Louis Maddox
@permutans.bsky.social
Combinatorially curious https://spin.systems
2026 spoiled user mindset: "how dare this semantically useful info be computationally any less than convenient"
The amnesty on bugs as you take a library you built from zero into the correct/performant phase is… humbling
is it idiomatic for data normalisation libraries to talk of invariants and determinants? well it is now
eyeballing 60% of WikiData Swahili values without a label in the language
Filleting out the entirety of English Wikidata might still change the game (Claude parquetmaxxing self-actualisation incoming)
Revisited schema inference in genson-core // polars-genson + made big speed gains + memory reductions 🏁 6+ day estimated processing time for Wikidata now <24h
Uncanny how accurately Claude can estimate the exact wall time change of individual perf optimisations
Packaged this, after some tape allocation adjustments it parses up to 3x faster than simd-json, gains most on files with large rows (Wikidata claims benchmark) 📁 cuJSON-rs github.com/lmmx/cuJSON-rs 🦀 Rust crates.io/crates/cujson 🐍 Python pypi.org/project/cujs...
Louis Maddox@permutans.bsky.social · 4d ago📄 CuJSON: A Highly Parallel JSON Parser for GPUs (ASPLOS '26) dl.acm.org/doi/pdf/10.1... 📁 cuJSON github.com/AutomataLab/...
Accidentally stumbled on what seems to be the same thing Claude Code (etc) use for displaying markdown: `cargo binstall mdcat-ng` (gives you mdcat and mdless)
📄 CuJSON: A Highly Parallel JSON Parser for GPUs (ASPLOS '26) dl.acm.org/doi/pdf/10.1... 📁 cuJSON github.com/AutomataLab/...
OK enough cute Opus visuals time to put some TB on the GPU
📝 hypothesmith: Hypothesis strategies for generating Python programs, something like CSmith github.com/Zac-HD/hypot...
GitHub - Zac-HD/hypothesmith: Hypothesis strategies for generating Python programs, something like CSmith
Hypothesis strategies for generating Python programs, something like CSmith - Zac-HD/hypothesmith
github.com
Anyone else been getting the system reminder saying ABANDON ALL HOPE
ah now to ship my perfectly working softwSEGV SEGV SEGV SEGV SEGV
Snake feels like a poor demo for Jev (in that it makes the latency conspicuous)
“a proof-carrying language for AI-written kernels” akitaonrails.com/en/2026/09/1...
New AI-Focused Language Just Released: Bend 2
Bend 2, from Brazilian researcher Victor Taelin, arrived promising C speed, GPU parallelism, and formal proofs that block AI mistakes. I tested it by porting two of my Rust projects right at launch. P...
akitaonrails.com
dKV-Cache: delayed caching arxiv.org/abs/2505.15781
dKV-Cache: The Cache for Diffusion Language Models
Diffusion Language Models (DLMs) have been seen as a promising competitor for autoregressive language models. However, diffusion language models have long been constrained by slow inference. A core ch...
arxiv.org
Trying this thing called a "blog" it's like when you type words out instead of prompting
dQwen3.5: Hybrid-Attention Diffusion Language Models arxiv.org/abs/2609.20751 dQwen3.5-2B huggingface.co/UT-IFML/dQwe... dQwen3.5-4B huggingface.co/UT-IFML/dQwe... dQwen3.5-9B huggingface.co/UT-IFML/dQwe...
dQwen3.5: Hybrid-Attention Diffusion Language Models
Adapting a pretrained autoregressive (AR) model is a cost-efficient route to a diffusion language model (DLM). While nearly all such adaptations start from a full-attention transformer, AR modeling ha...
arxiv.org
Hmmm CUDA kernel 10x suboptimal y'all mind if I spiral and crash out rq