Things are moving fast now. Two months ago Anthropic launched a new feature called “Cowork”, where Claude works autonomously within a sandbox environment on your computer with access to your files and various connectors. You choose the files Claude can work with, but you need to be careful – if prompted the wrong way there is a risk Claude removes or invalidates important files. This entire feature, Claude Cowork, was built in record time – only 10 days, and 100% of the code for it was built by Claude itself. Having agents run freely within a controlled environment is what…

Read more…

Last year in Tech Insights 2025 Week 30 I wrote: “In my last two seminars I got the question what main trend I see for AI in 2026, and to me it’s easy: computer control. It’s the natural next step, and once we get models specifically trained for this then they will very soon be much more efficient doing things at the computer than any human”. Last week OpenAI launched GPT-5.4 and guess what, it is specifically trained for computer use: “GPT‑5.4 is our first general-purpose model with native computer-use capabilities and marks a major step forward for developers and agents alike.…

Read more…

The main AI news of last week were the public announcements from Anthropic, OpenAI and the Department of War, where most people on social media seemed to summarize this as: Artists like Katy Perry even publicly posted on X that she has left ChatGPT and signed up for an Anthropic Pro plan, and it has actually been difficult to navigate through social media in recent days due to all these “Vote with your wallet” posts. So, I spent a significant amount of time reading up on everything I could on this matter, and in this case it is clear that…

Read more…

A year ago our top benchmark for measuring “smartness” in AI models, ARC-AGI, was becoming saturated. OpenAI o3 scored over 87% on the benchmark in a high-compute configuration, and it became clear we needed a new benchmark. A new benchmark was designed, and ARC-AGI-2 was launched in March 2025. The test is based on visual problem solving, where humans average 60% on the test. The best models in April 2025 scored less than 4% on ARC-AGI-2. In November 2025 Google launched Gemini 3 Pro that scored an impressive 31% on the ARC-AGI-2 test, and a month later OpenAI launched GPT-5.2-extra-high…

Read more…

The past two weeks you have probably read about the very popular AI agent “Clawdbot” that was later renamed to “OpenClaw”. OpenClaw is a viral, open-source AI agent platform that runs locally on your computer, allowing it to control your machine, manage files, and interact via chat apps like WhatsApp and Slack. Looking at the number of user-assigned GitHub stars, the repository for openclaw is on its way to become one the most popular repos on GitHub ever, in absolute record time. Just look at the graph below (updated on February 14). The red line on the right is no…

Read more…

If you ask the AI coding tool Claude Code to commit source code directly to GitHub, it will by default add itself as an author with the commit message: “Co-Authored-By: Claude Opus”. Most people I work with do not trust Claude to commit anything at all by itself, and those who do trust it typically disable this commit message. Now however, in February 2026, over 4% of all public GitHub commits contain this message. And the growth has been incredible. Greg Brockman, co-founder of OpenAI, said last week that “Since December, there’s been a step function improvement in what tools…

Read more…

“My name is Steve Chen. I was the CTO and one of the co-founders of YouTube. I love what you’ve built with Notebook Navigator 2.2! I wanted to send out a note of gratitude.” Steve contacted me last Tuesday, and is just one of hundreds of users that have reached out to me saying that Notebook Navigator + Obsidian fundamentally changed the way they are writing and organizing notes. This is how you get “paid” as an open source developer, and it’s one of the best feelings in the world to have created something that other people love to use.…

Read more…

Can you tell if a video is real or AI generated? Runway, the company behind the AI video model Gen-4.5 just did a test where they let 1,043 participants watch two five-second clips, where the first frame of the real video was used to create the animated AI-video. In this study, less than 10% could tell the difference between real and AI-generated video. If you have a few minutes to spare, you can do the test yourself, it would be interesting to hear if you succeeded. A year ago most discussions around AI generated video was around the problem with…

Read more…

Last week Anthropic launched a new feature called Cowork: “Give Claude access to a folder, set a task, and let it work”. You can ask Claude to go through your receipts and give you a summary, sort your files, find a specific text in a note you have somewhere on your desktop, create a presentation based on your draft documents, and so on. Instead of having to upload documents and files through a chat window, Claude can now work directly on your computer, with your files. This specific feature, Claude Cowork, was fully developed in just 1.5 weeks. And what’s…

Read more…

In an interview with CNBC, AMD CEO Lisa Su said that she expects us to go from “a billion active users in AI today to over 5 billion active users over the next five years”, requiring compute to increase 100 times to meet demand. This is what all major AI companies are working toward; a world where every connected human is using AI on a daily basis for almost every single task. It’s easy to become addicted to using AI. It knows everything and can do everything. You ask it and it performs. The big question is the long term…

Read more…