Solo Founder Operating System

SFOS is how I turned solo product work into an automated build-and-quality loop.

SFOS is the operating system around WittyKeys: AI-orchestrated development, automated UI/functionality work, QA regression, evals, observability, and release confidence.

AI-orchestrated dev Golden screenshots Espresso / UI Automator LLM-as-judge JourneyTracer
PlanCowork briefs, constraints, review gates
BuildClaude Code, worktrees, slash commands
CaptureReal device screenshots, logs, traces
CompareGolden baselines, visual drift, UX defects
TestEspresso, UI Automator, release checks
ShipHuman approval, Play Store discipline
SFOS
RoleProduct owner, architect, reviewer, and human approval gate.
AutomationCowork planning, Claude Code implementation, slash commands, and worktrees.
QAGolden screenshots, PixelDiffComparator, Espresso/UI Automator, and real-device checks.
WhySolo AI product work needs an operating system, not scattered prompts.
The problem

Building with AI is powerful, but only if the system keeps quality visible.

WittyKeys was not just an app build. It became a full operating loop: plan with Cowork, implement with Claude Code, deploy to a real Android device, capture screenshots, compare against baselines, run functional tests, score AI quality, inspect logs, and decide what ships.

That is what SFOS means in my work: the structure that lets a solo founder move fast without losing control of quality.

The loop

Automate development, testing, and learning without removing judgment.

  • Cowork creates phased plans and Claude Code instructions; Claude Code implements, builds, deploys, captures, and reports.
  • Parallel frontend and backend worktrees let independent AI sessions work without stepping on each other.
  • Golden screenshot regression catches UI drift through device captures and PixelDiffComparator baselines.
  • Espresso/UI Automator flows cover functionality across onboarding, keyboard, overlay, chat, tone, grammar, translation, and AI actions.
  • LLM-as-judge scoring reviews reply quality, tone fit, Hinglish naturalness, and context relevance.
  • JourneyTracer and backend logs make product behavior easier to investigate after release.
Development systemPlanning, implementation, reporting, and review happen through repeatable AI-agent workflows.
Testing systemUI and functionality are checked with goldens, E2E tests, real-device runs, and quality scoring.
Release systemLogs, traces, crash/performance data, and review gates keep shipping decisions grounded.
Takeaway

This is the operating layer I want to keep evolving.

I can build the product surface, but I also built the system around it: AI-assisted development, automated UI/functionality work, QA regression, evals, logs, release gates, and the discipline to keep a fast solo workflow honest.

Back to portfolio