It is the culmination of months of 6am alarms and questionable playlist decisions...
This is your last chance to:
โ support a brilliant cause
โ effectively bribe me to keep moving past mile 22
Every donation helps!
2026tcslondonmarathon.enthuse.com/pf/schrepel ๐
This Sunday, 26.2 miles stand between me and a medal I plan to wear to the supermarket (not really). ๐โโ๏ธ
I am running the TCS London Marathon for Starlight Children's Foundation UK helping seriously ill children smile through treatment.
โI bear witness to capitalism being reborn in the Arabian desert.โ
Jason Potts launches a new @networklawreview.bsky.social series on Vision 2030 as a digital-era competition policy program. He argues Saudi Arabia is running the most ambitious competition policy experiment in the world right now.
6/ Use cases extend past research. An agency can check a draft decision against its own prior record before issuance. A policymaker can map the eight EU digital regulations and find where their definitions collide.
Field-neutral. Works on any corpus.
LINK: papers.ssrn.com/sol3/papers....
5/ Two worked examples. One academic research corpus of 674 papers. One European Commission competition decisions dataset of 508 decisions from 1977 to 2025.
Same pipeline for both. Builds in a few hours. Free tools except Claude Code.
4/ The orientation is absence-first. In mature fields, the useful finding is a gap.
Keyword search gives you what the corpus says. The graph surfaces what it does not. Absent paths between concepts theory suggests should connect. Terms used everywhere and defined nowhere.
4. Claim-level extraction instead of document-level
5. A persistent hypothesis register
6. A complexity-theoretic diagnostic layer
3/ The guide adapts the method for research purposes. Six additions:
1. Schema design that encodes authority hierarchies
2. Centrality-weighted wiki generation
3. A six-step protocol for hypothesis generation
2/ The method comes from Andrej Karpathy. Keep a folder of sources you trust. Let an AI assistant extract concepts, map their connections, compile Wikipedia-style articles on every theme the corpus covers.
Every new query feeds back into the graph.
1/ A paper or decision you read ten years ago survives (at best) in memory as a conclusion.
New guide: โBuilding a Digital Brain for Research.โ A pipeline that turns a folder of PDFs into a queryable knowledge graph, here specifically designed for research purposes.
papers.ssrn.com/sol3/papers....
Always entertaining to see nationalism dismissed by many (rightfully as far as Iโm concerned), only to embrace the same logic at a larger scale under the banner of โEuropean digital sovereignty.โ Apparently itโs fine if you just zoom out a bit.
It is apparently, and I donโt think such a general rule makes sense, reality is more complex.
Well, you donโt have to imagine that anymore. The European Commission just announced this may be illegal if you dominate in these product markets. ec.europa.eu/commission/p...
1/2 Imagine opening the weights of your AI model. Imagine integrating it into other products and services to recoup your investment, but letting users and developers around the world fork it, build on it, and deploy it elsewhere.
โ Patent holdup
โ Killer acquisitions/Kill zones
โ Default bias
โ Common ownership.
Four influential antitrust theories of harm with limited empirical evidence.
youtu.be/n1XDsPweLqk
8/ Institutional design point lands well too. Antitrust is one instrument among many. Democracy law, tax policy, and campaign finance exist for a reason. Asking antitrust to solve everything is not ambition. It is category error.
Highly recommended. papers.ssrn.com/sol3/papers....
7/ The double standard section is, I find, the sharpest part. GLB are rigorous when attacking consumer welfare. They abandon rigor entirely when deploying the dashboard. That asymmetry is telling.
6/ Consumer welfare has real limits. Income effects, distributional blind spots. The authors do not hide this. Their point โ the fix is better methods within the framework, not a framework that is unworkable by design.
5/ The paper is equally sharp on how mainstream economics actually works. Economists narrow their models precisely to preserve reliability. Broad welfare frameworks exist in theory. They are not applied in practice.
4/ Without weights, the dashboard can justify any conclusion. A progressive uses it to block mergers. A libertarian uses it to permit them. No objective basis exists to say who is right.
3/ The core problem is simple and devastating. A dashboard requires welfare weights. Nobody knows how to set them. That is not a minor gap. It makes the whole framework indeterminate.
2/ The target: the โwelfare dashboard,โ the idea that antitrust should weigh democracy, inequality, small business protection, and more, alongside competition. GLB (Glick, Lozada, Bush) are the main proponents.
1/ What happens when two great scholars (@briancalbrecht.bsky.social & Erik Hovenkamp) get together?! They cut through a lot of noise in the antitrust goals debate. Worth reading carefully.
A few emails per year. Research and podcast episodes at the intersection of antitrust law and AI. No tracking, no noise, no cost.
80+ antitrust agencies are already part of the Stanford #ComputationalAntitrust network. The newsletter is how we share what we find.
docs.google.com/forms/d/e/1F...
New on Stanford #ComputationalAntitrust:
LLM-driven agents collude tacitly in a Bertrand duopoly, no instruction needed. This paper shows it can be detected with Smart Agent-Based Modelling. This is the future. law.stanford.edu/publications...
In 1999, Albert-Lรกszlรณ Barabรกsi tried to publish a paper on networks for 4 years. Every journal rejected it.
Then Nature and Science both published his work within months of each other.
NEW episode of #ScalingTheory!
โ open.spotify.com/episode/35e4...
โ youtu.be/ZdyFvJrtc9o?...
โ
Here with the one and only Jon Barnett (USC Law). Seminar today! alti.amsterdam/event/barnet...
This and more in ๐ฎ๐๐๐๐๐๐๐๐๐๐ ๐พ๐๐๐๐๐๐, an open-access series tracking the main developments in EU competition law, on the Network Law Review. networklawreview.org/competition-stories-two/
The EU just lowered the bar for compelled access to dominant platforms. No indispensability required, if you opened your platform voluntarily, you may have to share it. The gap with US antitrust just got wider. networklawreview.org/competition-stories-two/
