The delivery infrastructure
for AI agents.
AI agents can write code, but shipping software requires decomposing specs, independent review, deployment, CI repair, and audit trails. We build the orchestration layer that governs the full pipeline.
Problem
Companies using AI to write code still hire engineers for the work around it: decomposing specs into parallelizable tasks, reviewing code where the reviewer is independent from the author, deploying, repairing CI failures, and maintaining audit trails. The models are capable. The orchestration around them is the bottleneck.
Solution
prd to prod is an autonomous delivery pipeline built on GitHub’s Agentic Workflows. You file a product brief as a GitHub issue. The pipeline decomposes it into tasks, assigns agents to implement each one, has a separate agent review every PR, merges on approval, deploys, and self-heals when CI breaks.
The governance layer is repo-owned: an autonomy policy defining what agents can and cannot do, identity separation between builder and reviewer, dependency-aware orchestration, and self-healing loops with independent review on every fix.
Traction
First client build: Aurrin Ventures, a Calgary startup accelerator, is replacing a static site with a 12-module platform (event management, judge scoring, audience validation, mentor matching, Stripe payments, public directory). 80 agent-merged PRs across 133 issues in its first 6 days, as of March 27, 2026. Build is ongoing.
Platform credibility: Top 3 leading community contributor to GitHub Agentic Workflows. 28 issues filed, 24 fixes shipped across 8 releases, 15 release credits. We surfaced bugs that GitHub’s own 165-agent fleet did not catch.
Signal: “The New OSS” endorsed by Peli de Halleux (gh-aw creator, Microsoft Research), amplified by GitHub Next. “The Agent Interface” reposted by Don Syme (F# creator, Microsoft Research). Three active threads with Microsoft/GitHub, all initiated by the other side.
Market
The first buyers are founders and small product teams building on the standard SaaS stack who need to ship software without hiring a full engineering team for every change. Code generation is commoditized. Orchestration, review, deploy, and recovery are not.
GitHub has 100M+ developers and is building agent-native infrastructure. Factory raised $70M. Mendral is automating CI self-heal. These companies are solving pieces of agent-first delivery. The governance layer is unsolved.
Business model
- Paid initial engagements. We either deliver a working app with the pipeline baked in or install the pipeline in an existing repo.
- Expansion revenue. Recurring software revenue for orchestration, policy controls, review boundaries, and self-healing workflows.
Services fund the wedge. The long-term product is the control plane for governed autonomous delivery.
Team
Samuel Kahessay. Solo founder. Former Amazon SDE (Alexa Edge ML), ATB Financial (data science), University of Calgary researcher (NSERC grant).
Built the pipeline, wrote the thesis, surfaced upstream platform bugs, and delivered the first client.
Ask: design partners
Looking for early-stage companies that need to ship software and want to be early users of autonomous delivery infrastructure. Each engagement hardens the product. Strongest fit: founders building on the standard SaaS stack (Next.js, Postgres, Stripe, auth).
Also building relationships with investors who understand developer tooling, workflow infrastructure, and AI systems.