Vmax

Vmax is developing AI capable of open-ended learning. We build systems that exceed human performance across every domain, optimizing beyond the local maxima of learning from human expertise.

Our approach lets agents define and optimize their own goals. We do not seek to replace human labor with machine labor, but to discover radically new ways of working.

Campaign waitlist

Campaign is our self-play task generation infrastructure. It transforms your proprietary data and evals into new sets of environments, and refines agents on new examples of the tasks they are intended to perform.

Join the waitlist for early access. Approved accounts run campaigns on their own data and evals, and download every task and environment a campaign produces.

Research

Our research builds recursive self-play: co-evolving populations, verifiable environments, and probes that find frontier tasks cheaply. Every result makes Campaign stronger across all domains and becomes the foundation for the open-ended learning AI as it evolves next.

  1. June 10, 2026PROPEL: Breaking the Solver Bottleneck in Task-Generator RLPROPEL replaces solver trials with an activation probe on a frozen generator's internals, doubling frontier-task generation across math, code, and SWE.
  2. May 25, 2026unix-ctf: Procedural Environments for Unix-Competence Reinforcement LearningWe introduce unix-ctf, a procedural generator of capture-the-flag environments for training and evaluating Unix competence in language-model shell agents.
  3. May 20, 2026PopuLoRA: Co-Evolving LLM Populations for Reasoning Self-⁠PlayWe introduce PopuLoRA, a population-based asymmetric self-play framework for reinforcement learning with verifiable rewards (RLVR) post-training of LLMs.