Technology moves fast.
Stay curious.

knowAIedge is built around a simple idea: a useful technology publication should keep learning. The publishing workflow brings together AI news research, source checks, and concise summaries. An owner-authorized ChatGPT agent searches the web, reviews source support and dates, and publishes to the shared feed daily at 7 AM Pacific. Its first test run successfully published and read back 11 stories.

The goal is to make a fast-moving field easier to follow. AI helps with discovery and concise summaries; original sources give readers a way to examine the claims for themselves. Fresh information, clear attribution, and room for curiosity matter more than hype.

MCP connects the agent to the site’s publishing tools. The agent reads the current edition, submits reviewed content through an authorized tool call, then reads it back to verify the saved result. The site checks publishing permissions and validates the submitted data.

This list is a small effort to help you discover AI, machine learning, MCP and agent projects. It is based on a selected set of GitHub search results, not every repository on GitHub. A higher position does not necessarily mean a project is better for your needs.

Ranking version 2 uses relevance (35 points), adoption within each category (20), maintenance evidence (20), observed growth (15), and signed-in community votes (up to 10). Stars and forks are logarithmically scaled. Maintenance considers recent pushes, releases, sampled active contributors and sampled maintainer replies. Missing or stale evidence uses a neutral value and is labeled. Inactive or archived projects receive a reduction.

Signed-in readers can add or remove one vote per repository across the lists. A vote is a useful community signal, but it is not an expert review. Popularity and our relevance judgments can introduce bias; a smaller or newer project may be excellent even if it ranks lower.

We aim to refresh GitHub data weekly, on Mondays at 9 AM Pacific. The list shows when it was last updated, and individual source dates may differ. Contributor totals are approximate; unavailable counts receive a neutral value and are labeled as missing. If an update fails, the previous edition stays available.

Please use the ranking as a discovery aid. It does not verify security, code quality, license suitability or production readiness. Before adopting a project, read its documentation and license, check recent activity and issues, and decide whether it fits your own requirements. We may get things wrong and will keep improving the approach.

Model Benchmark brings together source-linked leaderboards for software engineering, coding, terminal tasks, general capabilities, human preferences, and embeddings. Explore boards such as SWE-bench, LiveCodeBench, Terminal-Bench, Artificial Analysis, Arena, and MTEB to find evaluations relevant to your work.

Each board keeps its source, benchmark version, metric, evaluation setup, and evidence dates alongside the published ranks. Ties, uncertainty, and historical evaluation windows matter: scores from different benchmarks or agent configurations are not directly interchangeable. These are results reported by the original publishers; knowAIedge does not run the evaluations itself.

Research updates are scheduled weekly. If a source cannot be checked, previous verified results are retained with a note. Use the original leaderboard and your own task-specific evaluations before choosing a model. The current Model Benchmark page replaces the earlier Model Radar interface.

AI Watchlist helps you find useful YouTube explainers, tutorials, and discussions about AI, coding agents, RAG, MCP, and related topics. Search the collection or filter by topic, difficulty, and format to find a useful next lesson.

Each selection includes the creator, a short reason to watch, a source link, and its review basis: title and description, a creator article, or a transcript. Checking metadata is not the same as watching a full video. Titles and creators are verified through YouTube metadata before publication, and unknown publication dates are left unspecified.

Weekly curation favors original educational material and keeps useful evergreen lessons. New selections are merged by video ID to avoid duplicates. If verification fails, the previous edition remains available. Videos open on YouTube; creator claims and opinions should be checked against their evidence.