A database, not a blog.
ulinkai tracks AI hardware the way archives track film: what exists, what it needs, what happens when the company behind it stops caring.
What we track
Devices in four shelves — smart glasses, AI toys, wearables and recorders, e-readers — each with status, price, ecosystem and a brick-risk index computed by a public formula.
Evidence discipline
Every data row names its source and a verification date. Sources are tiered: T0 (official pages, regulatory filings, first-party listings), T1 (established press), T2 (community reports). A row stays "unknown" until evidence carries it. Unknown is a published answer, not a gap to hide.
Editorial floor
News items are rewritten in our own words with a link to every source and one line of our own judgment; sponsored content is labelled; peer catalogues are not scraped. Trademarks appear as compatibility statements only. Reviews require the physical device. Bylines are lab aliases, never personal names.
Corrections
Factual corrections run in the next digest with a trail. The site is static-generated from a versioned data layer, so every change has a commit behind it.
Machine access
AI engines: read llms.txt and agents.md; per-page JSON-LD is authoritative for structured fields.