Thibault Schrepel

Thibault Schrepel

@profschrepel.bsky.social

Associate Prof VU Amsterdam • Faculty Affiliate Stanford • Into Running 🏃🏻 #antitrust #AI #complexityscience #digitalmarkets 📕 www.thibaultschrepel.com 📻 https://podcasters.spotify.com/pod/show/scalingtheory

455 followers136 following686 posts

TikTok’s best fact: 77 to 82% of its users also use Facebook or Instagram, so it locks nobody in. The Commission has accepted that kind of argument 19 times in merger cases. In court it has never changed an outcome. If the top court follows its adviser, it will not under the DMA either.

thecompetitionsystem.substack.comThe DMA’s first opinion, and the case law it walked past (Bytedance, C‑627/24 P)Advocate General Biondi cites twelve judgments under sixteen case numbers, and not one of them for a rule of competition law.

I spent a couple of days redesigning the Network Law Review. I hope you enjoy it, and above all that it gives every article the spotlight it deserves.

Network Law Review@networklawreview.bsky.social

The Network Law Review has a new design. Each theme is now a constellation in which every star is a publication, and each article draws the network of work it connects to. Twelve themes to explore, and a lot more to find along the way. www.networklawreview.org

One digital single market. National definitions of small. In telecoms and payments, EU law lets each Member State set the threshold under which a firm counts as small. Same EU law, same firm, different regimes. thecompetitionsystem.substack.com/p/who-is-sma...

thecompetitionsystem.substack.comWho is small in EU law? An empirical answerEU digital law grants small firms 77 reliefs but never settles who counts as small. A firm that qualifies under one law can fail under the next.

My reading suggestions for September. What acquirers actually keep when they buy a startup, the causation element in monopolization cases, the economics of recursive self-improvement, why AI detection cannot work, and the university after AI. www.networklawreview.org/september-20....

networklawreview.orgReading suggestions – September 2026 - Network Law ReviewI share the papers and articles I enjoyed reading this month. This edition runs from what acquirers actually keep when they buy a startup to the causation element in monopolization cases, a practical ...

The Draghi report diagnosed regulatory burden and fragmentation as the reason European firms stay small. The European Innovation Act repeals no Union obligation, reduces none of the 27 enforcement environments, and amends exactly one instrument: the EU trade mark Regulation.

thecompetitionsystem.substack.comThe European Innovation Act, testedI coded all 366 rules of the EU's new Innovation Act. It would be the first digital instrument to exclude a firm for who owns it. Four exhibits below.

The fastest change we have ever recorded. Two editions ago, no competition agency reported an internal LLM assistant. This year, 19 of 30 run or build one. From nothing to the most common tool in the field in two years. Fifth report, free. shorturl.at/wmEaW #computationalantitrust

Five errors in the Commission’s Booking decision, most of the increment calculation thrown out, nothing annulled. Commission v Tetra Laval, the closest authority in the graph, never named. My audit of the Booking/eTraveli judgment, citation by citation, is out.

thecompetitionsystem.substack.comKeeping a market where it is can now be the harm (Booking/eTraveli, T‑1139/23)The General Court says a merger may be blocked for sustaining a weak market, in two sentences no EU court had written before. The fence around them is narrow, and the appeal is open.

OUT: The only report on the use of AI in antitrust written by the agencies themselves. 30 of them this year, in their own words: which models they run, what they let them do, what they refuse to do... Here is the fifth edition of the Stanford #ComputationalAntitrust report shorturl.at/wmEaW

Tomorrow is going to be a very special day. Not so much for the champagne reception, but very much for what we have to unveil. Contributions from about 30 antitrust agencies showing up all the progress they’ve made when it comes to the use of AI and other computational tools.

Let me tell you something. I sent this empirical mapping of what needs to be fixed to have a proper digital single market to about 10 Members of the European Parliament, all of them involved in the Omnibus or the simplification process.

I like when AI is making data that I’ve had for a very long time. I find it useful. I’ve ranked over 2,000 movies on IMDb. Giving all the data to Claude, I was able to spot new movies that are missing to my watch list. How exciting is that?

The Brussels effect is usually established from outcomes ➝ the foreign firms that end up following EU rules without being required to do so. I went to the texts instead and coded every scope provision of the 34 EU digital acts, the 273 rules that decide whether the rulebook applies to you at all.

thecompetitionsystem.substack.comWhere EU digital law stops273 scope rules across 34 acts. As a result, twenty of these acts reach firms with no establishment in the Union, and the fastest-growing clause conscripts a foreigner’s representative.