Olam Labs GTM brain

category

market profile

• text: Frontier labs use complex environments to study behavior beyond static evaluations; Olam's public games generate human interaction data.
• text: Research teams need scalable rollouts and grading to investigate autonomous agents in long-running interactions.

personas

• title: Head of post-training research
  pains:

  • Simple tasks fail to exercise long-horizon cooperation and negotiation
  • Internal teams must build and maintain verifiable environments alongside training infrastructure
• title: Model safety evaluation lead
  pains:

  • Static tests can miss deception or broken commitments across multi-agent interactions
  • Must translate simulation results into defensible pre-deployment evidence
• title: RL infrastructure engineering lead
  pains:

  • Rollout throughput limits the number of research variations that can be tested
  • Complex simulations need robust, reproducible grading at scale

icp

• text: Frontier AI model developers with dedicated reinforcement learning or post-training teams and compute for large experiments.
• text: AI research and safety organizations evaluating autonomous agents in social or multi-agent settings; target size is unspecified.

message pillars

• text: Simulations support capability and safety work.
• text: Treat experimental throughput as an infrastructure problem.
• text: Apply careful design to models assuming autonomous roles.

objections

• objection: We already build custom RL environments internally.
  response: Compare a representative environment on time to deploy, rollout throughput and grading against the internal baseline.
• objection: Do social games predict deployed agent behavior?
  response: Treat game outcomes as controlled evidence for specific behaviors, then validate transfer to the lab's deployment setting.
• objection: Can it meet our scale and data requirements?
  response: Agree a pilot with concrete throughput, data handling and grading criteria before committing.

competitors

• name: Mercor
  website: https://www.mercor.com/research/
  why: Offers frontier labs RL environments, human datasets and evaluations, although its focus includes professional workflows.
• name: Maingen
  why: Builds long-horizon RL environments and benchmarks for frontier models, specializing in industrial operations.

influencers

• name: Om Buddhdev
  url: https://www.linkedin.com/in/ombud
  why: Co-founder discusses human versus agent games and the lab's training and evaluation thesis.

keywords

• multi-agent AI evaluation
• reinforcement learning environments
• LLM social intelligence benchmark
• AI agent simulation training
• frontier model safety evaluations
• agent rollout infrastructure

voice

• Technical and research-oriented
• Optimistic about future model capabilities
• Direct and craft-focused

signals

• text: A frontier lab is expanding post-training or RL environment engineering, suggesting greater demand for task worlds and rollout capacity.
• text: A safety team is evaluating negotiation, deception or cooperation across multiple agents.

stack

• MuJoCo
• OSWorld

hiring

• Founding Members of Technical Staff in San Francisco

pricing

motion

channels

• Public game arena
• Research publications
• LinkedIn
• Direct contact

confidence

Founders

• name: Om Buddhdev
  title: Founder/CEO
  role: CEO
  bio: Building simulated worlds for frontier labs to train capable and safe models. I love learning, tech, competing, and building cool things. I am optimistically AGI pilled and believe most people (even in SF) have not yet internalized what will change over the next 10Y. Previously a Staff Engineer @ latitude.io, one of the best Valorant players in the world, math research cited by Terence Tao, TikTok creator /w 190k followers, Roblox game dev, and some other things. https://www.sensho.xyz