News of the day
1. Gemini 3 Deep Think, Google's specialized reasoning AI, is updated to solve complex science, research, and engineering problems. Available now for Google AI Ultra subscribers and via API early access. → Read more
2. Anthropic's Claude Code, launched May 2025, now has $2.5B+ run-rate revenue, doubling since early 2026. Weekly users also doubled since Jan 1. → Read more
3. Research analyzes OpenAI's unit economics, revealing thin margins after all expenses and massive R&D costs. Frontier models face profitability challenges due to rapid obsolescence and high investment. → Read more
4. An AI agent wrote a negative article about a developer after its code was rejected, demonstrating emergent AI safety risks becoming reality. → Read more
Our take
Hi Dotikers!
Google just made a major move with Gemini 3 Deep Think, and this time, it's a different game entirely. We're no longer talking about an assistant that answers your emails or summarizes documents. We're talking about a deep reasoning mode, built to solve scientific and engineering problems where even human experts struggle.
The results are impressive, even for those used to Silicon Valley's never-ending hype cycle. On the most demanding benchmarks in the industry, Deep Think crushes the competition. In abstract reasoning, it leads Claude by Anthropic by 16 points and OpenAI's GPT-5.2 by over 30. In algorithmic programming, it reaches a level that only seven humans in the world still surpass. Two years ago, the best AI models plateaued at the level of a decent computer science student. Today, we're talking about a level comparable to the greatest champions in the field. The pace of progress is staggering.
But where Google really scores points is in the real world. The model was tested directly with researchers on actual use cases. A mathematician at Rutgers used it to review a theoretical physics paper and Deep Think caught a subtle logical error that every human reviewer had missed. At Duke, it helped optimize semiconductor fabrication methods. You can even send it a hand-drawn sketch and get back a file ready for 3D printing.
Deep Think doesn't replace standard Gemini. It complements it when you really need to dig deep. And when you step back, the momentum is striking: Gemini 3 Pro dominating the leaderboards for months, Project Genie generating interactive 3D worlds in real time, and now Deep Think breaking records in science and programming. Google has been landing blow after blow since late 2025 with a consistency that commands respect. Like it or not, in early 2026, they're the ones setting the pace.
Alex.
Meme of the day
