Whats In The Works

What's in Progress

forkit - A dependency scanner that finds external libraries and checks if you can just forkit; swap them out for self-hostable, open-source alternatives.
neural-arcade - A progression of neural networks built to play increasingly difficult games, from tic-tac-toe to emulated handheld consoles. Each game gets its own model and its own set of lessons.
advesarial-research - A research workflow that treats every thesis as something to be attacked, not confirmed. An agent gathers evidence from approved sources, extracts and classifies relevant claims, then hands them to opposing personas that argue for and against each point using structured academic debate. The goal is not consensus. It is a report that shows what survives scrutiny.

openbot-fork - A fork of OpenBot ↗ with CopilotKit Intelligence ripped out and local-first vLLM/Ollama support added. Same bot, no phone home.
trolly-autopilot - A simulation that feeds an LLM a queue of trolley problem variations and lets it pull the lever on each one. Then it has to explain why.
reportcard - A benchmark suite that puts LLMs through practical tasks and grades what they actually produce. The models write very confident essays. The scores say otherwise.
loradrone - A compact drone built for tactical comms and rapid deployment. Small enough to pack, fast enough to matter. The prepper's answer to "I need to know what's on the other side of the hill, now."
onebrain - One neural agent, increasingly different games. Start with tic-tac-toe, work toward emulated console games, and find out how far the same general architecture can stretch before it breaks.