We need to talk about this...
Matthew Berman showed 8 AI tools in this video on . 5 are verified available in the EU. Each row below jumps to the exact second. Part of week 2026-37.
Tools shown in this video, with timestamps
Creator says Opus 4.5 changed things overnight at the end of 2025, moving models from autocomplete to writing entire applications autonomously. On the METR chart it works nearly 5 hours autonomously.
Announced Feb 5, 2026 as OpenAI's first model instrumental in creating itself; the Codex team used early versions to debug training, manage deployment and diagnose evaluations.
Benchmark from OpenAI (April 2025) testing a model's ability to read research papers and write code to replicate their results.
Google's Gemini-powered coding agent that used AI to improve its own systems, including efficiency gains in Google's infrastructure architecture.
Model OpenAI released the previous week that the creator says blew the world away. OpenAI said an internal model, still training, is significantly more capable than it and was used to solve the Navier-Stokes problem.
Chart tracking how long models can work autonomously, rising from 9 seconds for GPT-3 to about 12 hours for Claude Opus 4.6 and 16 hours for Claude Mythos. The creator calls it a near-vertical acceleration.
Shown on the METR chart as able to work autonomously for about 12 hours.
Shown on the METR chart as able to work autonomously for about 16 hours.
