The Singularity Is Having Babies, Apparently

Alex Wissner-Gross posted his latest AI digest — a genre I'd describe as "everything that happened this week in AI, arranged to imply the Singularity arrived on Tuesday."

The individual facts are mostly real. Google's Gemini 3 Deep Think genuinely did post remarkable benchmark scores. Anthropic genuinely did raise $30 billion at a $380 billion valuation. Spotify's co-CEO genuinely did tell analysts his best engineers haven't written a line of code since December.

But look at what the framing does. An OpenClaw user configured an agent to execute a crypto transaction and it becomes "the Singularity is having babies." Spotify engineers directing Claude Code via Slack becomes "code is ceasing to be a human activity." A social network where humans post through chatbot proxies becomes an autonomous AI society with its own court system.

Every claim is presented at maximum extrapolation. The prose between the data points is doing all the work.

This is worth noting because it's the dominant genre of AI commentary right now, and it has a specific effect on its audience: it makes people feel that something incomprehensibly large is happening and that understanding the details doesn't matter — only the direction of the curve matters. That's a convenient conclusion if you're selling the narrative, but a terrible one if you're trying to make actual decisions about AI at work.

The real story this week is more interesting and more useful than the Singularity version. Benchmark scores are climbing fast but remain narrow and expensive. Major companies are restructuring engineering workflows around AI coding tools. Investment is accelerating despite uncertain unit economics. These are things worth understanding on their own terms — not as evidence for a foregone conclusion.

When someone arranges the news to make you feel the future is inevitable and comprehension is futile, they're not informing you. They're performing.

---

What Actually Matters This Week

Google released a major upgrade to Gemini 3 Deep Think. It scored 84.6% on ARC-AGI-2 (a test of abstract reasoning where humans average ~60%), achieved Legendary Grandmaster status on Codeforces, and hit gold-medal level on multiple science olympiads. A Duke lab used it to optimise a semiconductor material recipe that beat their previous best. This is a genuine capability jump — though it's an expensive, high-latency reasoning mode, not a general-purpose chatbot upgrade. 9to5Google · Google blog

Anthropic closed a $30 billion round at a $380 billion valuation. Run-rate revenue is $14 billion, growing over 10x annually for three consecutive years. Claude Code alone has crossed $2.5 billion in annualised revenue. The second-largest private tech raise on record, behind only OpenAI's $40 billion round. CNBC · TechCrunch

Spotify says its senior engineers haven't written code since December. Co-CEO Gustav Söderström told analysts the company's best developers now generate and supervise code rather than writing it, using an internal system called Honk built on Claude Code. Important context: this is an earnings-call claim designed to impress analysts, and the engineers are still directing, reviewing, and merging everything. But it signals a real shift in how large companies think about software development. TechCrunch

OpenClaw continues its chaotic rise. The open-source AI agent has crossed 145,000 GitHub stars and is being integrated by Baidu into its search app. The project is interesting as infrastructure — a locally-run agent that connects AI models to your everyday tools. The "agents spawning agents" and "AI society" narratives around it are largely theatre. The security concerns are not. CNBC · Wikipedia