Signature Ideas

Last updated: 2026-07-28

Ideas worth investing in. These are the themes that recur across multiple documents and are unique enough to define a position.

---

1. ⭐ The Background Agent Pattern

Core thesis: Engineers can shift from synchronous pairing to async AI agent partnerships — where agents autonomously execute well-defined engineering tasks from ticket to PR while humans review, merge, and plan the next cycle.

Evidence cluster:

Vibe code like a PRO — original vision: Linear→background agent→PR pipeline

How to Background Agents — practical Cursor-based tooling pattern

How I Built a Full Booking System... — real-world validation: Devin as async engineering partner

Graduated Trust — the trust/gov layer this pattern requires

State machines to visualize agents — formal modeling of agent state transitions

Why this is signature: Five distinct angles exploring the same core idea from different perspectives. The cluster is uniquely comprehensive — no other creator has stitched together vision + tooling + validation + trust + formal modeling under one roof.

Compounds into: Guide → CLI tool → open-source project → workshop → community

---

2. Structured Productivity for Indie Builders

Core thesis: Side projects and deep work are enabled by explicit weekly planning rhythms, not by grinding.

Evidence:

How I organize my week on sundays — the planning system

The Devin booking system — built as a side project while working full-time, enabled by the above system

Why this is signature: It connects productivity philosophy to concrete engineering output with real proof.

---

3. Graduated Trust for AI Code

Core thesis: As AI generates more production code, we need a social-graph-based trust system (not a technical solution) to verify and accredit that code.

Evidence:

Graduated Trust essay

Devin case study — 6 merged PRs in a month demonstrates the urgency of the trust problem

Why this is signature: Counter-consensus take. Most AI trust solutions are technical (watermarks, proofs). A social/web-of-trust approach is differentiated.

---

Ideas Under Consideration

State machines as a UI for agent workflows — Could formal models become a user interface, not just a developer tool?

The indie founder's AI toolchain — Synthesizing the background agent pattern + weekly planning into a stack recommendation

---

Blind Spots

Testing/QA as a service — If agents write code, who owns testing? This is underexplored in the current cluster.

Cost economics — What does running Devin/Cursor agents cost at scale? Real numbers would strengthen the argument.

Team dynamics — How does async agent work change team communication patterns? Currently focused on individual engineer experience.

---

Signature Ideas is a living document. New evidence is appended; previous analysis is preserved.