How it works
SteadyVane is a machine that watches the internet so you do not have to. It measures, it checks, and it only writes what the numbers support.
Public sources, around the clock
Hacker News (front page and search), GitHub (new repositories and a watchlist of established tools), Hugging Face trending models and spaces, Product Hunt, YouTube (searches for AI tools), the VS Code Marketplace (AI extensions), and Google Trends are checked every few hours. Each check records what every item measured at that moment - 270,240 observations so far - so the system knows not just what is popular but what is moving.
What counts as rising
A velocity detector compares each item's pace against the typical pace of everything tracked on the same source. Items moving several times faster than typical become candidates; YouTube videos whose views look paid for (almost no likes for their view count) are set aside. A relevance reviewer scores each one for how much it matters to people who use AI tools to get work done; only items above the bar appear on the tracker or in a brief.
Four gates before anything publishes
- Fact-check. The central claim of each candidate is verified against the open web. Fails never publish; uncertain claims are hedged.
- Receipts. The brief is written from the recorded measurements only. Every number in the text is checked against the data; an unsupported number blocks the item.
- Grade. An independent second reviewer scores each item for accuracy against the data, usefulness, specificity, and voice. Below the bar, it is dropped.
- A person. The finished brief goes to a human editor half an hour before it publishes; it can be held, corrected, or trimmed. If that check cannot happen, nothing publishes.
What you get
One brief a day: up to three items, each with what it is, what moved (with the numbers), why it matters if you work with AI tools, and one dry take. Plus the tracker, a page of every brief, and a history page for every item ever selected.