Pinned Tweet
A fun simple clip explaining how the Geometric Language Machine reasons.
- an example of how it understands how water becomes steam.
💠 github.com/DigitalEuan/GLM 💠
Next step after Lean4 is human checked (preferably x 3)
Some lean files could take a person a whole lifetime to be able to confidentiality understand and check
Some may be past our capacity to understand fully or maybe at all 🫠
DigitalEuan retweeted
The largest math proof in human history is now computer-verified.
Today, Axiom announces a complete Lean auto-formalization of the Classification of Finite Simple Groups, the "periodic table of symmetry."
About 15,000 pages, built by over 100 mathematicians across generations.
Consolidated GLM research document: academia.edu/179675455/The_G…
162 lean verified study documents squished into one master document then further condensed into a smaller companion version again.
Some interesting stuff in there, AI development, computational whatnot and so on
enjoy
Investigation into Wolfram's Rule 30 puzzle is a cool challenge and helps push the GLM development also.
- up to 10/50 ARC-AGI solutions
- all attempts at Rule 30 currently failing
- the GLM can clearly define and articulate Rule 30 but I need an original action!
Hard much!
Awaiting the next run
- can we have a deterministic Golay error corrected Leech Lattice?
- will that lead into a Quantum Monster Group? QMG!?
A fun journey with Wolfram's Rule 30 puzzle
A *mostly* accurate ai infographic to capture the adventure 👇
Experimenting with Google Vids to make an animation I have thought about for ages
- fun but slow and runs out of credits after a few scenes not to reset until Nov 1st!
- Gemini audio gen
- full on AI slop, but is it less sloppy than other slop?
Sort of an artistic experiment
🤖 Made with AI
DigitalEuan retweeted
Quantum execution for Golay-Leech:
Shots: 20,480
Ground-State Retention: 94.65%
Symmetry: Uniform single-bit error distribution matching IQM Emerald superconducting lattice specs.
Ran out of credits, my Refer code:
openquantum.com/signup?refer…
- help, someone use the code please!
Woah
The Lorentzian lift implemented and produces a valid 26D reflected vector for every ambiguous target solved!
GLM is growing slowly but surely
Quantum execution for Golay-Leech:
Shots: 20,480
Ground-State Retention: 94.65%
Symmetry: Uniform single-bit error distribution matching IQM Emerald superconducting lattice specs.
Ran out of credits, my Refer code:
openquantum.com/signup?refer…
- help, someone use the code please!
Lets try Quantum computing with UBP/GLM concepts.
- starting small with Golay and the Leech Lattice, do they exist in quantum space? - somewhat already proven but not quite the way I am aiming for.
- determinism in quantum 🫣 let's find out!
The Geometric Language Machine (GLM) is a Lean4 verified cognitive architecture built on the Universal Binary Principle (UBP), utilizing a 24-dimensional geometric substrate derived from the extended binary Golay code and the Leech lattice
github.com/DigitalEuan/GLM
fun
Getting closer each development iteration to Geometric Reasoning
github.com/DigitalEuan/GLM
Pushing into 96 dimension reasoning, superposition and cross-domain junctions (ARC-AGI)
- Failures 200 → 40 (80% reduction)
- ambiguity head is essentially perfect (97-100% across all tasks)
- F9 lattice
- computes D* = C + (B-A) in Q24)
github.com/DigitalEuan/GLM
Back to trying to "teach" the GLM some ARC-AGI
Seems the substrate chain is rock solid, just the actual intermediate step learning is the finicky part, pretty cool watching a machine learn something from a base substrate, solve some puzzles and come back to tell me what it needs












