The breakthroughs available here aren't new app categories. They're products built on archives almost nobody else has — and a few others where the technology has finally caught up to an obvious need.
Each of these rests on something already owned: a corpus, a pipeline, a venue, a relationship.
Point it at a family's boxes of tapes and photos. It transcribes, identifies faces, orders by era, and cuts a documentary with narration. The hard part — ingest, transcode, face-ID, transcripts — is already running here over roughly 5,000 real home videos. Nobody pitches this credibly without having done that first.
Buyers: estates, milestone anniversaries, StoryCorps-adjacent organizations, funeral homes. Boswell is the same product pointed at words instead of footage.
"Every clip of my father." "Every summer at the lake." Face-and-moment retrieval over personal video is memory, not media. Grief is the killer use case and almost nobody serves it. It's a retrieval index over the same pipeline above — the difference between a box of tapes and a shelf you can pull from.
Ancestry records plus 1776 Manhattan parcels plus NYC property records plus TimeWalk geometry. Not a name list — a map you can walk. Addressable family history is the thing genealogy sites have never done, because they were built as databases rather than places.
The health record is already shared with a physician. The unmet need is the appointment itself: a coherent, citation-backed summary the patient walks in with. Second opinion as a deliverable instead of a feeling.
TimeWalk's real breakthrough isn't a model, it's presence — standing at a corner and seeing 1776. AI is the content engine: documents into geometry, agents into crowds and voices. The one that makes people gasp, which is what press and fundraising are made of.
Immersive projection space is a medium nobody has content for. A projection room with no library is a venue; with one, it's a category. Generate the content.
A page about surviving without the internet is still a page. The product is a box that keeps answering when the network doesn't — post-scarcity as hardware instead of a manifesto.
Drafts get written off papers so the creative lead edits instead of starting cold. Industrial scale in comedy means volume of attempts, and volume of attempts is exactly what models are good at.
Read this before picking a date.
Start with the archive product. It can be built this month from code that already exists, and its buyers aren't the builder. The pipeline is 90% done; the missing 10% is a family-facing front end and a cut list. A weekend to prove, a quarter to sell.
Then street-level TimeWalk, because it's the one that makes people gasp.
The trap to avoid: another dashboard. There are plenty already. Build the thing where the archive does the work.