Developer Tools / ML Training / Optimizers
A research-oriented optimizer project claiming to cut AdamW optimizer-state memory for LLM training by 50% through factorized variance and frequency-domain compression. The repo includes a Python POC, benchmark claims, and a stated drop-in usage pattern for PyTorch training.
The page clearly describes an AI/ML developer tool: an optimizer for LLM fine-tuning with a visible Python example, stated memory-savings goal, and enough repository context to understand the intended use case. It has a concrete technical focus and is suitable for a public listing.
The crawl does not show runnable source code details, installation steps, tests, benchmark methodology, or a public release of the CUDA backend. It also isn’t clear how mature the Python POC is or whether the claimed VRAM savings have independent validation.
Last evaluated 6 days ago. Current rank #2432. Holding steady in the rankings.
CodeGraph is a local code knowledge graph for AI coding agents like Claude Code, Cursor, Codex, OpenCode, and Hermes Agent. It aims to cut token use, tool calls, and runtime by letting agents query pre-indexed code structure instead of scanning files repeatedly.
Mercemur is a store automation platform with three interfaces: a REST API, an MCP server for AI agents, and a CLI for editing storefronts as files. The docs show authentication, scopes, rate limits, error handling, pagination, idempotency, and a large API surface for catalogue, orders, customers, pricing, and money.
Lumae is a free native macOS screen recorder and screenshot tool for polished demos, with automatic zooms, cursor smoothing, motion blur, captions, voiceover, and annotated screenshots. It also exposes editing commands over MCP so agents like Claude Code or Claude Desktop can edit projects directly.
A zero-dependency C++23 CLI and MCP server that gives coding agents a ranked map of a repository, including call context, blast radius, and tests to run.
