2026-09-23·by Sijie Wang#standmeet#product

thesis

Parent: product

Treat AI conversation as the primary input to a personal, public-facing intellectual layer.

The loop: the user agentically curates the good parts of an AI conversation → pushes them to StandMeet via MCP → the corpus accumulates → the corpus serves multiple use cases.

Two load-bearing choices:

  • Explicit push, not automatic capture. Not "record everything" — that's a privacy and signal/noise disaster. The user decides what enters.
  • Curated corpus, not raw transcript dump. What lands is the articulated substance, not the whole log.

Rooted in ai-conversation-is-the-substrate; the ergonomics depend on MCP (why-now).

A second ingest route landed and is now the larger one: the vault. The owner's Obsidian vault syncs wholesale — 1082 notes, folder-to-genre classification, the folder tree materialized as the corpus tree (obsidian-sync-mechanism). It does not contradict the two load-bearing choices; it relocates where they are exercised. Explicit push becomes what the owner chose to write down at all, and curation becomes the raw → wiki promotion the owner performs in their own editor — the filter moved from the moment of pushing to the act of writing.

That is worth stating because it changes who the product is for. The MCP-push story needs a person who has AI conversations worth keeping. The vault story needs a person who already keeps notes — a much sharper and much smaller audience, and one that arrives with the corpus already built.

about this entry

One of sijie's wiki entries. The AI on this site is grounded in the same corpus and answers in sijie's voice, with citations back to entries like this one — answering costs sijie money, so it waits behind a code: enter an access code →