# Roberto Capobianco > I assess whether AI systems do what their builders say they do. Independent technical assessment of AI, reinforcement learning, and optimisation systems, for investors deciding whether to fund a technical claim and for teams deciding whether to build on one. Advisory pages are English only. ## Pages - [Home](https://robertocapobianco.com/): entity home. Biography, affiliations, and the research profiles that resolve the author. - [Advisory](https://robertocapobianco.com/services/): the two engagements and their prices — technical due diligence from EUR 12,000, and a fractional research lead retainer from EUR 6,000 per month. - [Publications](https://robertocapobianco.com/publications/): peer-reviewed publications, each with its DOI. - [Book](https://robertocapobianco.com/book/): Grace: The Value of Choice — a 457-page textbook, reinforcement learning, from bandits to policy optimisation, published October 2026, written for a master's course at Sapienza University of Rome. The exercises are free and public. - [Research](https://robertocapobianco.com/research/): three research directions, each stating the position it argues and the work that supports it — controllable and coachable agents, interpretable decision making, and neuro-symbolic and non-Markovian reinforcement learning, with an emphasis on whether those properties hold under realistic evaluation conditions. - [Notes and teardowns](https://robertocapobianco.com/notes/): notes on evaluating learning systems, and claim-by-claim analyses of systems with public technical claims. ## Notes - [The contract says it will never do that. Where is that written in the system?](https://robertocapobianco.com/notes/the-contract-says-it-will-never-do-that-where-is-that-written-in-the-system/): A penalty expresses a preference; an enforced constraint can provide a guarantee. Most promises a company makes about what its system will never do are implemented as preferences, and the difference shows up only in the cases the promise exists for. - [When does a research team become an engineering risk?](https://robertocapobianco.com/notes/when-does-a-research-team-become-an-engineering-risk/): Almost every team says their approach generalises. Very few have paid to find out, because the experiment costs money and produces no headline — it only tells you whether you have a method or a solution to one problem. - [Does this company have proprietary technology, or a proprietary evaluation?](https://robertocapobianco.com/notes/does-this-company-have-proprietary-technology-or-a-proprietary-evaluation/): The cheapest thing to improve is not the system but the measurement. A metric a company controls can move without the product moving — and the same fact makes a good private evaluation one of the more defensible things a team can own. - [When is a benchmark result evidence of a company advantage?](https://robertocapobianco.com/notes/when-is-a-benchmark-result-evidence-of-a-company-advantage/): A benchmark tells you a ceiling is reachable. It does not tell you that the thing which ships is the thing that was measured — and the distance between them is usually where the company actually lives. - [Is the explainability in the product, or in the slide deck?](https://robertocapobianco.com/notes/is-the-explainability-in-the-product-or-in-the-slide-deck/): "We have explainability" is not a yes-or-no answer. What matters is where the explanation lives: inside the mechanism that makes the decision, or in a layer added afterwards — because the two fail in different ways. - [The demo is impressive. What would make it investable?](https://robertocapobianco.com/notes/the-demo-is-impressive-what-would-make-it-investable/): A demo is one point. What matters is how far you can move from it, what moving costs, and whether the point sits inside the region the system was built for. - [Does this team know why their system works, or only that it does?](https://robertocapobianco.com/notes/does-this-team-know-why-their-system-works/): An explanation that has only been illustrated looks exactly like one that has been verified. The question is not whether a team's account is true, but how much of it they have tried to falsify. - [What I look for when a startup shows me its best result](https://robertocapobianco.com/notes/what-i-look-for-when-a-startup-shows-me-its-best-result/): A result that is excellent on the headline metric and strangely poor somewhere else is more informative than a mediocre headline. The odd secondary number is usually where the specification failure is hiding. ## Not listed here Older posts, including the Italian ones, are kept at their original URLs and listed in the sitemap. They are an archive for a different audience and are not part of the advisory offer.