Mirage Nexus Consulting
Physical AI, examined.
Most physical AI claims don't survive a seed change. We find out which ones do.
Independent technical review for investors, founders, and research labs. No decks. No demo-making. Written verdicts, with the code to back them.
questions we answer
- Does this demo survive a seed change, a data shuffle, a fresh environment build?
- Which of these benchmark wins are architecture, and which are tuning?
- Is this team's rig reproducible end-to-end, or is it held together by one person?
- What would falsify the headline result? And how cheap is that experiment?
engagements
- Deal review
- One to two weeks. Reproduce the central claim, read the team's code, deliver a written verdict on technical risk.
- Deep audit
- Two to four weeks. Stress-test a body of results. Reproduce or falsify specific claims. Written report with code.
- Advisor
- Retainer. Monthly technical take on a team's trajectory. Adversarial review before papers ship.
Not taking: pitch decks, demo engineering, generalist AI strategy, content partnerships.