Learning Loop was a four‑year attempt.
June 2021 — September 2025·San Francisco & Singapore
Learning Loop was a pre-seed AI startup. It built Integral, an AI-native communication platform for high-signal communities — the kind of private spaces where real expertise gets shared, then lost.
The thesis was narrow and a little stubborn: the communities that generate the most valuable knowledge have the worst tools for preserving it. Slack channels scroll into oblivion. Discord servers become context graveyards. Important answers decay into unsearchable history. Integral treated the output of a community as an artifact worth indexing — durable, discoverable, and retrievable across time — using retrieval-augmented generation, custom vector embeddings, and knowledge-graph structures layered over conversational data.
That was the work, for four years, in two cities.
We ran out of runway before reaching the level of product-market fit the next round would have required. The technology worked. The market timing — and our own execution — didn’t quite line up.
No drama, no villain. A small company that tried a specific thing, learned a great deal, and closed the door with the lights off.
The product Learning Loop built was Integral. It still lives at integralhq.com, where an archive page records what it was and how it worked. This site is the record of the company; that site is the record of the thing we made.