Today’s highlights: why policing insider trading in prediction markets is getting trickier — and new evidence that later school start times help kids sleep and learn.
Why catching insider trading is so difficult
Insider trading still pays, and it’s getting weirder. Recent headlines include an American service member who allegedly turned $33,000 into more than $400,000 using classified information. Even bigger sums have appeared on prediction markets like Polymarket, where well‑timed bets tracked outcomes such as whether the U.S. would bomb Iran or capture Venezuela’s Nicolás Maduro.
If a lawmaker or someone with privileged information places those bets, you’d think it would be straightforwardly illegal. In practice, it’s not. The enforcement challenge lies less in the bets themselves and more in how these platforms are built.
Some prediction market operators present a U.S.-facing, regulated front end, but their real activity runs on largely anonymous, crypto-based back ends operating internationally. That split makes it hard for regulators to see who is placing big wagers. Cryptocurrency can increase user anonymity, and when large stakes move through offshore or pseudonymous accounts, investigators struggle to tie trades to specific people.
Blockchain adds another wrinkle: transactions are permanent and transparent on-chain, but linking an on‑chain wallet to an identifiable person is often difficult. That permanence helps reconstruction once identities are known, but it doesn’t necessarily help catch or prevent suspicious trades in real time.
Senator Richard Blumenthal has proposed treating some prediction markets more like regulated sportsbooks, bringing them under stricter U.S. rules to reduce anonymity and limit opportunities for insider advantage. The idea is to close the gap between the compliant front end and the anonymous back end so platforms can’t effectively have it both ways. Whether legislation will be enough — or whether enforcement will keep up with evolving crypto structures — remains an open question.
Quick notes: other topics we’re tracking this week
– Advances in AI and model deployment continue to surprise — sometimes companies hold back powerful models for safety concerns, yet other similarly capable systems exist elsewhere.
– Geopolitical and institutional questions keep bubbling up: debates over NATO’s relevance and past tensions between presidents and the Federal Reserve are in the news.
News we’re watching
1) Can the grid meet surging electricity demand?
Demand forecasts tied to data centers and AI compute have shocked utilities. One major Texas utility reported asking for planning that would bring 122 gigawatts of demand to its system over the next five years — while the regional grid’s current maximum deliverable capacity is roughly 85 gigawatts. That implies increasing available capacity by around 143% for just that utility’s customers, a pace far beyond normal growth (typical annual increases are under 2%).
Former FERC chair Jon Wellinghoff calls this “off the charts.” There are reports of single projects proposing 5 GW in areas whose total current loads are a fraction of that — for example, a 5 GW project compared with roughly 840 MW for a whole city. Practical limits, permitting battles, community opposition and competition among developers mean a fair share of proposed projects may never get built; some developers have already canceled dozens of planned data centers amid local pushback.
In short: technically and politically, it will be extremely hard to add this much capacity quickly. Expect many announced projects to be trimmed, delayed or dropped.
2) Does more sleep improve student performance?
California’s 2022 law that pushed back middle and high school start times gave researchers an opportunity to study real policy effects. The results are encouraging: students did get more sleep overall; boys showed meaningful improvements in mental health measures; and many students improved their math and English scores. The biggest academic gains were seen among Hispanic students and those from economically disadvantaged backgrounds.
Why this matters: changing a schedule — not pouring money into new programs — delivered measurable benefits. Later start times are a low‑tech adjustment that appears to improve both well‑being and learning, especially for groups who have historically lagged on test outcomes.
Bottom line
Both stories are about incentives and structure. In markets, architecture matters: platform design, offshore operations and crypto rails can create opportunities for misuse and make enforcement hard. In schools, a simple institutional change — shifting start times — can produce better outcomes because it aligns system incentives with human biology. Policy and regulation can matter as much as individual behavior, for better or worse.