InsightKeeper

InsightKeeper is an internal AI watch agent built by a small firm to track how artificial intelligence evolves day by day. It scans a curated set of sources — X, LinkedIn, YouTube, Substack, podcasts, and arXiv — and pulls out the signals that matter. Each day it spots 5 to 10 relevant developments, qualifies them with summaries and key ideas, and stores everything in a searchable knowledge base. The tool was originally a consumer Mac and iPhone app, but the company stopped commercializing it and now uses it as internal infrastructure to feed their client work, R&D, and long-term guidance.

The agent works in three steps: watch, spot, and qualify. It monitors the selected sources continuously, then applies AI to identify the most useful signals. A human decision layer lets the team ignore, track, test, or integrate each signal. Behind the scenes, four specialized AI agents handle bugs, dependency updates, documentation, and R&D improvements. The system has been in development for 19 months, runs 5,470 automated tests, and has processed 16 billion tokens. The company emphasizes that InsightKeeper is not a demo — it is a maintained, stress-tested product that makes their work sustainable.

InsightKeeper is no longer offered to new users. It now serves as a strategic internal asset for the firm, which focuses on helping French SMEs and mid-market companies with AI diagnostics, training, and systems modernization. The tool directly informs their judgment: it helps them distinguish what is becoming possible, what remains fragile, what deserves a test, and what is mostly noise. For leadership teams working with the firm, InsightKeeper means their advice is grounded in real, current observations rather than hype.