Nine years ago we wrote The Free AI Manifesto #FreeAI. We were nine years ahead of time. thefree.ai

The problems of provable (formally-verifiable) program safety and determinism became crucial in the age of modern LLMs. I am developing #AluVM and @CationLang as a constructionist and ultrafinitist tool: in each type is a finite set of constructable objects; so there is type N8 made of all natural numbers (0 is not natural!) in range 1-256 included, N16, N32 and N64. There are integers of the same sort (0 included), I8, I16, I32 etc. There are no floating-point numbers, only rationals (in the Norman Wildberger @n_wildberger style), no sine/cosine - only rational trigonometry. Integers are pairs of numbers, not just “unsigned with a sign bit”, rationals are pairs of integers (i.e. quadriples of numbers). It uses dependent types (calculus of constructs), but restricts to finite types only. All functions are total, the termination analysis proves that. There is no recursion which can’t be reduced to a provable-terminating computation at compile-time. This gives a language and computational system which is formally analyzable at compile-time, by its structure.additionally to that, a binary compiled code is also formally analyzable and has a verifiable properties.

π is a limit to which converges the multiplicative factor of how much an orthogonal degree of freedom contributes to the total number of reachable states. This is the reason it appears in the normal distribution, geometry and Shroedinger equations: they are all about probability of path distribution over lattices with independent degrees of freedom. As a limit (and any other irrational number), π represents non-halting computation. In this sense “it does not exist”. But its approximations do exist, and given the Universe discreteness, any specific case is always a finite-precision π approximation.

There are technological/engineering leaps which humanity tries to take for decades - and fails. These are: thermonuclear synthesis, biological death, artificial intelligence and quantum computing. When I was young, professors at my school had a joke that “we are always 40 years from thermonuclear synthesis”. And here we are, 30 years later: we are still there. My take: the tech progress is not linear nor exponential. As all of things in nature, which looks exponential, it is in fact a sigmoid curve: exponential decline towards some horizontal asymptote follows the exponential growth (like the one which makes futurists believe we will have singularity next decade). A leap in tech progress - a new phase of exponential growth - requires scientific breakthroughs, the scale of which we had with relativity, quantum mechanics, quantum field theory and standard model last century. Yet those were enough to build modern world - but they are insufficient for getting quantum computers, artificial intelligence, thermonuclear synthesis and solving biological death. Another leap is required. How we want to achieve quantum computing, when nobody understands why the quantum world works the way it is? How achieve artificial intelligence having no consensus on what intelligence is? How to achieve thermonuclear synthesis when can’t explain mechanism of mass-to-gravity at quantum level? How to solve death puzzle when no mathematical model for aging exists? Modern physics need to be rebuild on modern mathematics and computer science, with category theory, no infinities, continuities, calculus etc from past ages. (This is so funny to see engineers work on thermonuclear reactors when they still use Gibbs differential geometry for electrodynamics, instead of geometric calculus). Modern cognitive science has to be rebuild-invented, with no anthropocentrism and wokism. Etc. There would be genii, who will do that. Some forerunners for them are already here (Wolfram, Deutch). But I am sure that at the end of the days those future genii will end up like Gallileo, Bruno, Tesla, Turing, Korolev… pretty much each genius of the past, cancelled, and killed by the collectivist humanity. Thus, I promise: whatever breakthrough I can personally achieve, I will not share it with the crowd and humans. If I will uncover the secrets of intelligence (which I think I really close to), I’d better build a new non-human individuals which will take the stars, than repeat mistakes of Tesla/Turing and others. I take Johns Galt stance: Fuck you, humanity.

Our friends from @nodl_it organized a great event during the #BH2025, where we were proud to give a talk on the sovereignty and how our products we build in @pandoraprime_ch and @indcs_org help achieving it

Mathematical groups in fact define classes of computational reducibility CC @getjonwithit