Subquadratic LLM Solution
Subquadratic LLM Solution
What it does
I discovered a flaw in classical computer architecture that explains A.I context window decoherence and created a universal patch that can make any LLM use VRAM subquadratically without becoming decoherent. The website compares a base vs patched Mistral 7b
Does the same job
all alternatives →
- KRKVSplit – Run 2-3x longer contexts on Apple Silicon2025 · github.com · ▲272
I discovered that in LLM inference, keys and values in the KV cache have very different quantization sensitivities. Keys need higher precision than values to maintain quality. I patched llama.cpp to enable different bit-widths for keys vs. values on Apple Silicon. The results are surprising: - K8V4 (8-bit keys, 4-bit values): 59% memory reduction with only 0.86% perplexity loss - K4V8 (4-bit keys, 8-bit values): 59% memory reduction but 6.06% perplexity loss - The configurations use the same number of bits, but K8V4 is 7× better for quality This means you can run LLMs with 2-3× longer…
- TVTiny-vLLM – high performance LLM inference engine in C++ and CUDAMay 2026 · github.com · ▲205
- 1B1-Bit Bonsai, the First Commercially Viable 1-Bit LLMsMar 2026 · prismml.com · ▲430
OpenLIT's Zero-code LLM ObservabilityOct 2025 · ▲136Trace LLM requests + costs with OpenTelemetry monitoring
- YAYet another memory system for LLMs2025 · github.com · ▲165
Built this for my LLM workflows - needed searchable, persistent memory that wouldn't blow up storage costs. I also wanted to use it locally for my research. It's a content-addressed storage system with block-level deduplication (saves 30-40% on typical codebases). I have integrated the CLI tool into most of my workflows in Zed, Claude Code, and Cursor, and I provide the prompt I'm currently using in the repo. The project is in C++ and the build system is rough around the edges but is tested on macOS and Ubuntu 24.04.
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 19d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
Launched alongside, March 2026
the whole month →
Switch from ChatGPT to Claude with import memory feature
AI · Mar 2026 · claude.com
