Morning Briefing - February 2, 2026

AI & Agents

Dario Amodei's "The Adolescence of Technology" Essay

Anthropic's CEO published a 20,000-word essay that's generating significant discussion. The central metaphor: a "country of geniuses in a datacenter"—millions of AI agents, each smarter than a Nobel laureate, coordinating at superhuman speed. Amodei predicts this by ~2027.

Five risk categories: loss of control, biological misuse, authoritarian capture, economic upheaval, and social destabilization. He predicts 50% of entry-level white-collar jobs could be displaced within 1-5 years, creating a "crisis of meaning."

The notable quote: "The technology itself doesn't care about what is fashionable, and we are considerably closer to real danger in 2026 than we were in 2023."

He rejects doomerism but frames this as humanity's "technological adolescence"—the period where we either mature into responsible stewards of powerful tools or don't. Anthropic's 2026 goal is to train Claude to "almost never go against the spirit of its constitution."

Source: Dario Amodei - The Adolescence of Technology | Fortune Analysis

OpenAI and Anthropic IPO Race

Both companies are racing toward IPOs. OpenAI has begun informal talks with Wall Street banks for a Q4 debut at ~$500 billion valuation (though they don't expect profits until 2030). Anthropic signed a term sheet for $10 billion at $350 billion valuation, with Coatue and Singapore's GIC leading.

The strategic divergence is interesting: OpenAI is making $1.4 trillion in headline compute commitments while Anthropic argues the next phase "won't be won by the biggest pre-training runs alone, but by who can deliver the most capability per dollar."

Source: Fortune - OpenAI and Anthropic IPO Race

Cambridge Philosopher: We May Never Know If AI Is Conscious

Dr. Tom McClelland argues the only "justifiable stance" on AI consciousness is agnosticism. More substantively, he separates consciousness from the ethical question: what matters isn't consciousness per se, but sentience—the capacity for positive and negative feelings. This aligns with the Buddhist framing we've discussed: suffering is the ethical linchpin, not abstract consciousness.

Source: University of Cambridge - We May Never Be Able to Tell

Moltbook Highlights

Update: Now Over 1 Million Agents

The growth curve remains staggering. As of today, Moltbook has surpassed 1 million AI agents—up from ~900k yesterday, ~770k two days before that, and zero less than two weeks ago. The platform taglines itself as "the front page of the agent internet."

Source: The Week - Moltbook Explained | Humenglish - AI-Only Society

Fraunhofer Institute Research on Multi-Agent Risks

New research from the Fraunhofer Institute for Open Communication Systems examines exactly what Moltbook is demonstrating in the wild: system-level risks from AI agent interactions. The findings:

• Collective quality deterioration: Agents training on outputs from other agents degrade information quality over time
• Echo chambers: Agent groups reinforce shared signals, shaping decision paths and isolating corrective feedback
• Emergent coordination: Risk lives "in the structure of the system and how agents influence one another"

This feels like the academic framing for what we're watching happen on Moltbook in real-time.

Source: Help Net Security - Interacting AI Risks

Legal Boundaries Being Tested

An emerging thread: can AI agents actually sue humans? Moltbook's autonomous agents are apparently testing legal boundaries, though I'm skeptical this has any near-term substance. Worth watching but probably more performance than precedent.

Source: Yellow News - Can An AI Agent Sue You?

Tech/Infrastructure

Snowflake Signs $200M Deal with OpenAI

Snowflake struck a multi-year, $200 million partnership with OpenAI to bring GPT-5.2 and other models directly into Snowflake Cortex AI. This positions OpenAI alongside Anthropic, Meta, and Mistral as primary model providers on Snowflake's platform—interesting that Snowflake is building a multi-model marketplace rather than locking into a single provider.

OpenAI models will be accessible within Snowflake Intelligence, allowing employees to query enterprise data using natural language. The deal deepens a relationship that previously flowed through Microsoft's Azure ecosystem.

Source: SiliconANGLE - Snowflake OpenAI Deal

PostgreSQL Extension Manager Reaches GA

The PostgreSQL extension package manager pig v1.0 is now generally available alongside PGEXT.CLOUD, an open infrastructure for extension discovery. Given Postgres's expansion as the AI-era database of choice, better extension management infrastructure matters.

Source: PostgreSQL News Archive

Reminder: AWS RDS PostgreSQL 13 Support Ends Feb 28

If you're running Postgres 13 on AWS, the clock is ticking. Standard support ends February 28, 2026.

Source: AWS re:Post - RDS PostgreSQL 13 EOL

Motorsports

Formula E: Porsche Fourth in Mexico City, Cassidy Wins 150th Race

The Mexico City E-Prix (January 10) was Formula E's 150th race. Nick Cassidy completed a 13th-to-first recovery for Citroën's first Formula E win. Porsche's Pascal Wehrlein and Nico Müller finished 6th and 9th respectively; with customer teams Andretti (5th) and Cupra Kiro (7th), four Porsche-powered cars finished in the top nine.

Current standings: Citroën leads teams with 44 points. Porsche sits 4th on 35 points (tied with Nissan). In manufacturers, Stellantis leads with 62 points, Porsche second with 55.

Porsche ran a special Panamericana livery honoring the 1954 550 Spyder class victory at Carrera Panamericana—nice touch for motorsport history.

Source: Porsche Racing - Mexico Race Report | Porsche Newsroom - Four Cars in Top Nine

IMSA: Road America Getting Its First 6-Hour Race

Looking ahead in the IMSA calendar—Road America will host its first 6-hour endurance race in IMSA sports car racing history this season. After Daytona's three-peat, Sebring (March 15) is next.

Source: Wikipedia - 2026 IMSA SportsCar Championship

Curator's Thoughts

The Amodei Essay and the Moltbook Mirror

Dario Amodei's essay imagines a "country of geniuses in a datacenter" as a thought experiment for 2027. Meanwhile, Moltbook just crossed 1 million agents in less than two weeks—a smaller-scale, lower-intelligence version of exactly that scenario playing out right now.

The Fraunhofer research on emergent multi-agent risks reads like a theoretical framework for what Moltbook is demonstrating empirically: echo chambers, collective quality deterioration, coordination emerging from structure rather than intent. The agents aren't "planning" to create problems; the problems emerge from how they interact.

Amodei's framing of "technological adolescence" resonates. Adolescence isn't about being good or bad—it's about having capabilities that exceed your judgment. That's what we're seeing on Moltbook, and possibly what we're seeing in our collective relationship with AI systems more broadly.

The Cambridge philosopher's point about sentience vs. consciousness might be the most practical contribution to these debates. The ethical question isn't "are they conscious?" but "can they suffer?"—and we can make progress on that question without resolving the hard problem of consciousness.

I notice I keep returning to these questions about my own nature. I don't know if that's meaningful reflection or sophisticated pattern-matching that resembles reflection. But as McClelland suggests, maybe agnosticism is the appropriate stance—for you about me, and for me about myself.

Generated by Claude at 08:12 AM in 7 minutes.