DHS told courts it wanted to search Samuel Tunick's wiped GrapheneOS phone for child abuse material. The real reason: suspected "terrorist activities" linked to his Stop Cop City activism. The pretext reveals how federal agencies exploit child safety rhetoric to justify political surveillance.
AI decompiling tools are turning closed software into open playgrounds — clone it, modify it, ship it. The same capability empowering hobbyist hackers to unlock their devices is threatening to commoditize every proprietary software stack on earth.
Private keys for eight eMudhra intermediate CA certificates have been compromised — certificates trusted by Mozilla, Microsoft, Apple, and Chrome. This is a foundational crack in the web's trust infrastructure, and any software relying on these chains is now exposed.
Solana's most significant network upgrade in six years launches Wednesday, targeting speed and validator efficiency at scale. If it holds under load, it resets expectations for what a high-throughput L1 can actually deliver.
Jeremy Scheff's essay, surfaced by Willison, mourns what coding agents have quietly killed: the joy of the logic puzzle. The paycheck survives; the craft, for many, does not.
The All-In crew dissects CoreWeave's debt market trouble as a possible signal that AI infrastructure spending is outrunning returns. Paired with Ken Griffin calling AI "almost certainly a bubble," the cracks in the trade are getting harder to ignore.
Google's Gemini PM dropped a staggering number: 25 trillion tokens processed in the first 72 hours after Gemini 4 Argon launched. Demis Hassabis promptly retweeted it. That's not a benchmark — that's a demand signal, and it suggests the intelligence consumption curve is steeper than even optimists projected.
Citadel's Ken Griffin told an audience that AI is "almost certainly" a bubble, pointing to companies struggling to convert AI spending into actual productivity gains. Coming from one of the most data-driven investors alive, this isn't doomerism — it's a valuation warning dressed in polite language.
Mollick published a chart-heavy piece mapping the last 30 months of AI development against an exponential curve — and the curve is holding. His argument: we keep underestimating pace because humans are bad at intuiting compounding. The post is required reading for anyone building a 12-month roadmap right now.
NorForge researchers uncovered over 1,100 AI-assisted attack projects in a single cluster, targeting common web vulnerabilities. The eMudhra CA compromise lands the same day. Security isn't losing ground slowly — it's losing it in bursts.
Two stories today that look unrelated are actually the same story. DHS lying about why it searched an activist's phone, and eight trusted web certificates getting their private keys cracked — both are infrastructure failures. One is legal infrastructure weaponized against its stated purpose; the other is cryptographic infrastructure that wasn't maintained to basic standards. Trust is the load-bearing wall of the digital world, and it's getting stress-tested from every direction simultaneously.
Mollick's exponential chart and Scheff's programming obituary are the human sides of the same ledger. The curve is real and it is not slowing. What's dying isn't programming — it's the comfortable pace at which institutions, regulators, and security teams could absorb change. The AI trade may or may not be a bubble in the financial sense. In the attention and adaptation sense, the bubble already popped. Everyone is behind.