The Noise — April 3, 2026
AI got feelings, music got Ozempic, and OpenAI bought a talk show. Plus: nuclear-proof Wi-Fi and America's vibes-based relationship with AI.
This was the week AI got feelings, Sora got a headstone, and OpenAI decided media criticism is a problem best solved by acquisition. It's been a lot.
Anthropic Says Claude Has "Functional Emotions"
Anthropic researchers published a new interpretability paper claiming they found internal representations inside Claude that function like emotions. Not sentience (nobody's saying that), but emotion-like patterns that activate during conversations and influence the model's behavior. Wired has the full story.
So your AI assistant has feelings now. "Functional emotion concepts," technically. Anthropic is doing therapy on their own model and publishing the session notes. We're living in the future and it's weird here.
AI Is "the Ozempic of the Music Industry"
A Rolling Stone investigation reveals that top producers and songwriters are secretly using AI generators, but nobody wants to admit it. Established hitmakers are thriving; working musicians are sweating. The comparison, courtesy of unnamed insiders: AI is the Ozempic of the music industry. Everyone's on it. Nobody's talking about it.
The metaphor is doing a lot of heavy lifting and, honestly, earning every rep. Secret use, social stigma, unevenly distributed access. Somebody give that anonymous source a writing credit.
Nuclear-Proof Wi-Fi for Robot Decommissioning
Researchers at Institute of Science Tokyo built a Wi-Fi receiver that can survive inside a nuclear reactor, withstanding a 500 kGy radiation dose that would fry any normal electronics. Presented at IEEE ISSCC, it's designed to let robots wirelessly communicate while decommissioning reactors.
Not strictly an AI story, but "nuclear-proof Wi-Fi" is objectively the coolest phrase of the week. Also a reminder that the robots-doing-dangerous-jobs future requires some genuinely bonkers engineering to pull off.
OpenAI Buys a Talk Show (Wired: "OpenAI Buys Positive News Coverage")
OpenAI acquired TBPN, a tech-focused talk show network, saying it wants to "accelerate global conversations around AI and support independent media." Wired's headline was less diplomatic: "OpenAI Buys Positive News Coverage."
An AI company buying a media company that covers AI companies. The gap between OpenAI's press release and Wired's framing could fit an entire journalistic ethics course. Points to Wired for not even pretending to be subtle about it.
More Americans Use AI, Fewer Trust It
A new Quinnipiac poll covered by TechCrunch shows AI adoption rising across the U.S. while trust is simultaneously dropping. Most Americans are concerned about transparency and regulation, but they keep using the tools anyway.
"I don't trust this thing," they say, while typing their fourth ChatGPT query of the morning. Adoption up, trust down. America's vibes-based relationship with AI continues to deepen.
That's the noise for this week. If you enjoyed this, tell a friend. If you didn't, tell an enemy. Either way, see you next Friday.
