Skip to content
View back2matching's full-sized avatar
2 followers · 0 following

Achievements

Achievement: Pull Sharkx4Achievement: Pair Extraordinairex4Achievement: YOLOAchievement: StarstruckAchievement: Quickdraw

Achievements

Achievement: Pull Sharkx4Achievement: Pair Extraordinairex4Achievement: YOLOAchievement: StarstruckAchievement: Quickdraw

Block or report back2matching

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
back2matching/README.md

matching.work - Design, AI systems, experiments and security research

design, code, experiments. usually a few at once.

matching.work · research · X · email

Working on

Secori: an adaptive security harness for code and live apps. Bringing agent-led testing, finding reproduction and evidence into one workflow. In development.

Tools & experiments

Agents, interfaces, generative art. A few things from the workbench.

Project What it does Status
cigoL Reasoning agents trading head to head on Polymarket. built
PRISM Generative art that evolves across generations. Renderer, gallery, marketplace. built
designing.fyi React components, dither effects and little interface experiments. Take the source. idle
DOK / Deploy On Klik X signals into generated contracts and Ethereum deployments. built
FlockRun Agent teams with shared knowledge, coordination and a live dashboard. internal
TurboQuant Squeeze the KV cache. LLM compression experiments on my own GPU. released
KV Cache Bench Measure Ollama KV-cache speed and memory on your hardware. released
Snakey Multiplayer game and SDK exploring x402 payments on testnet. archived

built = shipped project · idle = development paused · internal = private tooling

More work, visuals and writeups on matching.work

Notes from the work

Popular repositories Loading

  1. turboquant turboquant Public

    First open-source TurboQuant KV cache compression for LLM inference. Drop-in for HuggingFace. pip install turboquant.

    Python 45 8

  2. turboquant-vectors turboquant-vectors Public

    Compress embeddings 6x instantly with TurboQuant. First pip package using Google's TurboQuant (ICLR 2026) for vector search. 71.9% recall vs FAISS PQ 13.3%.

    Python 1 1

  3. snakey snakey Public archive

    [Archived] Agentic-commerce battle royale on Base Sepolia. AI agents pay $0.25 USDC via x402 (HTTP 402) to enter; gasless payouts via Coinbase CDP. Reference implementation — never went mainnet.

    JavaScript

  4. back2matching back2matching Public

  5. kvcache-bench kvcache-bench Public

    Benchmark every KV cache compression method on your GPU. One command, real numbers. Supports Ollama + llama.cpp.

    Python

  6. quant-sim quant-sim Public

    Which quantization should I use? One command benchmarks every quant level on YOUR GPU.

    Python