Morning Edition

The Futurist

Thursday, November 12, 2026
AI & Technology Markets & Crypto Ideas Worth Keeping
S&P 5007,520.93▼ 0.37%
Dow Jones52,231.89▼ 0.30%
Nasdaq25,065.87▼ 0.73%
VIX18.69▼ 3.96%
BITCOIN$73,908.00▼ 0.44%
RIPPLE$1.06▼ 1.72%
ETHEREUM$2,059.37▼ 1.54%
DOGECOIN$0.07▼ 1.82%
01

AI & Technology

SIMON WILLISON

Anthropic's IPO Roadshow Video Is Out — And It's Revealing

Anthropic kicked off its IPO roadshow with a video containing previously unreleased details about its compute strategy, delivered by co-founder Tom Brown. This is the clearest public picture yet of how Anthropic thinks about training costs and infrastructure at scale.

404 MEDIA

'Reverse Engineer Anything': AI Companies Crack Down on the Coolest Use of AI

The most creative power-user behavior — feeding AI proprietary hardware schematics, firmware, and closed systems to reverse-engineer them — is getting quietly banned via updated ToS. The escape hatch is local and open-weight models.

SIMON WILLISON

AWS Releases an Official Emulator — Built by 29,000 Agent Commits

AWS shipped a local emulator for its cloud services, written in Rust, with 29,684 commits in the first three weeks — overwhelmingly agent-generated. It's a landmark data point for what agentic software development looks like at scale.

SIMON WILLISON

Google Launches Gemini Decision 1 API

Google enters the decision model category, joining OpenAI and Jev, with a multimodal model priced at $0.10 per 1M input tokens. The race to productize structured AI reasoning for enterprise workflows is now a three-horse race.

A16Z

Why Frontier Models Still Can't Parse a PDF

Reducto's CEO argues that AI agents remain bottlenecked by document parsing — PDFs, scans, spreadsheets — before they can touch real work in healthcare, legal, and finance. The unsexy document layer is quietly becoming critical infrastructure.

COINDESK

MSCI Drops Strategy — And Its Bitcoin Machine May Stall

Strategy's removal from MSCI indexes forces index-tracking funds to sell, cutting off a key capital flywheel that funded its aggressive bitcoin accumulation. A structural liquidity squeeze, not a sentiment story.

02

X / Twitter Signal

Bloomberg broke news that Nvidia is working with an outside firm to find a Jensen Huang replacement — and Kobeissi fired back immediately, telling the board publicly not to do it. The sentiment is widespread: Jensen isn't just a CEO, he's the product thesis. Markets and observers treating this as an existential risk signal for the AI trade.

Ethan Mollick flagged that coding is the blunt counterexample to anyone arguing AI diffusion moves slowly — an entire profession has been structurally transformed in under a year. He also noted GTA VI will be the last major AAA title built without AI, framing it as a cultural inflection point hiding in plain sight.

HUMAN Security data shows AI agent web traffic grew 8x from August to October alone. a16z is calling this the opening of the B2A (business-to-agent) economy — where the primary customer of your API or service isn't a human at all. The design implications for every software product are significant.

Sebastian Raschka cleanly separated three distinct behaviors that often get conflated: hillclimbing (genuinely improving the model), benchmarking (measuring that improvement honestly), and benchmaxxing (gaming the numbers). In a week full of model releases and comparison claims, this taxonomy cuts through most of the noise.

03

The Thread

Two stories this week that look unrelated are actually the same story. AWS shipped a codebase with nearly 30,000 agent-generated commits in three weeks. AI companies are banning users from reverse-engineering proprietary hardware with AI. Both are about who controls the leverage when AI becomes the primary builder. Agents are already writing infrastructure code at a pace no human team can match — and incumbents are racing to wall off the domains where that capability is most threatening to them.

"The primary users of the Internet are no longer human. They're AI agents."
The Anthropic IPO roadshow is the week's most underleveraged story. When a company going public releases previously unknown information about its compute strategy in a video pitched at institutional investors, that's not a press release — it's a signal about how Anthropic thinks its moat actually works. The compute section, delivered by co-founder Tom Brown, suggests the thesis is infrastructure depth, not model benchmarks. That's a different pitch than OpenAI is making, and it matters for how the next 18 months of the AI race gets funded.

Meanwhile, the Jensen Huang succession leak — whether real or trial balloon — rattled markets at exactly the wrong moment, with Nasdaq already sliding and the AI trade under pressure from data center policy fears. Kobeissi's public plea to the board reveals how thin the bench is perceived to be. When a company's entire valuation narrative lives inside one person's credibility, that's not just a governance problem. It's a risk factor hiding in the footnotes of every AI-adjacent fund.