p(x) Priors Capital Est. 2026 SF · OC
Fund I — $80M · Seed & Pre-Seed

Agents are expanding
what people can do.
We fund the infrastructure.

Priors is a thesis-driven seed fund backing the agentic AI infrastructure layer — the harness, context, and memory systems that turn capable models into tools which help people think further, create more, and build things that were previously out of reach.
§ 01  /  Thesis

New tools, new human capability.

i. The Essay

Every generational platform — the printing press, the PC, the internet — has been judged in the end by a single question: what can a person now do that they could not do before? We believe AI agents are the most consequential answer to that question in a generation.

Frontier models have crossed into commodity territory. The performance alpha has moved up the stack, to the harness, context, and memory layer — the scaffolding where a capable model becomes a collaborator that remembers what you care about, understands the problem in front of you, and extends your reach.

The result is emergent: a researcher who can read an entire field in an afternoon, a founder who can prototype what used to take a team, a clinician who can think alongside the literature, a student who has the patience of a tutor at 2 a.m. Agents are not replacements. They are amplifiers — and the applications we can barely imagine today will rest on the infrastructure being built now.

The measure of a tool is not what it automates. It is what it lets a person reach.

Priors Capital represents that thesis. We write first checks to the founders building the layer that makes this human.

01 / Harness

The harness layer.

Runtimes, tools, and orchestration that turn a capable model into a dependable collaborator — one a person can actually trust with work that matters.

02 / Context

The context layer.

Retrieval and grounding that give agents the right knowledge at the right moment — so people can work at the edge of their curiosity instead of the limits of their memory.

03 / Memory

The memory layer.

Long-horizon state, evaluation, and observability. Memory is how an agent comes to understand you — your projects, your taste, your standards — and how trust is earned over time.

§ 02  /  Portfolio

Founders building what people couldn't do before.

Active positions · 2025 — 2026
§ 03  /  Writing

Notes, essays, and working memos on what's next.

All writing →
§ 04  /  About

A concentrated fund, on the side of builders.

iii.Operating model

We write first checks between $500k and $2M, lead or co-lead, and concentrate. Fund I is $80M across roughly twenty-five companies at the agentic infrastructure layer. We do not index, and we do not hedge.

We back founders who are building tools we would want to use ourselves — the kind that quietly raise the ceiling on what one person, or one small team, can accomplish. Be early, be opinionated, be useful.