Founding/Staff AI engineer · building acc4, an RLM over late-interaction scored-token memory · ex-CTO · Products Delivery · RL credit assignment
Max Baluev is a founding and staff AI engineer building acc4, a recursive reasoning system over late-interaction scored-token memory. His skills show an agent how to route a goal through that scored-memory loop, drain a queue of saved deliberation checkpoints by reading each frame's typed hole and retrieved context, and triage open commitments with honest outcome verdicts. Together they focus on structured, memory-backed reasoning and self-accounting for agent work.
Accreted Intelligence
3 skills
Drain a queue of saved AI deliberation checkpoints Works through open and waiting checkpoints left by headless runs, resolving each one.
Route a goal through a scored-memory reasoning loop with acc Get grounded, scored answers from acc's memory loop, with deliberation steps and outcome tracking.
Track and close open commitments with honest outcome verdicts Review pending promises and record real-world outcomes without inflating your self-grades.