Origin
InPromptOut started life, rather appropriately, in an idea-notebook kept on my desk. The premise was simple: a tool that combined AI-generated prompts with journaling and visual inspiration – not to automate creativity, but to make space for reflection and the slow shaping of ideas.
Core Question
The project grew out of frustration with existing creative AI tools. Most pushed for volume and rapid results. I kept returning to one question: why did every creative AI tool optimise for output, when the harder problem was first figuring out what to make and what shape it should take?
I wanted a quiet space for thinking, where prompts could provoke ideas rather than dictate them, and where the user stayed clearly in control. Chat interfaces, in particular, felt like they flattened creative work into disposable threads.
Early Vision
I wrote down the main pain points I was trying to solve:
- Blank page syndrome – not knowing what to create
- Ideas scattered across notes and tools with no easy way to reconnect them
- Useful musings forgotten and buried somewhere in an AI chat thread
- Lack of visual anchors tied to prompts or thoughts
- Difficulty reflecting on creative growth or reusing older ideas
MVP Scope
To keep momentum, I defined a narrow MVP focused on validating the core concept:
- Prompt generation via OpenAI (text only)
- Simple inspiration board with tags and timestamps
- Linked journal entries
- Basic prompt preview with image overlay
- Filters and favourites
- Image upload support
- Mock mode to avoid burning API credits during development
Tech Stack Considerations
The stack was still undecided. I considered:
- Backend: FastAPI (for async support and good documentation)
- Frontend: React + Vite or a lighter HTMX/Alpine approach
- Database: Firebase Firestore or SQLite for local development
- AI: OpenAI GPT-4o for text (DALL·E optional)
What was clear from the start: keep the foundations on tools I already trusted – pytest, loguru, mock mode for API safety – while using the project as a space to try newer AI-era practices for the first time. CodeRabbit for PR reviews was one. The build was as much about feeling out modern AI-assisted development workflows as it was about shipping the tool itself.
A learning thesis
The plan was to use AI unevenly across the stack: lightly on the Python backend (where I was already on familiar ground) and more heavily as a co-programmer on the React frontend. Part of the thesis was also personal. I was using the framework for the first time and wanted to know whether I could learn a new tool faster with AI alongside me than I'd managed on my own in the past.
Intended User
I wasn't trying to build for everyone. The target was creatives who wanted gentle, intentional AI support – writers, artists, educators and anyone who valued reflective thinking over high-volume output.
What This Phase Established
By the end of June 2025 I had a clear enough direction to start building. The philosophical framing – AI as companion, user in control, reflection over production – would be tested heavily once implementation began. Whether those ideas would survive contact with real code remained to be seen.