper published in Science on Thursday, the company’s Quantum AI team showed that a technique called magic state cultivation cut errors by about 40 times on its Willow processor. — Read More
Introducing Muse Spark Open,The Most Capable US Open‑Weight Model to Date
Muse Spark Open is available today and brings leading agentic capabilities to your desktop for the first time on Apple M5 Ultra Mac Studios.
Today, we’re releasing the weights for Muse Spark Open, the most capable US open-weight model to date. Muse Spark Open is a natively multimodal model with 816B total parameters (42B active) and a 1M-token context window. Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.
We believe the future is for everyone, and that includes the freedom to run advanced models wherever, including at home. Muse Spark Open is the first model in its class to run on your desktop using two Apple M5 Ultra Mac Studios with the latest MLX framework. — Read More
Software Abundance & The Trojan Horse That Comes With It
One fascinating thing about AI is that things seem not to change much until they do. Suddenly, AI can do something extremely effectively which it could not do at all just the day before.
And these past weeks are one of those times, and the “emergent capability” this time around is called “reverse engineering.” At least when it comes to software, “everything” is replicable now. — Read More
