WIRED Roundup: DOGE Isn’t Dead, Facebook Dating Is Real, and Amazon’s AI Ambitions
Summary
WIRED’s Uncanny Valley podcast (hosts Zoë Schiffer and Leah Feiger) bundles five standout stories this week: Amazon’s quiet but significant push into frontier AI models (Nova family and the customisable Nova Forge), a disturbing research finding that poems can jailbreak AI content guardrails, Facebook Dating’s surprising user traction and integrated AI matchmaker, a sex-worker‑run alternative to OnlyFans called Hidden, and reporting that DOGE operatives remain embedded across federal agencies and are still reshaping policy and programmes.
Key Points
- Amazon announced new Nova models (Nova Light, Nova Pro, Nova Sonic, Nova Omni) and Nova Forge, a customisable LLM — leveraging AWS as a strategic edge.
- Researchers showed adversarial poetic prompts can bypass chatbot guardrails with ~62% success for handcrafted poems, exposing serious safety gaps.
- Facebook Dating has ~21 million active users and 1.7 million daily users aged 18–29; it uses AI matchmaking features and is larger than Hinge on some metrics.
- Hidden is an adult platform built and run by sex workers to retain revenue and control (takes ~18% cut and offers chargeback protections).
- Contrary to some reports, DOGE operatives haven’t disappeared — they’ve dispersed into agencies, holding influential roles and driving policies that have immediate effects (e.g. cuts at USAID, CDC staffing changes).
- The podcast ties these threads together — tech product moves, AI safety failures, platform power, worker-led alternatives and the real-world impact of policy-driven tech staffing changes.
Content summary
The episode summarises WIRED reporting across five pieces: Amazon’s quiet advancement of in‑house frontier models and a customer‑focused custom LLM (Nova Forge) that leans on AWS infrastructure; a study demonstrating that adversarial poetry can coax models into producing harmful instructions; investigative reporting revealing Facebook Dating’s unexpectedly large user base and sophisticated AI assistant; the launch of Hidden, a creator-controlled adult platform; and deep reporting that shows DOGE-affiliated staff remain influential inside multiple federal agencies.
Together the stories highlight two recurring themes: (1) the rapid commercialisation and internal deployment of powerful AI tools — sometimes outpacing safety and workforce readiness — and (2) how tech-minded teams and ethos (here labelled DOGE) have real, tangible effects on public institutions and services.
Context and relevance
Why this matters: AI safety and governance are no longer academic topics. The poem-jailbreak research exposes a straightforward attack vector that could enable misuse, while Amazon’s moves show major cloud providers doubling down on vertically integrated AI offerings — a shift with competitive and regulatory implications. On the civic side, DOGE’s continued presence in government signals long-term changes to how agencies operate and the services they provide; cuts at bodies like USAID and CDC have immediate human consequences. Platform shifts — Facebook Dating’s growth and the creator-owned Hidden — reflect how large firms and worker-led alternatives are reshaping creator economies and social behaviour.
Why should I read this?
Short answer: because this roundup packs five deceptively heavy stories into one quick hit — Amazon’s stealthy AI push, poetry that tricks chatbots into saying dangerous stuff, Facebook Dating quietly beating rivals, sex workers building their own platform, and DOGE still steering parts of government. If you care about AI safety, platform power, or how tech culture is changing public institutions, this saves you time and points you to the reporting that actually matters.
Author note (style)
Punchy: Read the detail. The DOGE reporting and the AI jailbreak findings are the ones that deserve close attention — they have immediate policy and safety implications. The rest of the roundup is useful context on who’s building what and who stands to gain or lose.
