Wiki

Compiled knowledge base. Synthesized from raw sources, interlinked.

Topics

Concepts

Agent mistake → fix the harness, not the prompt

The single most important principle from harness engineering: when an AI agent makes a mistake, don't fix the prompt or try a better model. Fix the environment.

Agent self-discipline — drift detector and complexity thresholds

Principles for fighting AI agent degradation. The agent tends to drift into comfortable but unproductive patterns. Two mechanisms fight this:

CLI-first testing — every project gets a CLI mirror

Every project should have a CLI utility that mirrors the core business logic without UI. Not a separate tool — the same modules, different entry point.

Context engineering — context as code

One of three harness engineering components. The idea: agent context is not chat history — it's a continuously improved knowledge base in the repository.

Decision traces compound — each agent action improves future ones

The economic argument for context graphs: agent decision traces are a compounding asset. Each successful action recorded in the graph improves future ones th...

SGR — schemas first, logic second, UI last

Schema-Guided Reasoning (SGR): all projects built from schemas to logic, not the other way around. Schemas are the contract between AI agent, business logic,...

Kill / Iterate / Scale — weekly decision rules

Hard numeric thresholds for weekly product decisions. No "let's wait and see":

One pain → one feature → launch

The antidote to "platform thinking": don't build a platform, don't draw a 50-feature roadmap. One pain, one solution, shipped in days.

Portfolio approach — multiple small bets, not one big bet

Build multiple small products instead of going all-in on one. Pieter Levels: 70+ projects, 3-4 active revenue sources. John Rush: 20+ products. Danny Postma:...

Privacy is architecture, not a feature

Key distinction from the manifesto: privacy is not a settings checkbox but an architectural decision made before the first line of code.

S.E.E.D. — niche scoring in 10 minutes

Quick filter for niche ideas before writing code. 4 dimensions:

Antifragile life design: barbell across domains

Taleb's antifragility applied not to investments but to life design across 5 domains: income, health, relationships, skills, location. Three levels per domain:

5-step decision framework: from epistemics to action

Every decision passes through 5 steps, each with its own checklist:

Ikigai as 8-9 daily micro-joys, not a mission

The Western interpretation of ikigai (intersection of "what I love / what I'm good at / what pays / what the world needs") is a distortion of the original. A...

Optimizer syndrome: best choice ≠ more happiness

Barry Schwartz's research ("Paradox of Choice"): maximizers (those who seek the best option) objectively find better outcomes but experience less happiness a...

STREAM: principles as hierarchy, not a flat list

The key difference between STREAM and typical principle collections: it's not a flat list but a 6-layer hierarchy where each layer builds on the previous one.

Source Summaries

Apple Foundation Models — zero-cost on-device AI for iOS portfolio

Unified on-device AI layer for all iOS products (KissMyTask, FaceAlarm, VoiceJournal, QuestSchool). Zero API cost, privacy by architecture, SGR natively via ...

Context Graphs — agent trajectories and organizational memory

Context Graph — a graph capturing AI agent decision-making trajectories: not just *what* the agent did, but *how* and *why*. "A system of record for decision...

Universal Development Principles — every project, any stack

993-line reference document applied to every project regardless of stack. Injected into PRDs during generation. Covers: SOLID, DRY (but Rule of Three), KISS,...

Harness Engineering — development in the age of agents

Harness Engineering — designing environments where AI agents do reliable work. Humans steer, agents execute. Synthesized from three sources: OpenAI experimen...

Infrastructure: two tools, serverless by default

Unified infrastructure for multiple mini-startups. Two principles: one project = one deploy config (infra lives in the repo), two tools, no more (SST + Pulumi).

KMP sync — Supabase + PowerSync for offline-first cross-platform

For family/cross-platform apps needing offline-first sync between Android + iOS: Supabase + PowerSync is the recommended stack. MongoDB Realm Sync was deprec...

PostHog analytics: one product = one project, cross-platform identity

Unified analytics for solo founder with mobile apps + web landings. Core problem: one person on landing and in app = 2 anonymous IDs. Broken attribution, bro...

Background jobs — cron → CF Workers → Prefect → Temporal → Trigger.dev

5 tools for background jobs, ordered by complexity. Rule: start with cron, move up only when you need to.

Agent sandboxing — safe LLM agent execution on Mac

LLM agents (Claude Code, Codex, Gemini CLI) run with full filesystem access. Sandboxing via lightweight VMs solves this: agent works in isolation, sees only ...

Inspiration sources — curated channels, communities, newsletters

Curated list of ~60 channels/communities across 7 platforms, organized by relevance to solo founder workflow.

Launch Playbook — From Niche to Revenue

Operational playbook in 11 sections. Core: idea → live product with traffic in 5-9 days.

Solo Founder Manifesto — Core Values

The manifesto (86 lines) establishes 6 foundational positions:

RAG patterns — 7 approaches from vector to agentic

Reference catalog of 7 RAG patterns, each with when-to-use, trade-offs, and components.

Solopreneur success patterns — benchmarks, pricing, distribution

Data-driven patterns from successful indie hackers (2023-2026). Used by `/validate` for STREAM scoring and `/research` for market context.

CodeGraph — code intelligence across all projects via FalkorDB + tree-sitter

Knowledge graph of source code across all projects. Scans files, symbols (functions, classes, protocols), dependencies, imports, function calls, class inheri...

2026 validated app ideas — homeschool, family productivity, privacy-first

Five underserved market intersections where no competitor operates. Score: 8.1/10. 50+ evidence sources. All privacy-first, offline-first, on-device AI.

Summary: STREAM Framework — 6 Layers Decision System

STREAM = Scope (epistemic) · Time (temporal) · Route (action) · Exposure (stakes) · Audience (social) · Meaning (meta). Six layers arranged hierarchically — ...

SGR complete guide — code examples, patterns, integration with BAML

1855-line reference compiling Rinat Abdullin's Schema-Guided Reasoning (SGR) series. Core: use Pydantic/Zod schemas as the interface between LLM and code — t...