Independent · AI research & venture capital
RST Reserve is an independent AI research and venture capital firm. We back the infrastructure, models and agentic systems shaping the next generation of technology — and when the company doesn’t exist yet, we build it ourselves.
01 — Approach
Most funds separate the people who study a market from the people who write the cheque. We don’t. Research produces the conviction, capital acts on it, and the studio proves it when nothing on the market does.
An in-house group tracking capability curves, evaluation methodology and the real economics of training and inference. Every position begins as a written memo — the argument exists before the wire does.
Concentrated, patient and founder-first. We lead, co-lead and are comfortable being the first institutional cheque on the cap table. Reserves are held for the companies that earn them.
A venture studio for the gaps. When our research identifies a system the market needs and no one is building it, we incubate it — capital, engineering and go-to-market from day zero.
02 — Thesis
Intelligence is becoming an industrial input. The durable value accrues to whoever owns the layers that make it cheap, reliable and governable — not to the thin wrapper on top.
Architectures, training systems, distributed orchestration and the tooling that makes a run reproducible instead of lucky.
Long-horizon autonomy: planning, memory, tool use, sandboxing and the runtime layer that makes agents safe to put in production.
Accelerators, interconnect, scheduling and the physical constraints — power, thermal, supply — that quietly set the ceiling on everything above.
Pipelines, synthetic generation, provenance, vector and hybrid retrieval. The unglamorous layer that decides whether a model is useful.
Measurement you can defend, adversarial testing, model and agent security. Trust is the licence to deploy, and it has to be built.
Vertical systems in regulated, workflow-heavy industries where domain depth and distribution create a moat a general model cannot cross.
03 — Mandate
You should be able to tell in ninety seconds whether we’re a fit. So here it is, without the choreography.
“Diligence should teach the founder something. If our memo tells you nothing you didn’t already know, we haven’t done the work.”
“We would rather be early and wrong in public than late and right in private. Conviction is written down and dated.”
“Independence is the product. No strategic mandate, no platform politics — a single question: is this the system that should exist?”
04 — Process
A fast no is worth more to you than a slow maybe. We commit to both the speed and the reasoning.
You send the form or the deck. A partner reads it — not a screening associate, not a scoring model. You hear back within 48 hours either way.
Two working sessions with our research group. Architecture, evaluations, cost curves, failure modes. Technical, specific and genuinely two-directional.
We write the investment memo and send it to you — including the risks. You see exactly how we think about your company before any decision is made.
Yes or no inside fourteen days of the first call, with the reasoning attached. If it’s a no, you keep the memo.
