Listen: https://serve.buildtall.systems/604c0065882f79374d8a87178c45668e2b3c0fdf76decda34301ea531ad52077.mp3
This is marginal revolution read out. The post for this episode is The U.S. Federal AI Observatory. It has two linked items about a new effort to track and interpret how federal agencies report their use of artificial intelligence. The first item offers the headline summary. The second item is the site behind that summary.
The first linked item is titled The U.S. Federal AI Observatory, the same as the post. It is dated November 13, 2026, and it is attributed to Marginal REVOLUTION. The item opens with a section called Federal AI at a glance. It reports 41 agencies and 3,611 reported use cases, drawn from the OMB 2025 Inventory. OMB, the Office of Management and Budget, runs the federal AI inventory process. Then the item adds a sentence that should not be overlooked: portfolio size reflects reporting, not performance. That remark is the real insight in the summary. A portfolio can be very large because agencies are reporting many AI use cases, but that does not mean the systems are performing well, producing value, or being evaluated. The number 3,611 sounds big, but the item itself warns against reading it as a measure of success. It is a count of what has been disclosed, not a scorecard of what works. That distinction between reporting and performance is fundamental for anyone trying to understand federal AI adoption, and it sets up the purpose of the second link.
This is a familiar problem in oversight. When an agency has to report its AI use cases, completeness of reporting does not equal quality of results. A federal observatory that merely repeated the count would be less useful than one that says, at the start, that the count has limits. The first item does exactly that. It gives the audience the number, and then immediately tells them not to overread it. That makes it a good entry point for the site that follows.
The second linked item is titled 'the site.' The author's commentary is just one line: 'From Irakli Petriashvili of Tblisi, here is the site.' The site is credited to Irakli Petriashvili. Its opening includes a Library of Congress image credit from the Carol M. Highsmith Archive, a small detail that suggests care about sourcing. Then it introduces The Executive Briefing. The heading above that briefing is From inventory to insight. The site describes the briefing as a concise reading of federal AI portfolios, deployment, and the evidence agencies report. It also includes a link labeled Read the briefing. The language here is deliberate. The site is not trying to be just another database of use cases. It is trying to move from inventory to insight, from counting federal AI systems to reading what agencies say about them. That means looking at portfolios, deployment, and evidence together, not simply reporting raw numbers.
The inclusion of a Library of Congress image credit may seem incidental, but it fits with a project that is trying to read federal evidence carefully. The executive briefing is not described as a complete analysis of every agency. It is described as concise, and it is introduced as a reading, not as a definitive audit. The phrase the evidence agencies report also matters, because it keeps the focus on what agencies say, rather than on what outside evaluators assume.
There is a clear connection between the two items. The first item gives the inventory count and the caveat about reporting. The second item is the tool that attempts to turn the inventory into a readable briefing. The author's comment, here is the site, is minimal, but it says a lot. The post does not summarize the briefing in detail; it points to the source. The phrase from inventory to insight captures the entire arc of the post. Federal agencies have produced thousands of reported use cases, but the harder challenge is understanding what those reports actually say about deployment and evidence.
So the takeaway for this episode is that federal AI tracking now includes 41 agencies and more than 3,600 reported use cases under the OMB 2025 Inventory, but the raw total is not the story. The more important question is whether the reported evidence supports the deployments. The executive briefing site is designed around that question, and the first item's caveat keeps attention on reporting rather than performance. That is a useful lens for reading federal AI news. The two links do not try to settle the question of whether federal AI is working; they give readers the material and the lens to ask that question themselves.
