# Clean Slate Focus

> Agent-facing counterpart to the [human project page](/projects/clean-slate-focus/).

## Record metadata

- Record: 014
- Slug: clean-slate-focus
- Domain: Software
- Domain code: SW
- Type: iOS app
- Status: Shipped
- Period: 2025
- Portfolio role: Early work
- Publication state: Historical case study
- Case-study readiness: Draftable
- Compendium edition: 0.4

## Summary

An early Pomodoro-style iOS app built before coding agents became part of the workflow.

## Overview

Clean Slate Focus belongs to an earlier moment in AI-assisted programming, before an agent could inspect a project, edit several files, run the build, and respond to its own errors. The app was assembled through a much more literal conversation between ChatGPT and Xcode.

A small iOS focus timer built around a stripped-down Pomodoro workflow.

Purpose: Create a minimal personal focus timer and learn how to get an app from Xcode to a working phone.

## The problem behind the project

Josiah wanted a clean focus tool and, more importantly, a concrete way to learn the app-building and shipping process.

Josiah wanted a minimal Pomodoro-style timer, but the larger objective was learning how an iPhone application actually moves from an idea to a working device. The narrow feature set made that path visible without requiring a backend or complicated product model.

It primarily served Josiah as both a focus aid and a learning project. No broad user adoption is claimed.

## How it took shape

Swift and Xcode through a pre-agent workflow: copying code and compiler errors between ChatGPT and Xcode on Josiah's father's computer, then testing the result manually.

Working on his father's computer, Josiah copied proposed code into Xcode, copied compiler errors back into ChatGPT, interpreted the response, and repeated the cycle. That awkward loop still required product decisions, debugging judgment, manual testing, and persistence.

Josiah chose the product, assembled and debugged it interaction by interaction, made the design decisions, and completed the app despite not yet having an integrated coding agent.

A working app was completed as Josiah's second iOS application, though it did not develop a meaningful user or sales base.

## What the project means now

The app did not acquire users or revenue, and that is not the point of preserving it. It documents an early stage of Josiah's development practice and shows how much of the modern agent workflow was once performed manually across the boundary between two separate tools.

The surviving source and current compatibility need review, and the implementation reflects an early learning-stage process.

Even an inefficient copy-and-paste workflow can teach the full feedback loop of building, debugging, and finishing a real application.

Preserve representative screenshots and source history as an honest early-work record rather than modernizing away its context.

## Publication and interpretation notes

- Current classification: Shipped
- Portfolio readiness: Draftable
- Publication boundary: Historical case study

## Additional agent context

The primitive development workflow is part of the story. Do not retrofit modern agent-led implementation claims onto this early project.

## Related project records

- [Murphy’s CountDown](/projects/murphys-countdown/llm/) — A small iOS utility that turns end-of-shift register counting into exact instructions.

## Navigation

- [Complete project index](/projects/llm/)
- [Human version of this record](/projects/clean-slate-focus/)
- [About Josiah's working method](/about/llm/)
- [Agent discovery map](/llms.txt)