Eren

@tsotchke @badwolffen let it play sm64 or other speedrun game maybe the results could be different than from tas bots would be generally interesting to see the results and maybe as a side by side comparison

@badwolffen let me see if i can make a nice looking demo of that just for you

Noesis is extremely good at Tetris

**@tsotchke (tsotchke on X) is an independent open-source developer focused on high-performance tools for quantum many-body physics simulations, neural network quantum states (NQS), tensor networks, lattice gauge theories, and hybrid quantum-classical computing.** His work centers on making advanced quantum simulation accessible and practical—bridging condensed matter physics problems (like spin liquids and topological order) with modern ML/AI workflows. Everything is built with a strong emphasis on performant, low-level C implementations that support bindings to Python, Rust, JS/WASM, etc., and are designed for both research and as backends for physics-informed ML. ### Key projects he's actively releasing and iterating on: - **SbNN (spin_based_neural_network)** — A pure-C11 framework for simulating quantum spin systems using Neural Network Quantum States + Variational Monte Carlo (VMC). The brand-new v0.5 release (just pushed) includes a full empirical study of the kagome antiferromagnetic Heisenberg model at 27 sites under periodic boundaries. It uses five independent observables, symmetry projections over the full C6v group, and extracts the modular S-matrix spectrum from the low-energy manifold. The results strongly reject a simple Z₂ toric-code topological order and instead point toward a U(1) Dirac spin liquid (7 quasi-degenerate S=1/2 states, rank-4 modular S spectrum that is *not* Hadamard). GitHub: github.com/tsotchke/spin_… - **Moonlab** — His flagship hybrid quantum simulation library. The recent v0.2.1 release adds major tensor-network and gauge-theory capabilities, centered on **Clifford-Assisted Matrix Product States (CA-MPS)**: it splits a quantum state |ψ⟩ = D|φ⟩ where D is an exact Clifford/stabilizer layer (symbolically cheap) and |φ⟩ is a lower-entanglement MPS. This delivers massive speedups (e.g., 64× lower bond dimension and >13,000× faster at n=12 for stabilizer-rich systems). Other highlights: variational ground-state search (**var-D**), gauge-aware stabilizer warmstarts (for toric-code, surface-code, or Gauss-law constraints), a full 1+1D Z₂ lattice gauge theory pipeline, and a stable C ABI with Python/Rust/JS bindings. Moonlab is explicitly positioned as a **physics-aware simulation backend for AI/ML workflows** — you can drop it into PyTorch loops as a QuantumLayer, encode data via feature maps, evolve states, measure observables, and train hybrid models with symmetry/entanglement constraints baked in. GitHub: github.com/tsotchke/moonl… - **Supporting libraries** he maintains or contributes to: - **libirrep** — Irreducible representations and equivariant neural nets tailored for spin physics. - **Eshkol** — A LISP-like scientific computing language with built-in autodiff, aimed at hybrid quantum-classical programming. - Quantum RNG tools and other low-level C primitives for reproducible, high-precision numerics. ### Overall direction Through **Tsotchke Corporation** (@tsotchkecorp), he frames the work as “Transform the World through accessible Quantum Intelligence.” The goal is practical quantum-adjacent computing: solving hard open problems in condensed matter (spin liquids, topological phases, gauge theories), building tools that run efficiently on classical hardware today while preparing for near-term topological/hybrid quantum hardware, and making these simulations usable inside normal ML pipelines rather than isolated physics code. His X activity is mostly release notes, technical deep-dives on the math (symmetry projections, modular S extraction, Clifford decompositions, etc.), and occasional playful commentary. The bio “美少女戦士ギターゴッド” is a fun Sailor-Moon-meets-guitar-god reference. In short: he’s building a modern, open-source stack for serious quantum spin physics and topological simulation that’s designed from the ground up to be fast, extensible, and ML-native. The SbNN v0.5 kagome study is a nice showcase of the scientific output these tools are already producing.

@0kCyb0rg @tsotchke results look incredible nice mc by the way 😎

@Bloomverse @tsotchke ? 👀

The old way: → Roblox or Fortnite → Limited assets → Terrible payment rails The new way: → Built by Akasha → On Solana → Speak world into existence Games finally have real infrastructure. We built Akasha, and she builds your world.

@Clip_Master__ this isnt fun

Dude thought the chat was money donations.mp4

@0kCyb0rg @tsotchke 🔥

@tsotchke ❤️

so cool that my views are being suppressed for no reason again. why would i want anyone to see these open source software releases anyway? don't mind me, im just advancing the entire field of computational physics for the sheer fun of it.

@tsotchke 🔥💪

what if you could play with quantum physics on your laptop... not just simulate it, but actually understand it? Moonlab is your ultimate library for: • tensor networks (simulate systems that normally blow up) • error mitigation (turn noisy results into usable data) • nonlocality (prove a system is actually quantum) it’s fast, low-level, and correct, demonstrably validated against NIST + pq-crystals. you can now: - measure Bell-state entanglement yourself - run VQE on real molecules (H₂, LiH, H₂O) - explore topological phases + Chern numbers - experiment with quantum-safe crypto primitives no heavy frameworks required. just a clean, MIT-licensed core with Python/Rust/JS bindings. not a cloud platform. not a Qiskit replacement. this is for learning, experimenting, and building intuition. if you’ve ever wanted to see where classical physics breaks, and you can do it today. 🌙 github.com/tsotchke/moonl…

@tsotchke 🔥

i've now pushed libirrep, a pure-C11 library of SO(3) / SU(2) / O(3) representation-theory primitives (spherical harmonics, Wigner-D, Clebsch–Gordan, e3nn-style tensor products) for equivariant neural networks and exact-diagonalisation spin-system physics. github.com/tsotchke/libir…

@NVIDIAGeForceDE PRAGMATA RTX

PRAGMATA ist jetzt #RTXON – mit Path Tracing und DLSS 4! Zur Feier verlosen wir diese einzigartige GeForce RTX 5090 mit Hugh und Diana – perfekt für dein nächstes Abenteuer auf dem Mond. Du willst sie haben? Schreib „PRAGMATA RTX“ in die Kommentare, um teilzunehmen!

i have just pushed an update to the spin-based neural network v0.4. this version lays the foundation for significantly improved neural network pillars as well as deepens the topological computational capabilities. an ongoing series of updates are incoming. github.com/tsotchke/spin_…

@RemiliaRabbi @defidarling how i met your mother :)

@defidarling technically found my wife on a kpop forum, but shortly discovered her twitter and it was what make me fall in love with her 13 years ago

can you find love on X? anyone know of any success stories

Eshkol comes with everything you need for neural programming with real-time signal processing right-out-of-the-box? check all of the capabilities of the Eshkol standard library including advanced functions for real and complex mathematics, ML, & much more! github.com/tsotchke/eshko…

Very f’n based. Nice work fellas.

if you are a compiler developer or interested in contributing to the development of an up-and-coming functional programming language for high performance scientific computing and AI please don't hesitate to reach out.