Aurrin Ventures
× prd to prod

A Calgary startup accelerator evolving from a static site into a full-stack event and validation platform—built and maintained by an autonomous agent pipeline.

The challenge

Aurrin Ventures runs monthly pitch nights for early-stage founders in Calgary. They award microgrants, connect founders with judges and mentors, and track validation data from live audiences. Today, all of this runs on a Next.js marketing site with data stored in JSON files. Every content update requires a code change and a deploy.

What they need is a platform: event management, founder applications, dynamic scoring rubrics, audience validation via QR codes, mentor matching, Stripe payments, downloadable reports, and a public founder directory. Twelve core modules, six user roles, three delivery phases, and fifteen-plus database entities.

This is not a landing page refresh. It is a real product with authentication, real-time scoring, role-based access control, payment processing, and PDF generation.

Why agent-first development

The tools for autonomous software delivery exist now. GitHub shipped Agentic Workflows in technical preview. Copilot agents can implement features from issue descriptions. Claude and Codex can review code, diagnose CI failures, and reason about architecture. These are not code generators—they are agents that operate inside governed pipelines.

The traditional path for a project like Aurrin’s: hire a development team or agency, manage sprints, coordinate reviews, fix CI manually, and lose all context when the engagement ends. Every change after launch requires going back to the same team or onboarding a new one.

Agent-first development inverts this. You describe what you want as a GitHub issue. The pipeline decomposes it, assigns an agent to implement it, has a separate agent review it independently, merges when approved, deploys automatically, and self-heals when CI breaks. No sprint planning. No standups. No context loss.

This is the same shift that happened with cloud infrastructure: from “hire a sysadmin to rack servers” to “describe what you need and the platform provisions it.” The infrastructure for software delivery is making the same transition.

What Aurrin gets

A rough MVP. Not a polished product—an honest first build. Auth, database, admin dashboard, core modules. The starting point, not the finish line.

The autonomous delivery infrastructure. This is the real value. The repository comes with gh-aw workflows baked in:

• Auto-dispatch—file a GitHub issue describing a feature or bug. The pipeline picks it up, assigns an agent, and begins implementation.
• Independent review—a separate agent reviews every PR. The builder and reviewer are never the same identity.
• Self-healing CI—when CI breaks, the pipeline detects the failure, creates a fix issue, assigns an agent, and resolves it. Often before anyone notices.
• Continuous documentation—agents update docs as they change code. Documentation does not drift.
• Complete audit trail—every decision (who approved what, why, what evidence was presented) is traceable in the repo history.
• Deploy on merge—approved PRs merge and deploy automatically. No manual deploy steps.

The key point: Aurrin’s team does not need to call a developer to make changes. They describe what they want in a GitHub issue. The pipeline ships it. This works on day one and continues working indefinitely.

The infrastructure

A brief look at what actually lives in the repository. Not exhaustive—enough to understand the machinery:

• auto-dispatch.yml—routes labeled issues to the right agent
• repo-assist—the implementation agent that writes code from issue descriptions
• pr-review-agent—independent code review on every PR
• ci-failure-issue.yml—detects CI failures and creates fix issues automatically
• Autonomy policy—a machine-readable file defining what agents can and cannot do
• Identity separation—the agent that writes code never approves its own work

For the full thesis on harness engineering—why the orchestration layer matters more than the model—read the vision.

Where this goes

Aurrin is an early adopter of governed autonomous delivery. As the pipeline evolves:

• The pipeline learns from its own history—which patterns break, which decompositions work best
• Policy becomes configurable—Aurrin can tighten or loosen governance as the platform matures
• Multi-service orchestration—as Aurrin grows beyond a single repo, the pipeline coordinates across services
• The ecosystem grows—better agents, specialized reviewers, marketplace templates

The full unconstrained vision for where this leads is in the optimal vision document.

Aurrin Ventures gets a platform and a pipeline. The platform is the starting point. The pipeline is what makes it self-sustaining.

Read the full technical architecture →