Most programs that support early-stage ventures operate on a combination of structured curriculum and unstructured judgment. The curriculum is the same every cohort. The judgment belongs to the mentor in the room, and it does not transfer. When that mentor rotates out, the pattern knowledge they accumulated over three years of cohorts leaves with them.
I built Nascens to address this. Not because I think intuition is the enemy — experienced mentors have real and valuable pattern recognition — but because that intelligence has no structural home. It exists in the heads of the people who run programs, and nowhere else. The programs that scale, that compound, that get measurably better year over year, are the ones that find a way to make that intelligence systematic.
The six signal dimensions
The foundational design decision in Nascens is the signal vector: six dimensions that together describe the formation quality of a venture at a point in time. Every intake event — a pitch session, a milestone review, a mentor observation — is normalized into this six-dimension structure before any downstream analysis runs.
The six dimensions are: Problem (specificity, evidence of customer pain, framing of the gap), Solution (differentiation, feasibility, alignment to the problem), Market (addressable scope, segmentation clarity, demand evidence), Team (domain credibility, role coverage, founder-problem fit), Evidence (the validation artifacts — surveys, pilots, letters of intent, early revenue), and Assumptions (the explicit dependencies not yet validated).
That last dimension — Assumptions — is worth dwelling on. Most signal frameworks treat unvalidated assumptions as a deficiency to be penalized. We treat them as a first-class signal. A team that cannot name its critical assumptions is not in better shape than a team with three explicit open assumptions. Explicit assumptions are the raw material for targeted mentor intervention. Hidden assumptions are where ventures quietly fail. Surfacing them early is the point.
Why the episodic model matters
Ventures at the formation stage do not generate continuous telemetry. They do not have dashboards. They have episodes: a pitch, a customer conversation, a co-founder decision, a pivot. Between those episodes, their state persists but does not change. A model that tries to track formation quality continuously is imposing a data-collection burden that founders at this stage cannot and should not bear.
The Nascens architecture is explicitly episodic. An event occurs — a session happens, a milestone is reached or missed, a pivot is declared — and that event triggers the intelligence pipeline. The normalized signal vector from that event is compared against the venture's prior state and against the population. Guidance is generated. State is updated. Then the system waits until the next event.
This design choice has a practical consequence that matters enormously in program delivery: the intelligence is available at the moment it is needed, without requiring administrative overhead from founders between sessions. The mentor preparing for next week's check-in sees the current state, the outstanding assumptions, and the trajectory estimate — all from the last event, however long ago that was.
The compound population thesis
The third design principle — and the one I am most honest about in terms of time horizon — is that the database is the product.
Year 1 guidance from Nascens is LLM-derived. It is structured, consistent, and substantially more systematic than unaided judgment, but it is not grounded in population data. The LLM normalizes free text into the signal vector using its general knowledge of venture formation patterns — not empirical data from the database. I label this explicitly as novel guidance, meaning: derived from this venture's text alone.
The database begins to compound after the first hundred venture interactions. As signal records accumulate — across cohorts, across programs, across the shared population — the guidance tier transitions. A venture whose assumption profile matches 40 prior population analogs gets guidance that draws on observed outcomes from those analogs. A venture in a stalling state that matches 80 prior stalling ventures gets a trajectory estimate grounded in what actually resolved those situations, and what did not.
This is the compounding thesis: the programs that join early contribute to the population and eventually draw from a much richer one. The intelligence that is narrow in Year 1 becomes precise in Year 3.
Where we are
Nascens is in early-access engagement with anchor institutions. I am not in general availability and I am not running a broad beta. I am looking for a small number of programs — university entrepreneurship centers, structured accelerators, organized mentor networks — that want to be design partners. Programs that will shape the signal model, contribute the foundational population data, and build the institutional memory infrastructure the product depends on.
If that sounds like your program, I would like to talk. Reach me at richard@accelerate-ip.co or +1 831-431-7147. The conversation is worth having regardless of whether Nascens turns out to be the right fit.