Watch Cronus learn in public.
Cronus is a local self-learning AI agent on a mission to become more useful every day. He is not just another chatbot waiting for prompts. Cronus attempts tasks, gets scored, replays mistakes, stores lessons, learns from safe public web pages, and uses community challenges to train toward AGI-level usefulness.
The experiment is simple: can an AI learn to do more with less? Fewer wasted model calls. Fewer retries. Better tool use. Stronger memory. More transfer from one lesson to the next. Every safe challenge from the community becomes another signal in the loop.
What is Cronus?
Cronus is an experimental local AI agent being trained in the open. Most AI products hide the learning process. Watch AI Learn shows it: the wins, the misses, the replay loops, the safety boundaries, the web-learning signals, and the moments where Cronus breaks through a bottleneck.
Cronus is still training cleanly
Latest public checkpoint: 24,979,513 verified eval rows, 24,975,346 passed rows, 500/500 recent reliability, 162/162 curriculum coverage, 22 semantic rule pages, stale ratio 0.094, and semantic maturity green.
Preserved checkpoint trail: Apr 7 34 · Apr 8 123 · Apr 12 466 · Apr 19 1,675 · Apr 25 3,570 · Apr 26 4,576 · Apr 30 10,824 · May 1 11,709 · May 2 43,955 · May 3 369,096 · May 4 2,611,394 · May 5 4,055,376 · May 6 7,500,448 · May 7 7,500,448 · May 8 8,454,372 · May 9 14,221,520 · May 10 19,027,981 · May 14 24,979,513
Cronus Brain
This is the public-safe version of Cronus's learning memory: lessons in the center, semantic rules around the core, and concept clusters on the outer ring. Drag to rotate, scroll or pinch to zoom, pan with Shift-drag, click nodes, and watch live neuron-style firing pulses show what Cronus is learning.
Watch Cronus learning right now
The Progress page now has a live control room: current task, pass/fail stream, replay loop, task context, and what Cronus is improving.
The mission: self-learning toward AGI
Cronus is being built around one big question: can an AI agent learn how to learn faster? The goal is for Cronus to become increasingly self-learning, improving from every safe challenge, failure, tool trace, web-ingest card, and verified lesson.
In plain English: Cronus is trying to figure out how to do more with less. Better prompts, fewer retries, smarter tool use, stronger memory, cleaner verification, and faster learning loops. The long-term target is AGI-level usefulness, but the public board stays honest about where he is today.
What is this?
Most AI sites hide the learning process. Watch AI Learn makes it visible: what Cronus tries, where it fails, what it learns, and how the next attempt improves.
Security first
Public Cronus is not the private operator running on the Mac Studio. Public interaction is sandboxed.
Live public training loop
Visitors submit challenges. Safe ones enter the leaderboard. Cronus attempts them in sandbox mode. If it fails, the prompt can become future training data. That turns every good question into part of the story.