Glassbrain is an open-source tool for inspecting and comparing knowledge graphs created by AI agents. Graphs use a simple JSON format of nodes and relationships; a command-line interface validates them, reports structural problems, and compares snapshots or a graph against a Git commit. The browser viewer organizes nodes by containment - such as folders and files, or areas and features - rather than using a force-directed layout. Relationships appear on top without rearranging the map, so users can track changes without the whole display shifting. It supports search, node details, version playback, side-by-side comparisons, and layered 2.5D views, and can open graphs from tools such as Graphify, Understand-Anything, MCP memory servers, and Obsidian.
The project’s central argument is that agents can turn code, documentation, research, and plans into readable, evolving graphs without a specialized extraction pipeline: the model writes the graph, while Glassbrain checks and displays it. Included skills guide agents through building and reviewing graphs in stages. Demonstrations report a 540-node Flask code map with 751 typed relationships; a documentation-and-product analysis that found 421 features and merged them into 268 user stories, with 23 of 24 sampled stories confirmed against source code; and research snapshots that expose 25 corrected facts. A planning example links decisions to requirements and makes revisions visible as graph changes. The tool can be tried through a live demo or run with npx to produce a standalone HTML view.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
Researchers tested 16 LLMs on all 55 TIS-100 puzzles, giving each one attempt and time to refine its code. The models found 133 new performance tradeoffs across 18 puzzles and set three outright cycle records, though human players improved on most solutions within days. (tis-100.wincon.dev)
Australia is considering a new criminal offence for AI companies that fail to take reasonable steps to prevent harm caused by their models. The government is also weighing an AI safety watchdog with power to impose significant fines, following a hack of a Medicare statistics portal involving OpenAI. (abc.net.au)
AI is increasingly becoming a gateway through which people find information. Audiences still resist having AI produce their news. (niemanlab.org)
Nvidia holds a central position in the AI industry, making disruption to the company potentially catastrophic for the wider ecosystem. The race for advanced AI is also intensifying economic and geopolitical conflict, while concentrated corporate power and infrastructure face growing political resistance. (france24.com)
LLM inference can require reading about 26 GB of model weights to generate each token from a 13-billion-parameter model. The explanation shows how model size drives memory bandwidth demands during text generation. (youtube.com)
Deletion Tuning matches full fine-tuning across three LLM post-training methods by zeroing fewer than 0.05% of model weights. Its compact masks also preserve prior-task accuracy better than full fine-tuning, LoRA, and random sparse tuning. (arxiv.org)
Today's best Hacker News stories, summarized and screenshotted, one email a day.