Capabilities that have crossed the line. Oldest flip first — a chronology of the frontier.
10-Ks, earnings calls, MD&A sections. Buy-side analysts now spend more time prompting and verifying than reading. --- Current AI systems can process and analy...
Long a hard problem; mostly solved by 2023's contextual LLMs. Edge cases remain, but everyday detection is operational. --- Currently, AI systems can identify...
Banking ML models have been doing this for a decade; modern transformers improved tail-case detection again in 2024. --- AI can detect fraudulent credit-card...
Models that combine social signal, trailer engagement and historical patterns now beat box-office veterans on aggregate predictions. --- AI systems have made...
Diagnostic-companion models in 2024 found cases of rare conditions missed by clinicians in both training data and live trials. --- AI can diagnose certain rar...
State by state, including the road-sign questions and edge-case rules. Trivial for any modern frontier LLM. --- AI systems have made significant progress in n...
GitHub Copilot Workspace, Sourcegraph Cody, others — most modern engineering teams use AI-generated review comments as a first pass. --- AI can generate code...
Tools like FunSearch and AI-co-scientist released in 2024 surfaced novel hypotheses in materials science and biology that humans then verified in lab. --- Cur...
Specialised math models plus chain-of-thought tooling closed the gap to top human contestants in 2024. --- AI systems have demonstrated the ability to perform...
AlphaFold-Multimer and successors took this benchmark in 2024. --- Current AI systems have made significant progress in predicting protein-protein interaction...
The 'aha' moment problems that used to stump LLMs are now mostly solvable with good chain-of-thought tooling. --- AI systems have made significant progress in...
DeepMind's AlphaProof + AlphaGeometry 2 reached silver-medal level at IMO 2024 and approached gold by 2025 in geometry and number theory. --- AI systems have...
Precision-medicine assistants used at major academic medical centers. Final decisions remain with clinicians; suggestions are good enough that ignoring them cos...
Agentic systems run multi-step web tasks, file ops, calls to other agents. Not yet reliable enough for all jobs, but solidly working for many. --- Current res...
Add a statement to the atlas. We review weekly.
Site
stuffaicantdo.com © 2026 · made in NL
Pick a side, give it a one-line title, optional explanation, and a category. We review submissions weekly.