Executive summary · 2-minute scan Marco Avila · v1 · 2026-05-12

Pilot-first launch.
PMF before ARR.

A two-minute version of Marco's first GTM proposal at Zunou. The full 25-minute writeup, with derivations and sensitivity tables, sits at the canonical URL.


The argument, in one paragraph

Knowledge teams have stopped tolerating tool fragmentation. The architecture Zunou has shipped (MIT-aligned tool-based selective retrieval, 136+ production tools, Voice Agent on Realtime API in 18 languages, autonomous Relays) is the right shape for the moment. The launch problem is distribution + density, not product readiness. The recommendation is pilot-first community-led launch in Tokyo: TAI in end July / early Aug 2026 (Ilya-introduced), then layer in AI Tinkerers + Venture Café + one more. Hybrid venue sponsorship pays for actual conversion, not just seats. Spaces Event-Spaces is the Zoom-playbook viral surface alongside it. Pre-committed stage-gate Feb 2027: ≥4 of 6 PMF criteria → fuel; ≤2 → pivot; 3 → extend 60 days. Year-1 ARR ambition is ¥30M+, locked after month-3 paid signal, not before. PMF first. ARR last.

01 · The product is real

Three facts that anchor everything below.

MIT-aligned architecture

Tool-based selective retrieval. 136+ production tools. ~100× token reduction vs context-stuffing. Built before the MIT paper that validated it.

Voice Agent · 18 languages

Sub-second, on OpenAI's Realtime API. Shipped, not roadmap.

Autonomous Relays

Cross-tool actions without prompting. The product is already deeper than what zunou.ai shows publicly.

02 · The horizon · three windows

Year one runs in three phases. The ARR number locks at the end, not the beginning.

Months 1–3 · Activation
Does the product produce dense behavior?

Magic-number completion. Cohort retention. Qualitative "would be very disappointed" signal.

Months 3–6 · Paid validation
Do people pay? At what tier?

First paying logos. Pro $19 vs Business $39 preference. Early churn.

Months 6–12 · Compound
Do parallel tracks compound?

+communities. +events. +content / PR. MoM acceleration. ARR number locks here.

03 · The launch shape

The pilot

TAI · end July / early Aug 2026 (Ilya-introduced). 4,000+ Tokyo AI builders. Open membership, weekly events. Pilot deep first; learn; double into AI Tinkerers Ginza, Venture Café Toranomon, and one more. Reachable Tokyo population across 13 prioritized communities: 6,500 individuals.

The stage-gate · Feb 2027

Pre-committed, six months after Phase 1 launch. ≥4 of 6 PMF criteria → fuel. ≤2 → pivot. 3 → extend 60 days. Removes the founder-bias trap of "let's give it three more months" indefinitely.

04 · Year-1 ambition

¥30M+ ARR is the ambition we're working toward. The beachhead (TAI + a few communities) covers months 1–2 of PMF learning. The year-1 number compounds from parallel tracks layered on top: more communities, Spaces-powered large events (IVS-class), content + PR, partnerships.

The specific ARR number locks after month-3 paid signal, not before. Faking the forecast earlier would let founder-bias replace data.

What I'd like us to discuss
  • · The launch shape — pilot-first (recommended) or synchronized 4-community?
  • · The year-1 framing — ¥30M+ as the ambition, locked after paid signal?
  • · The hybrid venue-sponsorship model — flat fee + per-user-join bonus?
  • · Which 3 capabilities the TAI demo leads with, and which we hold back?
What I'm uncertain about
  • · The 5 / 1 / 3-in-14-days magic number is a working hypothesis. Real activation surfaces (Daily Debrief, Relays, Spaces joins) may substitute.
  • · The 10% personal-conversion rate from Event-Spaces is the investor-doc number. Tokyo-specific actual is TBD.
  • · Whether parallel tracks compound or run linearly is the biggest unknown. Year-1 ambition assumes compounding.
  • · Phase 0 capacity is plausible-not-committed. What can the team realistically absorb?

The next move

Vote on the 12 open decisions.
Push back where the inputs feel wrong.

This is a discussion document, not a finished plan. The full version below is the show-your-work version, with derivations, the community attack list, sensitivity tables, and references. Disagree with specific lines, not the whole shape.