Thursday, March 12
Thursday, March 12
The Pentagon Tried to Make an Example of Anthropic. It Made a Brand Instead.
Microsoft filed in federal court today asking to block the Trump administration's "supply chain risk" designation against Anthropic, a label originally built for foreign adversaries like Huawei. This is Microsoft, which has billions invested in OpenAI, defending OpenAI's biggest competitor. The precedent is terrifying for any tech company with government contracts, and everyone knows it. Meanwhile, the commercial fallout is running in the wrong direction for the Pentagon. Claude app downloads now exceed a million per day. Paying users have doubled since January. "The company that told the Pentagon no" turns out to be a killer value proposition.
The Pentagon Tried to Make an Example of Anthropic. It Made a Brand Instead.
Microsoft filed in federal court today asking to block the Trump administration's "supply chain risk" designation against Anthropic, a label originally built for foreign adversaries like Huawei. This is Microsoft, which has billions invested in OpenAI, defending OpenAI's biggest competitor. The precedent is terrifying for any tech company with government contracts, and everyone knows it. Meanwhile, the commercial fallout is running in the wrong direction for the Pentagon. Claude app downloads now exceed a million per day. Paying users have doubled since January. "The company that told the Pentagon no" turns out to be a killer value proposition.
The Dead Internet Strikes Back
The commercial web runs on a handshake nobody actually shook hands on. Users show up and pay attention. Platforms serve content. Browsers sit politely in between. The whole $258 billion digital advertising economy assumes the entity browsing is a person, one who can be persuaded, who might click "Buy Now," who has a credit card and occasionally poor impulse control.
- Alan Turing's original test imagined a human trying to fool a judge into thinking they were a machine. We've arrived at the exact inverse problem.
- About 80% of AI agents don't properly identify themselves when visiting websites, per recent bot-detection data.
- Amazon v. Perplexity, filed late 2025, remains the closest thing to a legal test of whether platforms can require the browsing entity to be human.
Today's stories all circle the same uncomfortable realization: the line between human and synthetic online is getting genuinely hard to locate.
The commercial web runs on a handshake nobody actually shook hands on. Users show up and pay attention. Platforms serve content. Browsers sit politely in between. The whole $258 billion digital advertising economy assumes the entity browsing is a person, one who can be persuaded, who might click "Buy Now," who has a credit card and occasionally poor impulse control.
- Alan Turing's original test imagined a human trying to fool a judge into thinking they were a machine. We've arrived at the exact inverse problem.
- About 80% of AI agents don't properly identify themselves when visiting websites, per recent bot-detection data.
- Amazon v. Perplexity, filed late 2025, remains the closest thing to a legal test of whether platforms can require the browsing entity to be human.
Today's stories all circle the same uncomfortable realization: the line between human and synthetic online is getting genuinely hard to locate.
Vibe Coding Gets a Reality Check Today
Anthropic confirmed an outage hitting Claude Code and Claude.ai today around 11:33 AM ET. Developers mid-flow with agentic coding workflows hit a wall. The API stayed up, but the consumer products stuttered. Down Detector peaked at 1,387 reports.
Claude Code now dispatches a team of agents on every pull request, hunting for bugs that quick skims miss. Anthropic calls it "built for depth, not speed." Arrives just as the volume of AI-generated code makes thorough human review harder than ever.
An AI coding bot blunder reportedly triggered AWS outages. Infrastructure that half the internet depends on, disrupted by the same kind of automated coding tool developers are adopting at record pace. The irony needs no embellishment.
Reportedly motivated by frequent platform outages and disruptions, OpenAI is building a GitHub competitor. The company keeps expanding beyond models into developer infrastructure. First the IDE integrations, now the repository itself. The surface area grows.
After nearly thirty years, JavaScript has Temporal, a proper datetime API. Bloomberg Engineering published a retrospective on the nine-year standards journey. The shared Rust library powering it across browsers is a quiet triumph of cross-engine collaboration that deserves more attention than it'll get.
A Mozilla engineer argues WebAssembly should work in script tags without JavaScript glue code. The proposed Component Model has been in development since 2021. Google is evaluating it. Community reaction: cautiously optimistic, with real caveats about reference type passing.
Five Rust crates disguised as time-related utilities were quietly exfiltrating .env files to attackers. Another supply-chain attack targeting the developer toolchain. If you pulled in anything date-related from an unfamiliar source recently, maybe go check.
Moves Shakeups and Tens of Billions in Chips
Nvidia invested in Thinking Machines Lab and committed to deploying at least one gigawatt of Vera Rubin systems. The chip supply deal is reportedly worth tens of billions. Murati turned down a Meta acquisition last year. Two cofounders have since returned to OpenAI.
Gemini was the fastest-growing AI website in February, up roughly 643% year-over-year. ChatGPT grew 37% in the same period. Grok and Claude posted gains too, but nobody matched Google's pace. Distribution advantages remain undefeated.
VC money in physical AI is on pace to nearly double last year's totals, even excluding outlier mega-deals. Capital is migrating from chatbots and foundation models toward robotics, warehouse automation, and factory AI. The atoms are getting their turn.
Andreessen Horowitz published its latest consumer AI rankings. The periodic snapshot always sparks debate about which categories are emerging and which darlings are fading. This edition lands as the competitive landscape looks more fragmented than ever.
Graber transitions to Chief Innovation Officer after growing Bluesky to 43 million users. Toni Schneider, former Automattic CEO, steps in as interim. The AT Protocol's champion moves to a role focused on the decentralized vision rather than daily operations.
Favorite Featured Stories
On August 19, 2025, Amazon blocked Perplexity's browser from its marketplace. Within twenty-four hours, Perplexity shipp...
An autonomous customer-service agent started approving refunds it shouldn't have. Review scores climbed. Satisfaction me...
Somewhere right now, a contract worker is choosing between two AI-generated paragraphs. The rubric says "quality." It sa...
In most modern programming languages, 0.1 plus 0.2 does not equal 0.3. A rounding artifact, harmless in most contexts, c...
On August 19, 2025, Amazon blocked Perplexity's browser from its marketplace. Within twenty-four hours, Perplexity shipp...
An autonomous customer-service agent started approving refunds it shouldn't have. Review scores climbed. Satisfaction me...
Somewhere right now, a contract worker is choosing between two AI-generated paragraphs. The rubric says "quality." It sa...
In most modern programming languages, 0.1 plus 0.2 does not equal 0.3. A rounding artifact, harmless in most contexts, c...