The instrumentthat tells you no.
For product teams & founders: drop a pricing change, a feature, or a story into a city of 1,000 opinionated buyers — and see who buys in, who pushes back, who quietly leaves. Before you ship it.
HOW IT WORKS
Three moves to a verdict.
Choose a market
Start from a template — B2B SaaS, DTC, Fintech. Holon assembles a directed buying network of ~1,000 agents in milliseconds.
Inject the change
A price move, a feature, a story-led rebrand. Pick where it starts, then watch it propagate.
Read the spread
Adoption, resistance, and silent churn — reported as p10·p50·p90 ranges, never a single soothing number.
METHODOLOGY
A thousand minds. One verdict.
Holon runs a vectorized Bass-SIR-D model — adoption, churn, negative contagion — over a directed weighted graph of 600–10,000 agents. Zero LLM calls per tick. A 90-day run on 1,000 agents finishes in ~21 ms. Replayable from a seed. Backtests public.
WHY HOLON
Four positions, non-negotiable.
Pre-filter, not replacement.
Test 50 hypotheses in an hour. Take the 3 survivors to real users. Holon doesn't replace research — it protects its budget.
Built to say no.
Role-based friction coefficients. Complex-contagion gates for deliberative buyers. Negative pressure exempt from sycophancy. Our agents resist on purpose.
Distributions, not oracles.
Every verdict ships with a p10–p90 interval over 500 seeds. Every backtest publishes its method and a 'what the model misses' section. Directional confidence — that's the only honest claim.
GDPR-native by design.
EU hosting only. No cookies. No IP storage. Automated purge. Free runs use public benchmarks — your data only enters when you calibrate.
Twelve archetypes, each with a documented reason to resist.
Real B2B buying committees aren't panels — they're networks. Champions amplify, CFOs gate, Lurkers vanish silent. Tap or hover any citizen to see the parameters that move them.
MOVERS
LIFT THE CURVEGATES
STOP THE CASCADEQUIET WEIGHT
DECIDE OFF-SCREEN* Compliance bias is the #1 documented failure of synthetic research. We engineer against it with role-based friction and complex-contagion gates. Field Note 02.
The method, in the open.
Every note ships with its math, its parameters, its sources, and a reproducibility section. Check our work — don't just trust it.
INTEREST CHECK
What are you trying to pressure-test first?
Pick the decision that's live right now. We invite beta users by urgency and fit — not by signup order.
Pick a decision to claim your beta seat →
SECTION 06
The hard questions, answered.
Directionally confident, not an oracle. Every verdict ships with a p10–p90 interval over 500 seeds. We publish backtests with method, seed, and a 'what the model misses' section. On calibrated cities (tier Scale) we report a per-account MAPE — never a global accuracy claim.
GDPR-native: EU hosting only, no IP storage, no cookies, automated purge via retention policies set per tenant. Free-tier runs use public benchmarks — zero real personal data needed. For calibration, your data stays in your EU project under Article 28 terms.
No. Holon is the pre-filter. Test 50 hypotheses here in an hour; take the 3 survivors to real users. That's the whole pitch — protect your research budget for the questions humans answer better.
Two things: a graph (agents influence each other through directed weighted ties) and friction (role-based resistance coefficients that fight the LLM compliance bias). The engine is pure vectorized math — zero LLM calls per simulation tick.
Holon's engine implements published work on diffusion (Bass), complex contagion (Centola), negative-WOM asymmetry (Baumeister, Anderson), and social churn (Nitzan & Libai). See Field Note 01 for the full reference list.