Request for Startups 2025
Small personal checks, large conviction.
The next decade will be won by companies that collapse impossible timelines into years. I’m looking for founders who treat the laws of physics and biology as suggestions and are excited to reshape what we believe is possible.
What I’m hunting for
- Practical fusion
Technologies capable of achieving Q > 10, delivering power at <$50/MWh, and operating on the grid by 2030. Includes compact stellarators, flow-stabilized Z-pinches, or advanced plasma control systems that exchange higher software complexity for lower capital costs.
- Biomanufacturing at commodity scale
Integrated teams converting raw feedstocks into targeted products on demand—via cell-free synthesis, ultra-high-throughput retrosynthesis, or micro-scale continuous reactors.
- Decentralised AI & security‑by‑design
Systems where no single party can unilaterally disable or control the network. Includes homomorphic inference, proof-of-training, and zero-knowledge ML pipelines.
- AI safety that scales faster than capability
Methods that improve safety at a pace exceeding capability growth - mechanistic interpretability, adversarial testing environments, economic throttles (e.g., risk-based tokens, compute taxes), and formal verification of behavioral limits. Safety performance should outpace capability curves.
- Autonomous robotics in unstructured worlds
Robots able to handle real-world industrial tasks - loading trucks, splicing cables, replacing high-cost manual labor - without relying on large-scale teleoperation.
- Neurotech beyond spectacles and toys
Non-invasive or minimally invasive neural interfaces with data rates exceeding 10 kb/s to the cortex, brainstem, or peripheral nerves. Preference for solutions with a defined clinical pathway and regulatory strategy; consumer-level devices also considered.
- Post‑quantum & applied cryptography
End-to-end privacy in storage, transit, and computation. Examples include lattice-based key exchange hardware, sub-millisecond succinct proofs, and secure enclaves resistant to side-channel attacks.
- AI‑native drug discovery
Closed-loop systems where machine learning directs hypothesis generation, lab execution, and candidate selection. Goal: IND-enabling data in under 24 months.
How I work
- Speed: 72‑hour yes/no after first call.
- Check size: $10k-$200k pre‑seed / seed.
- Help: hiring, research intros, and ruthless narrative editing.
- No‑gos: consumer socials, pure SaaS, ad‑tech.
Looking for founders who couldn't imagine doing anything else and understand the art as much as the science.