Back to How this system works

AI use log

The AI work logged here used Claude Code, Anthropic's coding tool. The prompts below are copied as written, including speech-to-text mistakes from dictating them.

Gap in this log: The earliest prompts that created the native app were not available when this log was written. The first saved prompt is from October 19, 2026, and the first commit in the repository is from October 20, 2026.

Part 1: FreeFlow Web, this class project

2026-11-02

  1. I'm going to submit this repo as a project for my systems class. If you go onto my ISM class in canvas and find the System making project you can see all of the requirements. I need you to go into canvas, find the project, build all requirements for the project and put it into one place for it. It's going to ask for a dfd. Although you need to be right I don't think you need to put in a lot of detail for the dfd mainly just the high level overview of how the processes work together. Ask any questions and as many as you want before you actually start.
  2. Decisions given to Claude's questions: build a browser prototype instead of submitting the native Mac app; target lecture-heavy business students; use the microphone with a typed fallback; allow Claude to push to GitHub Pages, publish this log, and register the idea on the board.

What the AI did: Read the handout, compared it with the native app, and found that the native app could not meet the browser, public link, and no-paid-service requirements. It then wrote the requirements, business rules, screens, diagrams, and tests.

Decisions the student owns: The target user, the value model, moving from a native app to a browser prototype, replacing the Claude API with rule-based cleanup, and keeping paste and proofreading manual. The business rule limits for word count, similarity, and retention were adapted from the native app or chosen for this prototype, and the student must be able to explain each one.

Part 2: FreeFlow, the native Mac app that inspired it

2026-10-19

Result: Adjusted the dictation bar, researched and improved cleanup, fixed the double-tap Fn hands-free mode, and added customizable sounds.

  1. Right now the bar is like in the way. It's too high up. and I, I, make the, the area in which like it starts to hover, make it a little bit smaller because it's like picking up too much.
  2. The cleanup feature isn't that good. It's not, it's not like tracking. I want you to go up in research, whisper, and specifically look for, um, the cleanup feature, like how it works, how it works, and you can test it and whisper if you want. But what I'm saying is, like, I used it and it just did my raw transcript and actually clean it up, like a Whisper does. So do research on that, fix it up, and then let me know.
  3. Whenever I double tap FN, it doesn't um, auto, like, it doesn't hold the dictation.
  4. Double-tap Fn hands-free works now.
  5. Uh, no, double tap FN hands-free doesn't work. And also whenever I press FN, it pulls up the emoji, like, I don't know, it's not working. And also, whenever I cancel it, it doesn't automatically, like, whenever I cancel it and then I go back to dictation, it doesn't show up with the little bar showing I'm talking.
  6. I'm testing FN right now.
  7. I'm testing FN right now. Whenever I do the double tap FM, the emoji thing pops up, and I don't like that, and also, I won't, to confirm my FN or to confirm my hands-free. I want to be able to press FN. And if I want to cancel it, I press escape.
  8. Still off. Here is what happened:
  9. Testing, FN, testing.Um, Double tapping FN still doesn't work. I mean, it does the hand-free mode, but then it pulls up the emoji thing, which is really annoying. And whenever I do the double tap FN and it pulls up the emoji thing, it makes my makes the text thing that I'm focused on the emoji search bar, which is really annoying.
  10. Testing. Testing, testing, op the emoji thing did not show up. Okay, now, um, if I'm hands-free, I want to be able to press control to turn it into uh, the cleanup mode.
  11. Hands-free then Control switched to cleanup and the result was cleaned.
  12. Can you go and tell me if the style is working? Um, and if it's detecting where I'm at?
  13. There are any other places where you think this could be really useful?
  14. Add all of it.
  15. Also, what other, I don't really like the chime. I think it's like too loud. Um, what other things could we do?
  16. I want to be able to con, I want to be able to customize not only the but the sound whenever I click FM, but also whenever I switch modes and all that.
  17. Um, can what else do you think would be good for this?

2026-10-20

Result: Fixed speaker labels in meeting transcripts, added Lecture mode, kept recordings from opening the main window, and wrote the interactive tutorial.

  1. The meeting recorder seemed to be tagging things that they said as things that I said. Go look at my latest meeting and fix it for the future
  2. should I use the meeting notes for class
  3. yes add the lecture mode i should press option n for the shortcut and then switch modes using the control button to toggle it and make it "lecture mode"
  4. when recording a lecture/meeting it should not open the app. It should just pop up on the bottom pill. I'm currently recording a lecture right now so don't restart the app right now.
  5. if I open up the app while in a meeting though it should auto bring me to the meeting I'm recording. Also does the "my thoughts section" get added to the summary?
  6. lecture's done, install it now
  7. Create a tutorial to onboard new members and add this to module hub repo, make it public.

2026-10-22

Result: Fixed false "no text field" errors, transcript scrolling, the pill size, and a Command Mode error.

  1. I'm noticing a bug that it says that there is not text field focused even if it is. Fix that please. Also while I'm recording a lecture it is consistently bringing my screen to the transcritpion actually happening instead of being able to browse the transcript as i'm recording. However if I scroll to the bottom, I want to be able to follow the transcript as it is being made Don't restart the app yet I'm recording
  2. I've stopped recording. Install the new build and relaunch FreeFlow.
  3. Commit the focus, transcript scroll and pill changes.
  4. Merge fix/focus-scroll-pill into main locally.
  5. I'm getting this error when trying to do command mode
  6. I'm not recording. Install the fix, relaunch FreeFlow and commit it to main.

2026-10-27

Result: Prepared the repository to be shared publicly with private files removed.

  1. Make it so that I can share this um, on GitHub with my friends, but obviously remove any like, uh, anything important, any environmental variables and stuff. So just make it a public repo that I can send.

2026-11-02

Result: Fixed a freeze when the microphone hands over to AirPods, and made Fn always stop recording.

  1. I've noticed there's a bug with their software. If I switch mic inputs, like I use AirPods, and I'm on my phone with my AirPods connected to that, and then I press FN to start Wispr, it realizes that it needs to use my AirPods and freezes my Wispr or my FreeFlow until it stops working, and I have to quit FreeFlow and relog it. I don't know if you can see the logs, but it's in there. Worst case scenario, if something like that happens, I just want the ability to press FN and for it to stop recording, even if it completely freezes. Can you fix that? Thank you.