02 is the product plan; 03 is the model. This document is the company on a timeline: what we intend to earn, hire, and spend over three years, and — in keeping with how this room works — the gates where the plan changes course, stated in advance. Figures are targets, not forecasts: they are the numbers we resource against and measure ourselves by, not predictions we claim to know.
| 2027 — Prove | 2028 — Convert | 2029 — Compound | |
|---|---|---|---|
| Theme | Paid evidence in each market | Pilots become recurring licences | Usage, renewals, and channels do the selling |
| Games | 2–3 design partnerships, LOI-backed, at least one paid | First world licences from converted partners | First shipped title paying per-NPC/per-player usage |
| Urban twins | First paid build engagement (Bologna as reference); EU proposals submitted with consultancy partners | First run licences; the first calibration-report renewal — the vertical’s moment of truth | Twin licences renewing; ask-phase counterfactual studies as attach revenue; consultancies reselling |
| Defence & disaster | One lighthouse civil-protection exercise, designed to be publishable | Repeat exercises at 2–3 organisations; first readiness licence | Prime/integrator channel active; exercises booked without us in the room |
| Engine | Runway Horizons 1–2 (02) | Horizon 3: enterprise shape, metering, multi-tenant | Standards layer shipped; calibration benchmark public |
| Team | 3 founders + first 2 hires | 6–8 people | 10–13 people |
| Funding | Seed closed; grant proposals in | Grant co-funding active; Series A optional, on metrics | Base case: default-alive on revenue; A only to accelerate |
The year has one job: turn running code into paid evidence in each market, however small the invoices. Evidence, not revenue, is the Year-1 output; the revenue follows it.
Games. Execute runway Horizon 1: the evaluator sandbox, the public proof surfaces, population scale. Commercial target: 2–3 design partnerships, LOI-backed, at least one of them paying for the pilot. The GDC-anchored motion from 03 runs all year; the commercial lead (first hire) owns the list and the close.
Urban twins. Bologna graduates from pilot to reference: a public, periodically updated calibration report on open data — the artifact a consultancy can put in front of a city. Target: one paid build engagement through a consultancy partner, and 2–3 EU proposals submitted with consortia (Horizon-class calls fund exactly the build phase; their decision cycles mean money submitted in 2027 lands in 2028 — we submit early precisely because of that lag).
Defence & disaster. One lighthouse exercise with a civil-protection organisation: scoped small, priced honestly (subsidised if the reference rights justify it), designed from day one to be publishable. The deliverable that matters is the after-action data package — it is the sales artifact for every exercise after it.
Team. First hires, in order: the commercial lead (the deck’s open role — owns games pipeline and the twin channel partnerships) and one senior engineer (population scale and enterprise hardening). Founders stay the twins/defence delivery capacity this year.
Exit criteria for the year: at least one signed design partnership, one paid twin engagement, one delivered exercise. Three markets, three first invoices.
The conversion year: everything piloted in 2027 either becomes recurring or gets demoted — see the gates below.
Games. Design partners convert to world licences — the first real recurring revenue line. The deck’s 18-month milestones come due mid-year: 10 paying customers or $1M ARR, 3 platform partnerships, 1000+ active API users. Horizon 2’s designer product is what makes the licence renew: the creative-director surface, not the CTO demo.
Urban twins. Build engagements flow into run licences. The single most important event of the year is the first calibration-report renewal: a city (or its consultancy) paying a second year because the report proved useful is the falsifiable test of the “calibration report is the SLA” thesis. EU-funded projects land and co-fund the build teams.
Defence & disaster. The lighthouse becomes a repeatable product: the same exercise shape delivered at 2–3 organisations, first readiness licence signed (the standing twin between exercises). Begin the prime/integrator conversations with delivered-exercise evidence rather than slideware.
Team. Grow to 6–8: a twins delivery engineer (the consultancy-facing role), a developer-relations/support engineer (the games funnel demands it by now), and a second core engineer. Still one commercial lead; channels are how we scale sales before salespeople.
Funding decision point (mid-2028). With the 18-month milestones measured, a Series A is optional: we raise to accelerate a motion that is demonstrably working, never to survive. The seed plus grant co-funding plus Year-2 revenue is planned to carry the company into 2029 at base-case burn.
Games. The first shipped title on the engine turns on usage revenue (per-NPC / per-player) — revenue that grows with the customer’s success rather than our sales effort. Horizon 3’s protocol stewardship makes the formats the default: every third-party shell and client strengthens the licence, not the alternative.
Urban twins. The motion inverts: consultancies resell the build phase, we keep the run licences and the ask-phase studies. Renewal rate is the KPI that matters; a twin that renews twice is infrastructure, not a project.
Defence & disaster. Exercises booked through the channel without us in the room; readiness licences at the organisations that exercised in 2028. Defence-proper engagements only via primes holding the frameworks, inside the published dual-use guardrails.
Team. 10–13: second commercial hire (channel management), an exercise programme role, engineering to ~6. Still small on purpose — channels and usage carry the growth.
Financial shape. Base case: default-alive — recurring revenue covers burn without the A. Targets, not forecasts: exit 2029 at $2.5–4M ARR, mixed roughly half games licences+usage, a third twins run/ask, the balance exercises and readiness — with the mix itself treated as a finding, not a commitment: one engine means we can follow whichever market converts fastest without rework.
| Gate | If it fails | Consequence |
|---|---|---|
| Mid-2027: ≥1 LOI-backed games design partnership | Zero LOIs despite the sandbox and the GDC motion | Games demotes to funnel-and-OSS presence; commercial lead re-points at twins. The engine roadmap barely changes — that is what one-engine-three-markets buys |
| End-2027: one paid engagement per new vertical | Twins or defence produce no paying pilot | The failing vertical loses its 2028 hiring slot and runs founder-led and opportunistic until a partner or grant changes the evidence |
| End-2028: first calibration-report renewal | Twin customers take the build, skip the run | Twins reverts to a project business through consultancies (still profitable, no longer a recurring thesis); the run-licence pitch is rewritten from evidence |
| End-2028: repeat exercise booking within 12 months of the lighthouse | One-and-done | Defence & disaster stays a capability we sell opportunistically, with no dedicated headcount, until the channel produces demand |
| Any point: the academic branch requests headcount | — | It doesn’t get it. Collaboration time, yes; a hire, no. It is leverage, not a line (03) |
| 2027 | 2028 | 2029 | |
|---|---|---|---|
| Design partnerships (games, LOI+) | 2–3 | convert ≥2 to licences | licences renewing + first usage |
| Paying customers (all markets) | 3–5 | 10+ | 20+ |
| ARR | first revenue | $1M target | $2.5–4M target |
| Twin calibration reports delivered | 1 (Bologna, public) | 3–4, first renewal | renewals > new builds |
| Exercises delivered | 1 lighthouse | 3–5 | channel-booked majority |
| EU grant funding secured | proposals in | first projects running | co-funding ≥ research spend |
| Team | 5 | 6–8 | 10–13 |
| Runway (months, base case) | 20+ post-seed | 12+ without A | default-alive |
No assumption that a model provider builds our layer for us, or against us — the category defence holds either way. No assumption that any single market carries the company: the plan survives any one of the three converting slowly, and the gates say which lever moves when it does. No headcount ahead of evidence: every 2028 and 2029 hire is attached to a 2027 or 2028 gate having passed. And no revenue from anything we said we would never do: inference, attention, the owner’s data, or work outside the dual-use guardrails.