An open-source methodology for building AI agent systems reliable enough to certify and insure.
AI consulting and a methodology for reliable agents.
Reliable Software = Spec Quality × Harness Enforcement × Continuous Evaluation
Methodology and book by Jhonn Moreno
Seamless forward momentum.
AI consulting from Vancouver, British Columbia
Attacca is operated by UNIFY ORCHESTRATION AI INC., a Vancouver-based British Columbia corporation. Start with an AI readiness assessment, plan your AI investments with Navigator, or organise your business information with Business Brain. Our open-source methodology explains how we approach specifications, evaluation and human oversight.
Free AI readiness assessment
A 10-minute diagnostic for Canadian mid-market leadership teams. Assess intent, workflows, data, trust and exposure to platform changes.
AI strategy consulting for BC companies
A three-month, CAD $25,000 engagement to diagnose opportunities, prioritise AI investments and specify your top three bets.
An AI-ready business knowledge base
Organise Google Drive information and put one AI workflow to work on your files, with a database and repository in accounts you own.
The bottleneck
The model isn’t the problem. The spec is.
AI agents generate code in minutes. But ambiguous specs produce ambiguous software. Agents don’t ask clarifying questions — they assume.
If each of 10 independent steps succeeds 90% of the time, the whole workflow succeeds about 35% of the time. Across 10 runs, that means about 6.5 failed runs on average.
The bottleneck isn’t the model. It’s the organizational context — the intent, the edge cases, the knowledge that took decades to build — that never makes it into the specification.
Attacca is a methodology that encodes that context into machine-actionable specifications, enforces the process through deterministic rails, and evaluates the output through structured adversarial testing.
The March of Nines
| Per-Step Success | 10-Step Workflow | Failures / Day (10 runs) |
|---|---|---|
| 90% | 35% end-to-end | ~6.5 |
| 99% | 90% end-to-end | ~1 |
| 99.9% | 99% end-to-end | ~0.1 |
The framework
Eight phases. Every gate enforced.
- 01IDEA
Capture the intent. Classify the risk.
- 02DISCOVER
Map the existing codebase.
brownfield only - 03SPEC
Write a specification precise enough for autonomous agents.
- 04BUILD
Execute on deterministic rails. Shift-left validation.
- 05TEST
Behavioral scenarios + factorial stress testing.
- 06CERTIFY
Human sign-off at tier-appropriate depth.
- 07DEPLOY
Production setup with verified checklists.
- 08MAINTAIN
Continuous flywheel. Drift detection.
- 01IDEA
Capture the intent. Classify the risk.
- 02DISCOVER
Map the existing codebase.
brownfield only - 03SPEC
Write a specification precise enough for autonomous agents.
- 04BUILD
Execute on deterministic rails. Shift-left validation.
- 05TEST
Behavioral scenarios + factorial stress testing.
- 06CERTIFY
Human sign-off at tier-appropriate depth.
- 07DEPLOY
Production setup with verified checklists.
- 08MAINTAIN
Continuous flywheel. Drift detection.
Every phase produces artifacts. Every gate has entry requirements. No phase can be skipped.
Trust architecture
Not everything needs the same rigor. Safety-critical systems need all of it.
- Tier 1
Deterministic
Annoyance if wrong. 7 scenarios. Full auto.
- Tier 2
Constrained
Wasted resources. Stress test variations. Logging.
- Tier 3
Open
Financial/reputational damage. Intent contract. Human oversight.
- Tier 4
High-Stakes
Legal/safety/irreversible. Domain expert. Full coverage.
A book · Out now
Specs In, Software Out.
The practitioner’s book on spec-driven AI development — why the bottleneck moved from code to specification, and how to build agent systems reliable enough to certify and insure.
15 chapters · 4 appendices · ~85,000 words · free to read
Start with a 10-minute readiness audit.
Follow the build.
Framework updates, case studies, and lessons from building AI agents that need to be right — not just fast.