← Fellow Signal dashboardSelection Simulator
Picking 50 from 1,000
Activate draws roughly a thousand applications per cohort and accepts about fifty (~5%). This is the core of founder discovery, and the one thing a dataset of accepted fellows can't show you. So here is an illustrative model of the decision: a synthetic field of ~1,000 candidate-like profiles, a few levers, one cohort. Drag the levers and watch the portfolio you'd be building change.
Illustrative model. A synthetic field of ~1,000 candidate-like profiles, drawn from selected-fellow and field-signal priors, to show how a selection strategy shapes a cohort.
Not real applicant data, and it does not estimate applicant quality or selection probability, we don't have the applicant pool. Profiles are built from acceptedfellows and public field signals as a stand-in; sectors are sampled from Activate's vertical-mention counts, not unique primary sectors.
Try a strategy:
1,000 candidate-like profiles, clustered by sector50 selected (5%)
Hover any dot for its signals, what drove its score, and why it's in or out. Drag a lever to see the sector swap here.
Signal weights
Research, funding, and whitespace are field signals (same for everyone in a sector); only depth is the founder's own. So weights pick the sectors, depth picks who within them.
Research momentum60
favor fields whose science is accelerating
Funding momentum45
favor fields where federal money is accelerating
Founder research depth50
favor deeper, more-cited founders (percentile-ranked)
Whitespace / frontier30
favor fields Activate is barely in yet
Experimental levers
Illustrative of the range; a real version would be tuned to Activate's actual process.
Cohort size50
Lean into a sector (strong institutional override)
Diversity capoff
ceiling, max share any one sector can take
Breadth flooroff
floor, reserve up to N per sector (real cohorts span the frontier)
The cohort you just builtWeight robustness: Stable (8% churn, ±10, this field)
A frontier-leaning, moderately concentrated cohort with high founder-depth dependence and narrow breadth; robust to weight nudges.
32%
top sector concentration
54%
top hub concentration
56%
whitespace
92
median depth pct
Sector mix
Computing16
Transportation & Mobility15
Energy Generation & Delivery10
Carbon Management6
Built Environment3
Hub balance
Berkeley27
Boston8
Anywhere8
New York6
Houston1
On the margin
The decision lives at the edge. These are the closest calls under the current strategy, with the primary binding reason each landed in or out (the dominant constraint from the selection pass, not a full counterfactual rerun). Scores shown are the weighted index, not a probability.
Weakest selected (last in)
Transportation & Mobilitymade the cut67
Carbon Managementmade the cut67
Transportation & Mobilitymade the cut67
Carbon Managementmade the cut67
Built Environmentmade the cut68
Strongest excluded (first out)
Energy Generation & Deliverybelow the cut67
Computingbelow the cut67
Built Environmentbelow the cut67
Carbon Managementbelow the cut67
Computingbelow the cut66
There is no objective best 50, only a strategy made explicit.
Lean on funding momentum and you build an energy and climate cohort; lean on research depth and you build a publication-heavy one. And notice: pure scoring leaves whole sectors dark, because real selection isn't pure optimization. Activate's actual cohorts span every sector, so set a breadth floorand watch the frontier fill in. The job isn't optimizing a score, it's choosing which portfolio you want, breadth and all, and being able to say why.
What the numbers can't see
This triages a thousand profiles to a shortlist on field signals and research footprint. It cannot measure what most often decides a fellow: the originality of the idea, the founder's vision and grit, the magnitude of potential impact. Those surface in interviews, references, and diligence, never in a dataset like this. So read the cohort as a triage that earns a candidate a closer look, not a verdict. The mechanism for adding that human judgment, per-candidate conviction with an audit trail, is the next layer in the proposal.
Phase 1. The design, including strategy dials, auditable manual overrides, and a stakeholder-vote layer, is in the proposal. See the live data in the dashboard.