TRIBE v2 Long-Form Brain-Response Breakdown
Cortical-response model output for How I Turned Pi Into the Ultimate Coding Agent, a 21:26 YouTube video processed on the RTX 5090 rig.
TRIBE v2 predicts cortical-response maps from video, audio, and language features. The timeline is useful for comparing where the modeled response rises, falls, or shifts. It is not a virality score, preference score, clinical interpretation, or proof that a specific edit caused retention behavior.
The supplied YouTube Studio screenshot shows an average view duration of 6:24, an average percentage viewed of 29.9%, and 62% of viewers still watching around 0:30, which Studio labels below typical. The screenshot is contextual evidence, not a machine-readable retention series, so this report does not compute a numerical correlation coefficient yet.
Timeline
The first panel shows overall cortical-response magnitude and its 95th-percentile envelope. The second shows second-to-second map change. The third compares left and right hemisphere magnitude. Vertical guides mark the strongest response-magnitude moments.
Sampled Cortical Maps
Twelve evenly spaced cortical maps summarize how the predicted surface response develops across the video.
Strongest Response Moments
These are the largest per-second mean absolute cortical-response values. Nearby transcript words are editing landmarks, not causal explanations.
| Time | Mean absolute response | Nearby transcript words |
|---|---|---|
0:31 | 0.2045 | Pi coding agent. I know you've probably heard of a lot of different coding agents and seen plenty of videos |
0:37 | 0.1921 | me and hear me out on this one. This one is different. The way it works is not like any of the other |
0:06 | 0.1523 | impressions video where I talked about the SDK a ton, some brief thoughts on the |
10:34 | 0.1514 | It's fully cleared. It all just kind of works. I can do slash resume whenever I want to, to go back to an existing |
6:18 | 0.1505 | our CI either in GitHub Actions or in some solution that tries to fix GitHub Actions. But the |
11:13 | 0.1435 | is fully open source. I will link it down below. If you want to go clone it onto your machine, you can have the exact |
10:16 | 0.1427 | the last message and as context, so I can paste into a new session to clear out the context |
6:22 | 0.1413 | Actions. But the problem is GitHub Actions is fundamentally broken. And I think at this point it is much better to |
Largest Map Transitions
These seconds have the largest change in the predicted cortical surface map versus the prior second. They are useful places to inspect cuts, topic changes, visual reveals, and pacing shifts.
| Time | Change RMS | Nearby transcript words |
|---|---|---|
5:00 | 0.1357 | That's it. I now have a fully custom extension added to my PI instance. That's literally all |
8:20 | 0.1099 | for my terminal. This is a project that was released a couple months ago, I'm pretty sure. I was a |
6:40 | 0.1078 | which means that it is fully compatible with GitHub Actions syntax. The only thing you need to do to migrate |
0:28 | 0.0938 | with GPT 5.5 on low reasoning and the Pi coding agent. I know |
0:01 | 0.0926 | A couple of weeks ago, I put out a video about Pi. This was kind of a first |
21:26 | 0.0895 | And until next time, enjoy your new pie. |
11:40 | 0.0838 | to hyper customize your pie into exactly what you want it to be. Mine should |
10:29 | 0.0829 | custom extension that I made. We'll talk about that in a second. And then I can just do slash new, make a new session. |
Aggregate Diagnostics
| Metric | Value | Meaning |
|---|---|---|
| Mean absolute response, average | 0.0874 | Average magnitude across time and cortical vertices. |
| Mean absolute response, maximum | 0.2045 | Largest one-second response-magnitude moment. |
| Change RMS, average | 0.0301 | Typical second-to-second shift in the cortical surface map. |
| Change RMS, maximum | 0.1357 | Largest second-to-second map transition. |
| Left hemisphere magnitude, average | 0.0842 | Average model-output magnitude on left-hemisphere vertices. |
| Right hemisphere magnitude, average | 0.0906 | Average model-output magnitude on right-hemisphere vertices. |
Run Details
- Source
- Public YouTube watch page
- Input SHA-256
fec1c4506028d2be99a62026c6a7c56069c0ef2ae709065fc919e0a3a5609a6c- GPU
- NVIDIA GeForce RTX 5090
- CUDA
- PyTorch 2.7.0+cu128 · compiled CUDA 12.8 · capability [12, 0]
- Numerical sanity
- Finite predictions
float32from -1.0531 to 0.9615 - Events
{"Audio": 21, "Sentence": 341, "Text": 1, "Video": 21, "Word": 5181}- Boundary handling
- Report-facing outputs are constrained to the original 1286.27-second MP4 duration. Preserved raw artifacts retain the untrimmed model output for audit.
The released TRIBE v2 source and weights are licensed under CC-BY-NC-4.0. Keep this analysis in the research sandbox unless commercial permission or an acceptable legal path is confirmed.