AI changed who writes code. Kin changes what they build on.
A new code repository built around connected software.
Troy Fortin / Founder / Firelock, LLC / September 2026
The problem
The same map
gets rebuilt.
Change a shop's discount rule. The cart, checkout and receipt may share the same code.
The next agent and reviewer can repeat that investigation.
The solution
Keep the map
with the code.
Kin keeps code, recorded relationships and history in one inspectable repository.
The code and its recorded structure move through commits and branches together.
Build from a persistent record.
Not only the current session.
Repository model from supplied founder deck. Kin can miss connections; compilers, tests and review still matter.
Why now
AI adoption is widespread.
Confidence still takes work.
Review burden
85%
agree AI shifted the bottleneck to reviewing and validating code.
Generating a change is not the same as establishing confidence in it.
Different surveys, questions and populations. These indicators are not a combined statistic. Sources and scope in A5.
Product evidence
Ask the repository
what connects.
8
Recorded callers.
One React helper connects to input, select and text-area code.
Kin v0.7.3
React 9b7a0d40
kin refs getToStringValue --kind callsReactDOMInput.js
hydrateInput
initInput
updateInput
ReactDOMSelect.js
hydrateSelect
updateOptions
ReactDOMTextarea.js
hydrateTextarea
initTextarea
updateTextarea
Recorded callers, regrouped by file. Not a new live run.
Supplied September 7 deck, p. 8. Single-commit snapshot; call coverage partial. Full provenance and capture limitations in A1.
The distinction
Context beside Git.
Or part of the repository.
Git + separate context
People and agents query the derived record.
Kin's repository model
and history, togetherCommits and branches include the structure.
People and agents query repository state.
The bet: a shared software record that stays with the revision being changed.
Indexes can be persistent and revision-aware. Selected models, not a performance ranking. Product examples and official sources in A5.
Market and entry
Start where context
already costs time.
Initial target hypothesis
AI-heavy teams.
20 to 200 engineers.
- Buyer
- VP Engineering or platform lead.
- Trigger
- Agent-driven changes outpace review capacity.
- First workflow
- Shared-code changes in one supported repository.
The scale of the workflow
43.2M
pull requests merged per month on GitHub in 2025.
Workflow volume, not Kin's market size.
One engineer and their agent.
One recurring change.
Beside the existing Git workflow.
GitHub Octoverse 2025. Narrowed segment proposed in the review. Bottom-up sizing and pricing tests in A3.
The business
Earn the developer.
Become the team's shared record.
Kin
Local repository and graph queries.
CLI, editor and MCP access.
No hosted account required.
KinLab
Shared repository state.
Team access and permissions.
Persistent change evidence.
Active-developer subscription or team fee plus usage.
Test willingness to pay through fixed-scope paid evaluations.
Local Kin is Apache-2.0. Hosted paid scope and pricing are design-partner hypotheses, not available self-service features.
Progress
The alpha is built.
Now earn repeat use.
6 months
Founder-built public alpha.
Native repository operations.
Git import and export.
CLI, editor and agent access.
Early conversations / September 8
Measure next
Founder-confirmed September 8 snapshot: no revenue or signed design partners. Installation completion and retention have not been established.
The founder
Built from
firsthand experience.
Troy Fortin
Machine learning engineer building production generative-AI systems. Previously at The Home Depot.
Architected and shipped Kin's public alpha solo.
Self-funded. Full-time on Kin at financing.
First technical hires / risk removed
Repository / runtimeProduction reliability
Developer / agent integrationsAdoption and repeat use
Cloud / platformShared deployment and security
Founder materials, September 7, 2026. Firelock, LLC. Hiring is staged against product readiness and customer pull.
What the round must produce
Capital should buy evidence.
Not just features.
01 / Reliable use
Qualify the workflow.
5 to 10 external repositories in daily-use qualification.
4 weeks of continuous Kin-on-Kin use.
Track activation, coverage and setup cost.02 / Retained teams
Earn recurring work.
5 to 7 design partners.
Measure four-week return and repeated use on real changes.
Expand from one workflow to another.03 / Paid evidence
Prove the commitment.
2 to 3 paid pilots.
At least 2 referenceable customers.
Independent reproducible evidence.
Track conversion, expansion and margin.Operating targets for the 24-month plan, not achieved results. The repository count is a qualification cohort, not an acquisition ceiling.
The next chapter
Build the software record
for the agent era.
Raising seed financing
From public alpha
to daily team use.
A 24-month plan for repository reliability, agent adoption and shared KinLab workflows.
What this enables
Retained external use.
Paid team workflows.
Independent evidence.
Firelock, LLC
Operating plan and capital allocation in A4.
Appendix / A1 / Recorded evidence
The result is inspectable.
So are its limits.
Eight recorded callers of getToStringValue, grouped by file.
ReactDOMInput.js
hydrateInput
initInput
updateInput
ReactDOMSelect.js
hydrateSelect
updateOptions
ReactDOMTextarea.js
hydrateTextarea
initTextarea
updateTextarea
Recorded callers of getToStringValue
v0.7.3 / React 9b7a0d40. Single-commit snapshot. Full history was refused on that build.
Reported JavaScript call coverage: 13,819 of 36,234 / 38% / partial.
Historical recording in supplied September 7 deck, p. 8. Re-typeset, not rerun. It does not establish speed, completeness or end-to-end correctness.
Appendix / A2 / The next evidence
Prove the workflow.
Not only the query.
Compare Kin and an indexed Git workflow on the same pinned change.
Hold constant
One controlled task.
Repository revision, change request, model, context budget and tests.
Include graph setup and maintenance, not only warm queries.
Measure the outcome
Useful work, end to end.
Time and tool calls to establish impact. Missing connections. Review rework. Finished-change quality.
Repeat on unfamiliar supported repositories.
Coverage, import scale and workflow support remain alpha constraints. Compilers, tests and review remain necessary.
Proposed evaluation, not a completed experiment. Live proof scope separates older retrieval results from the shipping release; no current accuracy or savings claim is added.
Appendix / A3 / Market and pricing
Size the buyer.
Test what they pay for.
accountsCount the initial segment
per accountMeasure participation
per userTest paid commitment
Annual seat-based opportunity model. A team-plus-usage plan needs a separate account model.
Test 01
Active developer.
Subscription tied to actual use.
Test 02
Team + usage.
Shared platform fee with usage.
Test 03
Paid evaluation.
Fixed scope and an explicit fee.
The review proposes these experiments. No validated account count, price, market total or future revenue is supplied or implied.
Appendix / A4 / Capital plan
Fund the team.
Stage the commitments.
24-month base plan.
Founder plus three technical hires. No uncontracted revenue assumed.
A smaller close changes the scope.
The separate 18-month plan is not the same milestones for less money.
Go-to-market hiring is gated on at least two paid pilots and repeatable onboarding.
Planned allocation and operating assumptions from the supplied founder deck, pp. 14 and 15.
Appendix / A5 / Sources
Clear sources.
Bounded claims.
Company and evidence
Founder deck, September 7, 2026: product model, React recording, outreach, founder and operating targets.
Positioning review supplied September 8: narrowed buyer, pricing experiments and evaluation plan.
Kin source
Current public proof scope
The React result is historical. Outreach numbers reconfirmed September 8. Conversations with developers, potential design partners and investors have just begun.
Industry signals
DORA / 2025 and 2026
90% AI use at work.
Stack Overflow / 2025
46% distrust AI accuracy.
GitLab / Harris Poll / 2026
85% review-bottleneck agreement; 1,528 respondents, six countries.
GitHub Octoverse / 2025
43.2M merged PRs per month.
Selected alternatives
GitLab Orbit / beta
Sourcegraph / SCIP indexes
Augment / Context Engine
Greptile / graph context
GitHub / Copilot review
Official product descriptions checked September 8, 2026. Not a feature ranking or a measured comparison.
Indexing does not mean ephemeral context or absence of revision support.