I’ve been working on porting the app from Windows to Mac, and initial testing on Apple Silicon shows noticeably better CPU performance across the board, which is promising. The app’s Metal support is in place, but the CUDA parts are still outdated and need updating. I reorganized the hub events system to improve serialized communication, which should help with some of the event ordering and content ID issues I’ve been seeing.

The app officially moved to version 0.5, but I’m still figuring out how to position the Pro version better for Mac users since the current setup feels off. Testing the packaged app on both Windows and Mac hasn’t happened yet, so I don’t know if there are other issues beyond the transcription problem I fixed—turns out overlapping watch and output folders were breaking transcription, and that’s resolved now.

There are still several medium-priority bugs lingering, like Metal Performance Shaders running slow on the Mac Mini, transcription resummarizing the same file repeatedly, WAV file timestamp problems causing repeated summaries, and some serialization/order problems with hub events and content IDs. These are frustrating because they affect reliability, and I haven’t nailed down fixes for all of them yet.

On the customization front, I implemented creation, import, and export of custom user modes through the mode builder, and set up a GitHub repo to manage these modes, releasing version 1.0 there. The website now has responsive starter mode ideas and a built-in GitHub issue tracker to submit custom modes. I added a toggle switch to flip between light and full modes in the custom user mode section, but I’m still unclear on what exactly those modes should mean, so that’s up for discussion on GitHub and the site. I also moved the contribute button and import/export controls inline into the user dropdown menu to clean up the UI.

Licensing has been a big focus. I moved expiration dates out of plain text storage into a more secure method and started building webhooks for license and cryptography management. I’m planning to use Netlify as the backend for license distribution, but connecting reliably has been tricky. I want to implement a defense mechanism so the custom mode tab won’t populate unless there’s a valid license, but I’m still figuring out how to handle users who downgrade—should they keep previously installed modes or lose access? That’s an open question.

I found a newer Torch version optimized for Apple Silicon and am setting up a virtual environment to support Torch 2.10. Testing is ongoing to make sure it doesn’t break anything, but there’s some concern about stability with this update.

Some decisions are settled: the app is at version 0.5, the license structure will be secured, the defense mechanism for license validation will be implemented, and the backend restructuring will be tested soon on Apple Silicon. I’m hesitant about adding a helper in the mode builder panel because I’m not sure if it’ll help or just clutter things. I’m also considering splitting the large configuration file into smaller parts and adding a simplified mode builder page for Nucleate Labs, plus an alert system to warn about performance impacts from config file size.

Still on the to-do list is thorough testing of packaging on Mac, Windows, and Apple hardware, and maybe running a pre-release beta test, but I haven’t committed to that yet. There are a lot of open questions around licensing, mode management, performance, and UI that I need to resolve before moving forward. Overall, progress is steady but there’s a fair bit of uncertainty and some technical debt to clean up.

Two-Line Takeaway

Key step forward

The most important forward progress step is the successful porting of the app to Mac with improved CPU performance on Apple Silicon and the implementation of Metal support.

Significant barrier or risk

The most important barrier or risk is the unresolved medium-priority bugs affecting reliability, ongoing licensing and mode management uncertainties, and the need for thorough testing on multiple platforms before moving forward.

Generated with Nucleate by Watchlight Studio — distilled from full transcription.