Alexis Rondeau

Featured

"We don't write the rules. We just discover them. (With AI.)" — discovering scheduling heuristics with an LLM autoresearch loop A job-shop scheduling problem dressed up as a lunch rush: one grill, one shared range, plates running late, and sequencing as the only lever. Instead of hand-writing the rule, an agent proposes candidates, runs them across simulated services, and keeps what beats the standing champion — trained on six scenarios, held out against three more to check it found a principle, not a memory. Based on Nima H. Siboni's "The Heuristic Scientist: Open-Ended Algorithm Discovery with LLMs" workshop.
"Which city in Paris are you staying in?" - A post-training study on small-model specialization with an autoresearch loop Across 77 experiments — every lever from prompting and retrieval to fine-tuning and 2025's rubric-grounded RL — an on-device 3B model landed within about one rubric point of a frontier cloud model on every quality dimension: close, but consistently a notch below. None of the post-training playbook closed the last notch, and the controlled tests show why — the gap is a broad capability limit, not a single missing skill.

Unblah A macOS app that tracks your talk-time in meetings, so you can speak up — or zip it.
JamGPT A keyboard shortcut to summon ChatGPT instantly, from anywhere on your Mac.
Streamline A distraction-free, stream-of-consciousness writer that feeds straight into Obsidian.
Pudding A macOS app that records every keystroke to prove your writing was made by a human, not AI.
Algorand Ballet An interactive tool for visualizing wallet activity and transaction networks on the Algorand blockchain.
Pudding, live The Pudding editor running right in your browser. Try it.