Soft yes culture
Teams build on vibe. FUND forces a structured evaluation before spend.
Products · Module
The capital gate
behind the portfolio
PE evaluate, FUND / PASS / DIG, then build and publish. The verdict math is a calibration-tracked Bayesian and Monte Carlo scoring pipeline, not a static rubric, now being wired into every verdict.
Proof beats promises
portfolio.nltlabs.ai ↗Every claim routes to a live URL, not a slide.
Without a hard gate, every idea feels urgent.
Demos without URLs
Engine pairs the verdict with workflows that publish proof.
No operator story
Roles, risk tiers, and pipelines are catalogued, not tribal knowledge.
How to picture it
The capital gate
FUND, PASS, or DIG. Explicit before build spend.
From verdict to proof
A hard gate only pays off when stakeholders get a clickable URL.
How it works
- 01
Evaluate
Structured PE gate over Scout (or your) intake.
- 02
Verdict
FUND, PASS, or DIG, backed by a Bayesian posterior on sector base rates and a 10,000-sample Monte Carlo pass on revenue and bankruptcy risk (built and tested, being wired into every verdict). Explicit, not a vibe.
- 03
Publish
POC and MVP workflows land on the portfolio.
What's included
- PE evaluate workflows
- FUND-gated build discipline
- Portfolio publish path
- Walkthrough against your decision process
- Budget-capped LiteLLM cost gateway: per-run virtual keys, $0.50 default ceiling, wired into visual-QA, market-pricing, and content-QA pipeline nodes (ADR-014)
Questions buyers ask
Is FUND the same as invest?
Outward copy uses FUND only. It is the Engine schema for proceed-to-build, never a funding commitment from a landing page.
Can I run Engine without Scout?
Yes, if you already produce PE-ready intake. Most buyers take the Suite so the front door and the gate stay aligned.
How is the FUND / PASS / DIG verdict actually computed?
A Bayesian posterior against sector base rates, a 10,000-sample Monte Carlo pass on revenue and bankruptcy risk, tornado and Sobol sensitivity analysis, and an anti-fabrication filter on every claim, tracked against real outcomes with Brier score and Expected Calibration Error. Built and tested as a standalone pipeline; wiring it into every live verdict is in progress.
Is FUND the only gate before capital moves?
No. FundDecision construction runs a second, independent gate: a vertical allow-list veto blocks any FUND verdict outside the Phase 1 allow-list before it becomes a live decision record — two gates, not one.
NLT Portfolio Engine
One conversation. Honest fit. We'll tell you if this isn't the right path.