I Built an App That Turns My Thoughts Into Posts

Introduction

This post was recorded, not typed. A few hours ago I built a small app called Just Thoughts: it records what I say, transcribes it locally with Whisper, and then hands it to an AI agent that turns it into either a blog article or a thought summary. Here is why I built it, how it came together, and what I plan to do with it.

Summary

• Just Thoughts is a minimalistic app I built to record, transcribe and process spoken thoughts, and it is public on my GitHub.
• It solves my scattered note-taking across Apple Notes, Obsidian, Word, notebooks and stray A4 pages.
• It also removes the friction and perfectionism that stop me writing blog articles consistently.
• Transcription runs locally with OpenAI's Whisper model, which is fast and very accurate.
• Two functions exist: convert a recording into a blog post, or into a markdown thought summary.
• Over time, a compilation of summaries could reveal patterns in my thinking and rescue forgotten ideas.

Hello everyone. Today I'm recording this conversation in an app that I created myself just a few hours ago. The name of this app is Just Thoughts, because it's just about thoughts: saving thoughts and converting thoughts into blog posts or thought summaries. It is available publicly in my GitHub repo. In this recording I'm going to talk a bit about the development of this app, what made me develop it, and how I'm going to use it in the future.

Why I built it

As a PhD student, and as a very curious person in general, I know how many thoughts come across my brain and how many things I process throughout the day. It can be one topic, it can be several topics, it can be thousands of different topics a day that I process in my mind. What I usually do today is take notes in Apple Notes, sometimes in Obsidian, sometimes even in Microsoft Word. I take a lot of notes here and there, and I also use physical notebooks. I sometimes write notes on plain A4 paper that I usually put in the bin at the end, because they somehow pile up on my desk.

For some time I'd been thinking about the idea of an app that could transcribe what I say, just like this, save it somewhere, and then process it using AI to summarise it. That was the first requirement I identified: transcribing my thoughts and saving them as text files for future reference.

The second requirement came from my own website, which you are using now to read this blog article. I sometimes write blog articles out of nowhere. Something just comes to mind and I decide to write an article. You will see quite a few articles on my website, which is the new version of my personal website, because I had a different version before that I deleted permanently. The problem is that I'm not always motivated to write articles. It takes time, it needs preparation, and sometimes I get a bit too perfectionist about it. All these things stop me writing blog articles consistently, even though a lot of ideas come to my mind all the time. That's why I wanted to automate the process. Now I record my thoughts and turn them into a blog article using AI.

The third reason is simply curiosity. With AI and vibe coding, everything is very convenient, and it gives me so much pleasure to build things as a vibe coder.

How I built it

It was around 4pm today when I decided to build this app using Claude Code. I opened Claude Code and said I needed a very minimalistic app for this purpose, then explained the purpose. I wanted to build it using Apple's UI concepts, because Apple's human interface guidelines are very solid and have proved successful for billions of people. I wanted to stick to that, so I uploaded a screenshot of my Voice Memos app and asked Claude to take inspiration from it. On top of that, I found an Apple design skill for Claude Code online and asked Claude to install that skill as well. With those concepts and my guidance, Claude created a very nice interface on the first attempt, and then I followed up with more and more suggestions. That's how we built this app.

Initially there was only one interface, where I could just talk like this and it would save my thoughts as text. I'm using a locally run speech-to-text model called Whisper, from OpenAI. It runs locally and it is extremely fast: as soon as I finish recording, it transcribes my voice really accurately and saves it as a text file, which is then shown on screen. That was the first function.

(The transcript view — Whisper's raw output from a recorded thought, before any AI processing)

Then I added another function, where a Claude-based LLM agent converts my text into a blog post. Because you're reading this blog article, it has been generated by Claude from my voice input.

(The same recording, turned into a structured blog post draft with a title, summary and headings)

Thought summaries

Then I thought, I'm not going to write blog articles all the time. I might just need to summarise my thoughts for future reference. So I created the second function of the app: thought summaries. When I record my thoughts, which can be just talking to myself, whatever it is, it will be saved as text as before, and a different Claude agent will summarise it and save it as a markdown file.

(The same recording again, this time distilled into highlights and themed sections instead of a blog post)

You already know the use case for the blog post, because you're reading it now. As for the thought summaries, it would be nice to keep my thoughts as summaries over a long period. If I create a compilation of my thoughts, recording them every day or whenever possible, I might be able to find patterns in my thinking. I also usually forget very innovative ideas that come to my mind, but if my thoughts are documented I can always refer back to them and say, I had this idea back then, on this day, and maybe bring it back onto the desk and improve it. There may be more use cases in the future, but for now that is the use case.

Final thoughts

I believe this is a very innovative way to record my thoughts and document them, both for my own reference and for anyone who is interested, because knowledge is knowledge. Everyone has something to learn every day, every minute and every second. Even though AI-based tools have started putting a lot of text onto the internet, these thoughts will not be too much for the internet.

That's it for today. Hopefully I will make more content in the future. Thank you.

The app is open source, you can find the code on GitHub.

Most powerful idea: By pairing local Whisper transcription with AI agents that either publish or archive, the app removes the friction between having a thought and keeping it, turning everyday speech into a searchable record of your own thinking.