The Ai Signal on our homepage isn't a list somebody edits once a week and forgets about. It's a live feed, and we get asked often enough how it actually works that it seemed worth explaining plainly.

Where it comes from

Under the hood, a small server-side function pulls the AI-coverage feeds from a handful of outlets we trust — TechCrunch, VentureBeat, The Verge, and MIT Technology Review among them — merges everything together, and sorts it by publish time. That function refreshes on its own; nobody on our end is manually copying headlines into the page, and nobody has to remember to update it before it goes stale.

What "our team, not an algorithm" actually means

You'll see the line "pulled and sorted by our team, not an algorithm" above the feed, and it's worth being precise about what that means now that the pipeline itself is automated. The mechanics — fetching, merging, sorting by recency — are handled by code. The judgment isn't. We chose which outlets are trustworthy enough to pull from in the first place, and just as importantly, which ones we don't: no engagement-optimized aggregators, no SEO content farms repackaging someone else's reporting, no unverified rumor blogs. That selection is the editorial layer. The line isn't a claim that a human copy-pastes every headline by hand — it's a claim that a human decided what "worth showing you" means before any code touched it.

What counts as "shipped"

Not everything an outlet publishes about AI is a story we'd want leading the feed. We're generally looking for:

  • A real product, feature, or model that's actually available — not a leak, a rumor, or a "sources say" piece with nothing to point to yet.
  • Funding, acquisitions, or business moves reported by an outlet that did real verification, not just a press release rewritten as news.
  • Coverage that affects people actually using these tools — a policy shift, an outage, a safety incident — over pieces that exist mainly to generate clicks off a company's name.

The pipeline can't make that call on its own, which is exactly why source selection matters more than any filtering logic we could write.

Why it isn't sorted by engagement

It would be easy to rank the feed by whatever's getting the most clicks or shares, and plenty of aggregators do exactly that. We sort by recency instead, from outlets we've already vetted, because a feed optimized for engagement tends to reward outrage and speculation over anything actually useful to know. The goal of the Signal is to answer "what happened in AI recently that's worth two minutes," not to maximize how long you stay on the page.

The pipeline is automated. The judgment about what deserves to be in it isn't.

If something looks off

Automated feeds occasionally pull in something that shouldn't be there, or miss a story that should. If you spot either, tell us in the community — the source list gets adjusted based on exactly that kind of feedback.