The evidence behind the argument.

One ecosystem. Separate responsibilities.

The architecture below is a composition map. Solid associations describe documented responsibilities; the hosted operating layer is proposed. No automatic universal integration or policy enforcement is claimed.

Capability and maturity matrix

Inspect the recorded run

Authentic local execution; deliberately seeded fixture. The assistant authored a small coding task, recorded its failure, corrected the formula and ran the checks. This is not a customer incident, live production agent run or independent attestation.

Select an artifact to inspect it.

CaseFile captures command exit status and bounded output. Eval checks the explicit conditions. Files are trusted-local artifacts; SHA-256 detects changes relative to this manifest, not host compromise.

A correctness fixture, not a performance claim

Method: Node assertions check 6/30 = 20% and 3/30 = 10%; the initial denominator intentionally adds the raise and fails. CaseFile captures both executions. Eval runs the assertion command and checks the corrected source expression. Observed result: two configured Eval checks passed, with the earlier assertion failure retained.

Reproduce with python3 scripts/record-demo.py from the source deck, using the sibling CaseFile, Eval and Kujo 1.9.0 checkout. This overwrites the recorded fixture artifacts; retain the original before comparing runs. Timings are recorded but not a statistically meaningful benchmark. No competitor benchmark, latency advantage, comprehensive correctness or security certification is claimed.

Future benchmark plan: publish fixed repositories, pinned environments, task corpus, failure taxonomy, complete attempts, repeated seeds and independent reproduction. Measure reviewer time, evidence completeness and failure recovery alongside task success.

Strong alternatives define the bar

Representative comparison, source/documentation review dated 2026-11-09, not hands-on parity testing. GitHub Copilot and LangGraph already offer substantial capability. Observability systems such as OpenTelemetry standardize telemetry; CI systems such as GitHub Actions execute workflow checks. Kujo should complement or replace specific responsibilities only after evidence.

Potential defensibility: stable evidence contracts, integration quality, reproducible operations and retained developer use. These are future advantages to earn. Open interfaces reduce lock-in and ease adoption; no exclusive technology moat is asserted.

Edit the bottom-up assumptions

All counts and contract values are unvalidated planning assumptions. The model estimates a narrow organization-level service opportunity, not the entire AI industry.

Expansion scenario: 10,000 organizations × $60K = $600M annual pool. This is a separate, longer-term hypothesis, not additive TAM. Validate organization counts, budget ownership, competition, retention and willingness to pay.

Proposed pricing needs customer evidence

Commercial validation sequence: instrument active repositories → investigate repeated use → recruit design partners → paid pilots → retention and expansion. Infrastructure cost, support load and gross margin require actual usage data.

Funding scenarios and dilution

Working scenario: $6M at $30M post-money. These are planning inputs, not agreed terms or validated valuation. Simplified ownership = raise ÷ post-money; SAFE mechanics, option pools and other instruments can change outcomes.

At fixed $30M post-money: $3M → 10%; $6M → 20%; $8M → 26.67%; $10M → 33.33%. A later round that dilutes all holders by 20% reduces a 20% holding to 16%.

$3M prioritizes one flagship workflow and a smaller team. $6M supports parallel runtime, execution and developer experience work. $8M/$10M permit more integration and commercial capacity, conditional on evidence of demand.

Twenty-four-month operating plan

$6M case: $3.30M engineering; $600K infrastructure; $600K distribution; $600K operations; $900K reserve. Non-reserve operating average: $212,500/month. Gross capital divided by 24 months: $250,000/month, including reserve.

Founder base: $150K/year, or $300K over two years. Illustrative 25% payroll burden makes $375K, included in engineering. Remaining engineering: $2.925M; at $240K per engineer-year, approximately 12.19 engineer-years. Location and hiring ramp materially affect this model.

Reserve stays $900K. No revenue offset assumed. This simplified allocation is not a monthly cash-flow forecast. Benefits, taxes, legal, security and customer support must be priced before commitments.

Hire against measurable milestones

1. Months 0–6

   Hire execution reliability and developer experience engineers; harden one flagship evidence workflow. Publish reproducible benchmark protocol and recruit external reviewers. Proposed target: 10 independently activated teams.

2. Months 6–12

   Expand runtime/integration capacity after activation evidence. Proposed targets: 5 design partners, 3 paid pilots and measured four-week retention. Publish all benchmark limitations.

3. Months 12–24

   Add customer engineering when support demand is demonstrated. Test repeatable acquisition, renewal and expansion. 100 customers at $24K is a scenario to test, not a committed forecast.

December 2026 campaign is planned. Measure unique activated repositories, successful evidence runs, repeat usage, qualified conversations and paid conversions by cohort. Partnerships are prospective.

Technical ambition must survive commercial reality

Language adoption

New language adoption may take years. Sell useful workflow infrastructure independently of migration.

Integration complexity

Tools have distinct contracts and runtime pins. Fund one reliable flagship experience before promising seamless coverage.

Trust and security

Editable local evidence is not independent attestation. Hosted custody, tenant isolation and retention need dedicated design and review.

Incumbents

Agents, CI and observability vendors can expand into accountability. Prove lower operational burden and repeat use.

Demand and pricing

No verified customer, retention or revenue data supplied. Validate the beachhead and pricing through actual pilots.

Founder concentration

Broaden maintainership, documentation and release ownership. Confirm contributor attribution and biography before distribution.

Source register and investor follow-up

Research cut: 2026-11-09. Release inventory dated 2026-10-22 is not a blanket current-HEAD guarantee. Full claim register is retained in source at research/CLAIMS.md.

• T01 / kujo / pinned source
• T02 / dispatch / pinned source
• T03 / workcell / pinned source
• T04 / eval / pinned source
• T05 / spec / pinned source
• T06 / runledger / pinned source
• T07 / shipcheck / pinned source
• T08 / casefile / pinned source
• T09 / watchdog / pinned source
• T10 / kennel / pinned source
• T11 / mcp / pinned source
• T12 / packwrite / pinned source
• T13 / kujo-workflows / pinned source
• T14 / presentations / pinned source
• M03/M04: Stack Overflow 2026 AI survey
• F01: Robert DeVore public biography
• Release inventory
• Kujo official site

Before investor distribution: supply verified adoption and commercial metrics, confirm founder attribution and contact details, validate pricing/budget assumptions, and obtain financing advice on actual instrument terms. No current customer or investor commitment is represented.