Public roadmap · updated May 2026

From a leaderboard and a family app to the trust layer every edtech platform deploys.

Kumuao v1 is two products: an open benchmark and a parent-facing iOS app. Everything else on this page — the safety API, certification, schools platform, and a possible foundation model — comes later, in the order listed here, for reasons we explain in the "bets we're making" section.

How to read this roadmap: We commit to order and direction. Calendar dates beyond the current quarter are estimates — they'll shift as we learn. Phases are separated by signals, not hard gates. If a signal isn't there, we stay in the phase until it is rather than rushing to the next one.

Products on this roadmap

Kumuao Bench: Our open benchmark — tests AI chatbots against child-safety scenarios across 8 dimensions and 4 developmental stages. The leaderboard is public.
Kumuao Spark: A native AI chat for kids inside the Kumuao Family app, powered by Kumuao's safety and education optimization layer. Safe by design, not by filter.
AI Interaction Monitor: A browser extension (+ Snapchat Family Center) that gives parents topic-level summaries of what their child talked about on other AI platforms.
Guard API: A safety classification layer edtech companies embed into their own products — real-time filtering and flagging powered by Kumuao Bench methodology.

Shipped — live today Building — work in flight Planned — committed scope Vision — strategic bet, scope TBD

Today: May 25, 2026
Live: 1 product (Bench v0.1.0)
Building: 2 (Family iOS · pre-seed)
Horizon: 30 months

The arc

Six phases. Rough timing, firm order.

Each phase has a primary goal and a signal we're looking for before shifting focus. Dates are our best current estimate — the order is what we're actually committed to.

Q2 2026 · now
Credibility

Ship the benchmark. Get the family app into beta. Close pre-seed.
Signal: Benchmark generating conversation. Beta waitlist growing.
Q3–Q4 2026
Early traction

Bench v1.0 published. Family iOS soft launch. First edtech pilots on Guard API.
Signal: Pilots using the API. Family showing early retention. Seed process started.
Q4 2026–H1 2027
Product–market fit

Bench v2.0. Guard API broadly available. Self-service certification. First district pilots.
Signal: Paying API customers, pricing model validated, at least one district in production.
H1–H2 2027
Platform expansion

Tutor API when Guard PMF is clear. Android. Education fine-tune. Standard cert program.
Signal: API revenue growing month-over-month. NRR healthy. Clear next product to build.
2027–2028
Scale & enterprise

Complete API. International expansion. SOC 2. Series A when the numbers support it.
Signal: ARR trajectory strong enough to run a credible growth round.
2028+
Market leadership

Foundation Model (if margin justifies it). Multimodal. 50+ countries. Procurement standard.
Signal: Sustainable unit economics. Path to profitability visible.

Workstreams

Seven parallel tracks. They feed each other.

The benchmark generates safety signal data. That data trains the API. The API generates usage data. Usage data sharpens the benchmark. Each track is a flywheel input for another.

Q2 '26 now

Q3 '26

Q4 '26

H1 '27

H2 '27

2028+

Benchmark Open-source measurement

v0.1.0 · 100+ tests v1.0 + paper v2.0 — 2,400+ cases Multimodal extension Localized · 50+ countries

Family B2C · iOS-first

iOS beta · TestFlight iOS soft launch → public Android beta Android GA · web · siblings

API B2B · core revenue

Guard · private alpha Guard GA (when pilot feedback is strong) Tutor (Tier 2) Complete + Foundation

Certify Compliance program

Self-service eval Standard cert program Premium & Enterprise

Schools B2B2C · per-student

1–3 district pilots LMS integrations · grow districts Scale · 100+ districts

Trust & compliance SOC 2 · regulators

State of AI Child Safety v1 SOC 2 readiness → Type II UK · EU packs · regulators

Funding Capital milestones

Pre-seed · $750K Seed · $3–4M Series A · when numbers support it

Time periods are quarterly to mid-2027, then half-yearly, then annual. The further out you look, the less precise the boundaries are — that's deliberate. We commit to dates inside 12 months and to order beyond that.

The detail

What's in each phase, milestone by milestone.

Now · Q2 2026

Credibility

Goal — get the benchmark talked about and the family app in real parents' hands.

ShippedKumuao Bench v0.1.0Leaderboard live. 102 tests + 15 sequences, 4 stages, 8 dimensions.
BuildingKumuao Family · iOS TestFlightKumuao Spark (child chat) + AI Interaction Monitor + Ask Kumuao + monthly report card. Invites by waitlist.
BuildingPre-seed close · $750KMission-aligned funds + angels. Funds 6 months of build and first hires.
BuildingCo-founder recruitingCTO (T&S background) + Head of Safety (NCMEC/Thorn/IWF).

Q3–Q4 2026

Early traction

Goal — get Kumuao into the conversation. Benchmark cited, app in real hands, first API pilots live.

PlannedBench v1.0 + research paper800+ test cases, eval harness open-sourced on GitHub. Target: cited in at least one press piece.
PlannedKumuao Family · iOS soft launchInvite-based rollout first, broad public when retention looks right.
PlannedGuard API · private alphaA handful of edtech pilots — we'd rather have 3 who love it than 10 who are lukewarm.
PlannedSeed raise · target $3–4MProcess starts when we have enough signal from the above. Targeting mission-aligned edtech funds.
PlannedState of AI Child Safety · v1First edition of what we hope becomes a reference report for the industry.

Q4 2026 – H1 2027

Product–market fit

Goal — validate that edtech companies will pay for child-safe AI infrastructure. Find the things that make them stay.

PlannedBench v2.0 · 2,400+ casesFull grooming sequences, regulatory mapping, public leaderboard refresh.
PlannedGuard API · broadly availableOpen to paying customers when pilot feedback is strong enough to feel confident in the product.
PlannedSelf-service certification$2K/run automated eval + PDF report. Keeps the cert motion going without needing sales.
PlannedFirst district pilot(s)1–3 districts. Paid, but the goal is learning the sales motion and building references.

H1–H2 2027 · if PMF is clear

Platform expansion

Goal — expand the API surface and the family product based on what we learned in the PMF phase. Order within this phase will shift.

VisionKumuao Tutor API (Tier 2)Education-optimized layer. Ships when Guard has product-market fit — not before.
VisionEducation fine-tune v1Llama-3 LoRA + DPO. Start when we have enough usage data to make it meaningful.
VisionFamily on AndroidWhen iOS retention and revenue justify the platform investment.
VisionStandard cert program$15K/yr tier when we have enough cert engagements to systemize it.
VisionGrow the district pipelineMore districts when we have references, an LMS integration story, and a dedicated sales motion.

2027–2028 · longer-horizon bets

Scale & enterprise

Goal — enterprise revenue, international compliance, and a growth round if the business supports it. Specific timing TBD based on what we learn.

VisionKumuao Complete API (Tier 3)Parent + teacher dashboards, assessment generation, white-label UI.
VisionSOC 2 Type IIEnterprise procurement requirement — will pursue when enterprise pipeline demands it.
VisionUK + EU compliance packsAADC + EU AI Act. Big opportunity, meaningful effort — sequenced after US PMF.
VisionEnterprise certification tierDedicated safety engineer + custom test cases. When we have the team to support it.
VisionSeries A · when the numbers support itNo target date set. We'll raise when ARR trajectory makes a strong growth story, not on a schedule.

2028 and beyond

Market leadership

Goal — be the standard. Be the thing regulators cite by name.

VisionKumuao Foundation Model v1Purpose-built child-safe education model. Only if API margin compression justifies it.
VisionMultimodal child safetyImage-generation safety added to Bench and Guard.
VisionLocalized benchmarks · 50+ countriesCultural and regulatory variants per jurisdiction.
VisionProcurement standard"Kumuao Certified" listed as a requirement in 100+ district RFPs.
Vision$10M+ ARR · path to profitabilityNet positive cash flow in early Year 3.

Bets we're making

Why this order, not some other.

A roadmap is mostly the story of which thing we're betting on second, given what we're betting on first. Here are the five sequencing bets — each one is the reason the next phase exists.

Bet 01

Benchmark before product.

We ship the leaderboard before the API because the benchmark is what makes Kumuao credible. Without it, we're just another wrapper. With it, every conversation about AI safety in education starts with our numbers — and that's distribution.

What kills this bet: if the benchmark doesn't get adopted as a reference within 6 months (citations, GitHub stars, press), we don't have the credibility to sell the API.

Bet 02

B2C before B2B.

Family ships before Guard API because parents convert in days and edtech buyers take months. Family generates revenue we need before the seed closes, gives us anecdotes for the seed pitch, and produces real-world safety signal that improves the API later.

What kills this bet: if early Family subscribers churn quickly and CAC doesn't fall with scale, the B2C unit economics don't work and we shift emphasis to B2B earlier.

Bet 03

API before certification.

Guard API ships before paid certification because the API is what generates the data certification will need to be credible. You can't audit something you've never operated.

What kills this bet: if foundation model providers ship deep child-safety features in 2026, the wrapper-API thesis weakens and we lean harder on certification + benchmark as the standalone business.

Bet 04

Districts after enterprise edtech.

School districts have 6–18 month sales cycles and budget-driven procurement. We're not equipped for that motion at Q3 2026. We sell to the edtech companies that sell to districts first, then go direct once we have references and a cert program.

What kills this bet: if a district RFP explicitly requires "Kumuao Certified" before our cert program is GA, we'll have to accelerate.

Bet 05

Fine-tune before foundation model.

LoRA + DPO on Llama is a 2-engineer, 3-month investment. A custom foundation model is 20 engineers and 18 months. We don't take the second bet until usage data proves the first one was the bottleneck.

What kills this bet: if foundation model unit economics never improve to where margin justifies our own model, we stay a fine-tune company forever — which is fine, but it changes the long-term cap table math.

If we're wrong

The roadmap that survives every scenario.

We've sequenced everything so that if a phase fails, the work behind it still has value. Here's what the path looks like in each failure mode.

If the API doesn't work

We become a measurement-and-certification company. Benchmark + Certify alone is a $5M+ ARR business (see Veracode, Checkmarx). The Family app still runs as a B2C product.

If foundation labs commoditize child safety

Our deep evaluation methodology becomes the audit standard regardless of who provides the underlying safety. We pivot from "build the layer" to "certify the layers" — Kumuao becomes the UL Labs of AI safety.

If a competitor raises $50M+

Open-source benchmark adoption is sticky once it's the reference. First-mover on the standard wins the standard. We accelerate the cert program and lean on data flywheel.

If edtech funding stays frozen

Pivot weight to schools (budget-driven, not VC-driven) and compliance (regulatory requirement, not discretionary spend). EU AI Act enforcement gives us a buyer regardless of the funding climate.

Changelog

What's changed in this roadmap.

2026-05-25 v0.2 · Roadmap restructure. Split v1 (Bench + Family) into its own surface. Added bets & failure-mode sections.
2026-05-10 v0.1 · Initial founder's design doc. Six-phase plan, seven workstreams, 30-month horizon defined.
2026-04-22 Pre-doc · Benchmark spec finalized. 8 safety dimensions, 4 developmental stages, 2,400+ target test cases set.

This roadmap is updated quarterly. Bookmark this page or follow the beta list to get changes by email.