Proof of work · systems architecture
Human + AI Operating Architecture
I decompose real operating work into decision rights, deterministic workflows, bounded agentic tasks, tools, retrieval, and human approval gates. The goal is not more automation; it is a system where parallel work can move without losing ownership, context, or control.
Career & Proof
Capability signal · market feedbackTurns completed work into visible evidence, market learning, and targeted opportunities.
- Headhunter workflowRole discovery · fit analysis · application packets
- Marketing researchPain → buyer → reachability → smallest useful test
- Capability LabDemos, architecture proof, and reusable case evidence
- Proof publishingTranslate operating wins into artifacts outsiders can evaluate
Mortgage Operations
Revenue system · human-in-the-loopProduction workflows keep judgment with the loan officer while automation handles repeatable coordination and enrichment.
- Database 2.0Operational system of record · intake, pipeline, funding, registry
- Database 3.0 cuesPre-qualified signals routed into the working desk
- Project MortgagePurchase and refinance execution paths
- Data enrichmentHomebot · Remine · title and property data
Knowledge & Orchestration
Context · retrieval · bounded delegationShared context, retrieval, preferences, controls, and orchestration support every lane without creating a separate “brain” for each one.
- Tech AdvisorBounded synthesis → one high-leverage recommendation
- Opportunity ResearcherResearch and evidence gathering across defined questions
- Follow Up orchestrationRoster assembly from operational context and enrichment
- Knowledge layer (“Compound”)Canon · preferences · live retrieval · write-back
Peregrine Solution Lab
Evidence-led experimentsTests named buyer problems before committing to a larger build, then converts validated work into reusable proof.
- Solution LabProblem hypothesis → smallest paid or observable test
- Recruiting conceptSeparate client-facing workflow from my own job-search lane
- Opportunity R&DMarket gaps, constraints, and solution feasibility
- Venture testsAdvance only when evidence justifies the next step
Decision & Execution Flow
Architecture Principles
| Principle | Design choice | Why it matters |
|---|---|---|
| Decision rights | Human authority is explicit rather than implied | Automation can move quickly without silently taking ownership of consequential decisions |
| Right tool for the work | Use deterministic workflows for known paths and agentic judgment only where interpretation adds value | Reduces unnecessary model risk, cost, and complexity |
| Shared state | Retrieval and write-back connect lanes to a common knowledge layer | Work compounds instead of restarting from zero in each tool or conversation |
| Observable completion | Outputs become records, artifacts, checks, or published proof | Makes “done” inspectable and creates material for the next iteration |
Drill Down
Lane Handoffs
| From | To | What crosses the boundary |
|---|---|---|
| Knowledge & orchestration | Mortgage operations | Enriched records, prioritized cues, retrieved context, daily outputs |
| Mortgage operations | Career & proof | Completed workflows, operating results, architecture evidence |
| Peregrine solution lab | Career & proof | Test results, prototypes, consulting evidence, lessons |
| Career & proof | Knowledge & orchestration | Market feedback, application learning, positioning signals |
| System owner | All lanes | Priority, risk tolerance, spending, commitments, merge and release authority |