Not a judge. A progress report.
AI is talked about more than it is tested. We gather both, label which is which, and let you see the gap.
Every claim about AI on this site sits next to the evidence for it: who tested it, how, and when. A launch is not a test, a test is not a study with real people, and a study is not wide use. We keep those apart and show how far each task has got.
That is why the site is called Intelligence Progress: it reports the progress AI is actually making, task by task, rather than handing down verdicts. We never give our own verdict on whether an AI product is good. The stages are computed from the evidence by rules we publish, and nobody can pay to change one. AI helps us find and summarize material, and it is never allowed to set a stage, a count or a quote.
Blue is what was actually tested: an independent benchmark, a study with real people, a regulator's decision or a court record.
Apricot is what someone says: a maker's own numbers, a launch announcement or a named person's prediction, quoted word for word and dated.
Lilac is what forecasters and prediction markets currently expect, read on the day we show it, kept apart from both.
Everything AI, checked.
From the tasks you might hand to AI today to the chips, power and rules behind it.
Can AI do it yet?
Everyday tasks, each placed on a six-step ladder from a maker's claim to wide use with a verified record.
OpenSuperintelligence watch
Dated predictions by named people, set against the measured trends and the crowd's forecasts.
OpenAI agents
AI that does whole tasks on its own, scored against people, with what happens when it goes wrong.
OpenAI and coding
What coding tools score on independent tests, and what studies of real developers found.
OpenAI and your work
Job data, trials of AI at work, and what the research says about your job.
OpenChips, data centers and power
The largest AI data centers, the chips inside them and the electricity they use.
OpenAI rules
Laws and deadlines in the US, the EU, the UK and China, each with its date and source.
OpenWhat's new
The latest launches, studies, rules and failures, linked to the original source.
OpenWho's behind it.
Yale Marc
Founder and publisher
- About 20 years building software
- Projects for NASA and DARPA
- Three books about AI
I'm Yale Marc, the founder and publisher of Intelligence Progress. I'm a senior software engineer with about 20 years of experience, including projects for NASA and DARPA. These days I build software with AI tools, so I see first-hand what they do well and where they still fall short.
I also write about AI. My three books on it tell its history, look at its coming impact, and, in a novel, imagine the first five years after an AI wakes up in a lab.
This site doesn't ask you to trust my opinion, or anyone's. It gathers the news every morning, sets what AI makers claim against what has actually been measured, and links every claim to its source.
Why I built it
AI is announced faster than anyone can check it. Whether a model can really do a task, whether a tool is worth paying for, or what AI means for your job usually has an honest answer, but it's buried in benchmark results, research papers and court records that few people have time to read, and drowned out by launch hype. I wanted one free, independent place, in plain English, that shows what AI can actually do today and keeps up as it changes.
My books about AI
I also run Simplify Software, which publishes this site, and publish its two sister sites, Robotics Verdict and Longevity Verdict.
Three advances. One question.
Intelligence, robotics and longevity are the three leading technological advances of our time. Intelligence Progress and its two sister sites, Robotics Verdict and Longevity Verdict, each cover one in depth and ask the same question: how far has it actually got?
Intelligence Progress
What AI can actually do today, set against what its makers claim: independent tests, studies with real people, regulators and court records.
Robotics Verdict ↗
Which robots are real and how far each has got, from humanoids and home robots to robotaxis and surgical robots.
Longevity Verdict ↗
Every new longevity study, trial, test and clinic, gathered each morning, with what named experts say about it.
All three are built the same way. Each sets what is claimed against what has been shown, says plainly when something is only a demo or a promise, and links every claim to its source.
The fields overlap more each year. AI now runs inside many of the newest robots, and it is being tested on reading scans and searching medical research. Where a story belongs to one of the other fields, we point you to the sister site that covers it in depth.
Get in touch.
Intelligence Progress is published by Simplify Software, 3823 Tamiami Trl E, Unit #252, Naples, FL 34112.
Use our contact form for advertising and sponsorship, press, corrections or anything else.