Who is asking
I built this site to follow an argument people have had for decades: whether machines will one day outrun us. Some say it is decades away, some say it is already starting, and the trouble is agreeing on what would count as an answer. The site records what two groups have said about artificial general intelligence, or AGI, a machine that can do most of the thinking work a person can, with every quote dated and linked to where it was said. One is the academics and writers who have talked about this for decades. The other is the people who run the AI companies, who are actually building it and who tend to run well ahead of the evidence with their predictions. Both belong in the picture, and the point is to see how the talk has moved, not to score anyone. Beside that record, the site asks a harder question, are we in the singularity yet, and sets a test that anyone can check.
Why the name?
"Zeitgeist" is a German word for the mood of a time: what people are saying and what they expect. Two kinds of people shape that mood. The long-time thinkers give the long view; the leaders are the ones making it happen, and what they say is the clearest sign of what the industry expects right now.
Method and data
Everything on the home page is worked out from tables anyone can download, and every row in those tables links to where it came from. What people said and expected is kept in one set of tables, and what the world actually did (how good AI got, what the companies reported, how the economy grew) in another; the two are never mixed. Every search that was run to find a quote is written down, so that someone else can run it again. The working documents behind the site are published as they are written, in the project's own terms, for anyone checking the work: the method, the rules for sorting statements, the rules of the test, and every change, with its date and reason. The home page is the plain account of the same material.
Most statements are found by the same fixed searches, run for every person on the panel. Some I come across myself, in an interview or an article, and send in; these are taken in the same way as any other, coded by the same rules, and count in every figure, and each is marked "added from a tip" (the full list, with how many each person has). So that what I happen to read does not decide whose words are on the record, every person on the panel is searched for new statements every week (from 26 October 2026), and every chart that compares people says how recently all of them were searched. When the searches miss something I sent in, a better search is added automatically and run for everyone from the next week, not only for that person; searches are only ever added, and each one is listed on the changelog with its date and the tip that prompted it. How tips are taken in.
I wrote the rules for finding and sorting statements, and I decided every rule and every line of the test; Claude, an AI model made by Anthropic, did the searching and sorting by those rules. The forum posts were sorted by three AI systems from three companies with no person checking them, and systems from different companies can still share blind spots.
Cite as: Fredrickson, J. A. (2026). AGI Zeitgeist Tracker [dataset]. Version of 2026-10-31 (a7bd17a). https://agizeitgeist.com. Data under CC BY 4.0; scripts under MIT. The charts are pictures; the table linked beside each one holds the numbers drawn. The test's lines were fixed on 8 October 2026; the record of that decision.
The story so far
The main changes to how the record is kept and how the test is scored, newest first. Each links to its entry in the changelog, which keeps every change in full with its date, its reason and who decided it.
- October 21st, 2026 Counts of machines doing paid physical work are shown beside the economy part of the test. They are not scored. The change in full.
- October 21st, 2026 Every statement I come across and send in counts like any other, and from October 26th, 2026 every person on the panel is searched for new statements every week. The change in full.
- October 19th, 2026 The record of what the industry's leaders predict becomes one of the site's two voices, beside the long-time thinkers. It records their words; it is not a scorecard. The change in full.
- October 19th, 2026 That record widens from two people to fifteen: the people who lead the companies building the most advanced AI, or the chips it is trained on. The change in full.
- October 18th, 2026 A way in for statements I find myself, in an interview or an article, between the full searches. The change in full.
- October 17th, 2026 A dated record of each industry leader's predictions is set up, with its rules written before any prediction was entered. The change in full.
- October 17th, 2026 The headline becomes how far along the way the parts of the test are, on average, with the slowest part named and a sentence on where current trends lead. The change in full.
- October 17th, 2026 The measurements the pace part is scored on have stopped being updated. The question of what replaces them is opened, and nothing is changed until it is answered. The change in full.
- October 8th, 2026 The test of whether the singularity has begun is adopted, with the line each part must cross fixed in advance. A fourth part, the economy, is added, because it is the only one measured outside the AI world. The change in full.
- October 6th, 2026 The panel of experts whose words the site follows is fixed, so that it cannot drift toward the story. Any change to it is a dated entry of its own. The change in full.
- September 30th, 2026 The project starts. Two outside reviews of the first plan reshape it into the second version of the method. The change in full.