# True, Then Beautiful - an AlephBrain Field Note

> Five passes at one brain-scan file. Getting it true was the easy part; getting it beautiful was the work. One picture skipped the truth and was lovelier for it. A Beyond Reason field note by Eric Porres on decoding a living-brain transcranial-ultrasound file with ChatGPT and Claude.

On June 25, 2026 the brain-interface lab Aleph posted a raw 2.4 MB binary of microbubble tracks from transcranial ultrasound localization microscopy (ULM) of a living human brain, with one line: "claude should be able to figure out how to render them." ChatGPT and Claude both took the dare from the identical one-sentence prompt. Every render that actually parsed the bytes converged on the same flow field; one image-generation pass painted a gorgeous vascular tree connected to none of the data. The piece is the five passes side by side, with interactive 3D viewers.

Full text for LLMs: https://porres.com/alephbrain/llms-full.txt

## The field note

- [True, Then Beautiful (article)](https://porres.com/alephbrain/): The full field note - the truthy-vs-true distinction, the five passes, and what the bytes say.
- [Full text for LLMs](https://porres.com/alephbrain/llms-full.txt): The entire field note as plain text, plus the decode facts and the interactive-viewer index.

## Interactive renders

All five decode the same 2.4 MB microbubble file (ULM; 6,211 tracks, 690,254 localizations). Open any viewer to rotate, filter, and scrub time.

- [ChatGPT GPT-5.5 High - v1](https://porres.com/alephbrain/gpt-v1.html): First contact - depth-colored projections and a browser viewer; the decode is roughly right.
- [Claude Opus 4.8 Max - v1](https://porres.com/alephbrain/claude-v1.html): The honest blob - the same decode as a raw point cloud, a comet because every track starts at one origin.
- [Claude Opus 4.8 Max - v2](https://porres.com/alephbrain/claude-v2.html): Converged - rebuilt as a line viewer with track lines, opacity, length filter, and projections.
- [ChatGPT GPT-5.5 Pro - v2](https://porres.com/alephbrain/gpt-v2.html): The finest render of the five - a real line viewer with a time sweep that flows the bubbles through the vessels.
- [Claude Opus 4.8 Max - v3](https://porres.com/alephbrain/claude-v3.html): Time, learned from Pro - each point carries its acquisition frame, so a moving "now" trails a comet tail behind every bubble.

## What the bytes say

- Format: ULMT v6 (undocumented binary, reverse-engineered to the byte).
- Tracks: 6,211. Localizations: 690,254. Bounds: 54.3 x 19.6 x 29.4 mm. Max speed: 0.829 mm/frame.
- Structure: a 64-byte header, a 16-byte track table, and four quantized 16-bit point streams; position is delta-encoded (a track is the running sum of its steps).
- The missing piece: the spatial origin. The file records how each microbubble moved, not where in the skull it moved. The one field that could place each track turned out to be a frame counter, not a coordinate.
- Models: Claude Opus 4.8 Max; ChatGPT GPT-5.5 High and Pro.

## Sources

- [Field note on Substack](https://promptedbyeric.substack.com/p/true-then-beautiful-ai-and-the-brain): True, Then Beautiful. Beyond Reason, June 2026.
- [Aleph's original dare on X](https://x.com/alephneuro/status/2070241600853458961)
- [The data file (tracks_v5_v6.bin.gz)](https://alephneuro.com/data/tracks_v5_v6.bin.gz)
- [Aleph (alephneuro.com)](https://alephneuro.com/)

## About

- [Eric Porres](https://porres.com): Chief AI Officer at Logitech; author of Beyond Reason. Contact: eric@porres.com.
- [Beyond Reason](https://promptedbyeric.substack.com): Eric's Substack on AI and the economy for practitioners.
- This field note is one property on porres.com; site index at https://porres.com/llms.txt.