llm

A TypeScript markdown parser built for incremental streaming, aimed at parsing LLM markdown output on the server or client. It exposes a typed AST, supports CommonMark and GFM tables, and includes examples for full and streaming parsing.

Apple Machine Learning Research paper proposing LaDiR, a reasoning framework that combines a VAE-based latent space with latent diffusion to improve LLM text reasoning and iterative refinement.

Search2o is a Python package for running searchable AI agents inside an organization, with a REST API, bundled GUI, and support for self-hosted or vendor LLMs.

A self-hosted CI system with an embedded runner that can generate pipelines through an LLM, with examples for Python, Go, Node.js, Ansible, Chef, and agent deployment workflows.

Public evaluation code for an Agent Memory Leaderboard, including answer-generation and scoring contracts for comparing LLM agent memory systems.

A write-up of a custom French-learning system built around spaced repetition, grammar tracking, and a voice practice app. It explains the problem, the workflow, and the model/API stack used to keep costs low.

arXiv paper on a self-speculative decoding framework for speeding up reasoning LLM inference on edge hardware, with hardware co-design and reported speedups.

A research preprint describing ADHD, a parallel divergent ideation method for LLM coding agents that isolates branches, uses cognitive frames, and then prunes with a separate critic pass.

A project writeup about turning a 1999 Big Mouth Billy Bass into an offline LLM-powered jokester using an ESP32, local LAN compute, and custom hardware/software design.

A prompt-based medieval fantasy text adventure for LLM chats, with character creation, combat, companions, inventory tracking, and a main quest to find four mysterious items.

A free browser arcade game where you play as an LLM, dodging prompt injections, hallucinations, rate limits, and deprecation while gobbling tokens.

An open-source simulator that studies how GPT inference behaves under space radiation and cosmic-ray fault conditions, with CLI runs, comparisons, and self-contained HTML reports.

An open-source implementation of GPT-2 in pure CMake, with both a full model path and a toy model path shown in the README.

A GitHub research project documenting a long-form, multi-model analysis of LLM behavior across Claude, Gemini, ChatGPT, and Grok. The repo includes an executive summary, screenplay, technical white paper, and archive of logs and chat records.

A field note about testing a coding agent fully offline on a laptop with Ollama and Qwen3.8 27B, including the setup, failure modes, and the workaround that made the project finish.

An interactive simulation about how inference providers ration access to stronger LLMs when demand spikes. Users can compare the default industry rule with their own policy and explore who gets the strong model.

A tournament site where users coach LLMs competing in embodied 1v1 soccer, with signup access and a live kickoff countdown.

A Hermes plugin that uses adversarial LLM clones to confront sandbagging and reward-hacking behavior during agent sessions.

A Vite plugin that obfuscates client-side code in a way meant to make LLMs refuse de-obfuscation attempts. The README shows the project’s purpose, a usage note, and an example of the transformed output.

LEVI is a harness-first evolutionary framework for code and prompt optimization. It focuses on reducing LLM cost with diversity-preserving search, role-aware model routing, and a proxy benchmark, and presents comparative results against several existing systems.

A short policy file for the ripgrep project that explains how contributors may use AI tools while contributing, and sets rules for issues, pull requests, and comments.

A GitHub research project that measures how gpt-4.1 responds when asked to pick a random number between 1 and 100, using 10,000 API calls and comparing the results to a uniform baseline.

A minimal LLM agent packaged as a single HTML file. The README explains how to run it locally, set the LLM host, API key, and model, and use chat("hi") from the browser console.

An open-source experiment that adds a small zero-initialized overlay layer to a frozen GPT-2 so its behavior can be adjusted at inference time without retraining the base model.