So... The system wrote the app. Let's see how it did.

To start with, OpenCode wrote the app "blindly".It never ran it to verify that it works, so all the work was done "mentally". (Oooo, mental!)

It has created the directory yt-watcher and the files. But, we have to create the virtual environment, download dependencies, and then run app.py. According to the instructions it gave, that is.
So:

nobait@Dev-Station:~$ cd yt-watcher/

nobait@Dev-Station:~/yt-watcher$ ls

app.py requirements.txt static templates

nobait@Dev-Station:~/yt-watcher$ python3 -m venv venv

nobait@Dev-Station:~/yt-watcher$ source venv/bin/activate

(venv) nobait@Dev-Station:~/yt-watcher$ pip install -r requirements.txt

Successfully installed Flask-3.1.3 blinker-1.9.0 certifi-2026.4.22 charset_normalizer-3.4.7 click-8.4.0 idna-3.15 itsdangerous-2.2.0 jinja2-3.1.6 markupsafe-3.0.3 requests-2.34.2 urllib3-2.7.0 werkzeug-3.1.8 yt-dlp-2026.3.17

(venv) nobait@Dev-Station:~/yt-watcher$ python app.py

* Serving Flask app 'app'

* Debug mode: on

WARNING: This is a development server. Do not use it in a production deployment. Use a production WSGI server instead.

* Running on all addresses (0.0.0.0)

* Running on http://127.0.0.1:5000

* Running on http://172.17.212.246:5000

Press CTRL+C to quit

* Restarting with stat

* Debugger is active!

* Debugger PIN: 965-698-609

The app runs—this is a good first sign.

We can connect to the app, and we see the web page.

It works for a few seconds, but after that, it produces an error.

The backend has this log:

172.17.212.79 - - [18/May/2026 08:34:26] "GET / HTTP/1.1" 200 -

172.17.212.79 - - [18/May/2026 08:34:27] "GET /static/style.css HTTP/1.1" 304 -

[youtube] Extracting URL: https://youtu.be/SaiyMBwW0Hc

[youtube] SaiyMBwW0Hc: Downloading webpage

WARNING: [youtube] No supported JavaScript runtime could be found. Only deno is enabled by default; to use another runtime add --js-runtimes RUNTIME[:PATH] to your command/config. YouTube extraction without a JS runtime has been deprecated, and some formats may be missing. See https://github.com/yt-dlp/yt-dlp/wiki/EJS for details on installing one

[youtube] SaiyMBwW0Hc: Downloading android vr player API JSON

[info] SaiyMBwW0Hc: Downloading 1 format(s): 401+251

[download] Destination: /home/nobait/yt-watcher/uploads/audio_f09b68dbb578454589c75825642826f7.wav.f401.mp4

[download] 100% of 282.16MiB in 00:00:14 at 19.59MiB/s

[download] Destination: /home/nobait/yt-watcher/uploads/audio_f09b68dbb578454589c75825642826f7.wav.f251.webm

[download] 100% of 8.93MiB in 00:00:00 at 17.79MiB/s

[Merger] Merging formats into "/home/nobait/yt-watcher/uploads/audio_f09b68dbb578454589c75825642826f7.wav.webm"

Deleting original file /home/nobait/yt-watcher/uploads/audio_f09b68dbb578454589c75825642826f7.wav.f251.webm (pass -k to keep)

Deleting original file /home/nobait/yt-watcher/uploads/audio_f09b68dbb578454589c75825642826f7.wav.f401.mp4 (pass -k to keep)

172.17.212.79 - - [18/May/2026 08:35:29] "POST /summarize HTTP/1.1" 500 -

That indicates that the transcription phase was completed, but the summarization could not start.

Now, we could of course do a bit of debugging ourselves, BUT this is a test, and we want the robot to work alone...

Or I could restart from scratch using another model to see if it would handle the request better, or I could start from scratch with a prompt that would also ask the system to verify operation before declaring success.

Instead, I thought about informing OpenCode about the situation and asking it to fix it.

OpenCode went back to the drawing board and used my RTX 5090 fully for a few minutes.

(For the record, I have capped the power on the RTX to 550W instead of the default 600W. The loss of performance is almost unmeasurable, but this way my air-cooled system works at a max of 85ºC even for hours.)

A couple of minutes later, the system identified the issue and corrected it: