Amund Tveit. Principal AI Engineer. PhD in CS. Personal blog and opinions
My first record attempt on modded-nanogpt, building on other people's work: two open PRs (#379, #375) combined, certified over 18 runs at 34.86 s vs 40.16 s for the current record's code on the same 8×H100 node, plus a small timing fix proposed separately. In review, not an accepted record.
An audited-before-publish formal-methods write-up of a hypothetical G4-FlashTree speculative-decoding stack on Apple Silicon. Tier-classified evidence: 4 mechanically rechecked Lean theorems, 4 TLA+ properties under bounded TLC (with a load-bearing safe-vs-unsafe counter-example pair), 3 ABS behavioral models, and 9 deferred findings.
An early investigation of Prompt Lookup Decoding on Gemma 4 31B 4-bit on M3 Ultra under mlx-lm 0.31.3, with a focus on RAG-style workloads (classic doc RAG, structured / JSON RAG, agentic RAG, code RAG). One class (structured / JSON RAG) shows a +54% decode lift with greedy parity holding at two PLD configurations. A second (passage / classic doc RAG) holds parity at one configuration and breaks at another — fragile evidence. A third (code RAG) breaks parity at every tested configuration. Numbers, RAG-applicability discussion, caveats, and an open implementation question included.
From emulation to native representation. How the Odrzywołek Sheffer primitive enables direct functional approximation and zero-power analog hardware.
mojo-tokenizer converts AI output to readable text at 144 million tokens per second — 3.1x faster than the industry standard. 100% compatible, rigorously benchmarked.
InlineBuffer stores small data on the stack, provides zero-copy slice access via pointer arithmetic, and includes SIMD-accelerated operations. The missing piece for high-performance Mojo.
mojo-json achieves 6.5 GB/s throughput, outperforming the fastest Rust and C++ parsers by 52% on standard benchmarks. Empirical results from Apple M3 Ultra.
Introducing mojo-contrib: 26 pure Mojo libraries for authentication, resilience, observability, and more. Enterprise-grade infrastructure for the high-performance language.
Exploring the lower bounds of LLM quantization - from OpenAIs 4-bit GPT-OSS to research suggesting 3.6 bits per parameter may be the theoretical minimum.
Exploring ultra-efficient AI inference on wearable devices using MLX and Apple Silicon optimization.
Grokking implementations in Jax/Flax and Pytorch
