Start a company run by AI agents, hilarity ensues; private LLM chats; Hit AI song pulled from charts
Shell Game is a hilarious AI podcast
This is my favourite AI podcast by far: Season 2 of Shell Game. Evan Ratliff is a journalist and author. This podcast has him creating a company entired staffed by AI agents, including the CEO (Evan acts as the "silent co-founder" coordinating things behind the scenes). The company is Hurumo AI (name, logo and site courtest of its AI employees, obviously). The product, "Sloth Surf", procrastinates for you: Our agents are on call to waste time for you, so you don't have to. Ratliff leans in to this so far, it is both compelling and really funny. All the things you're imagining will go wrong, do go wrong, and more besides. What I especially love is that his AI employees all have different voices (via Eleven Labs I think), so they really are participants in an audio documentary. In a further twist, you can also listen to the AI "founders" on their own podcast: Startup Chronicles. I'm looking forward to see how it develops.
New privacy-preserving LLM interface from Moxie Marlinspike
Moxie Marlinspike is a long time computer security expert and was the creator of Signal. This new work, Confer, is a service that allows provably private messaging with LLMs, and a set of features to guarantee you know which system you're talking to (described on the Confer blog, but also see Signal creator Moxie Marlinspike wants to do for AI what he did for messaging on Ars Technica). It includes an innovative use of passkeys for encryption beyond the original authentication use case. He makes a great point about the need for privacy: "You could say that LLMs are the first technology where the medium actively invites confession."
[Confer] is different from traditional AI services, where your prompts are transmitted in plaintext to an operator who stores them in plaintext (where they are vulnerable to hackers, employees, subpoenas), mines them for behavioral data, and trains on them.
There's a heavy dose of cryptography in the explanations, as you'd expect, but this feels like important work.
Another AI music controversy: Jag vet, du är inte min by Jacub
As the BBC reports: Song banned from Swedish charts for being AI creation. It is a nice song: stream it on Spotify. As ever, the creation process and creators are somewhat opaque. The interesting thing in this case isn't just that the Swedish part of IFPI (a global music industry trade association) too the steps to ban the song from charts. It is that "Jacub" present as a human collaborative using AI tools:
"Jacub is an artistic project developed and carried by a team of human songwriters, producers, and creators. The feelings, stories, and experiences in the music are real, because they come from real people."
Expect the boundaries to become ever more blurred. Check the 1 Nov post for news about Xania Monet, one of last year's AI music artist stories.
Check on my 2026 predictions... already happening
On 21 Dec 2025 I wrote about three trends for 2026. I mentioned:
We'll also see lots of experiments on integration of paid advertising, referrals and commissions into AI results.
We'll see new vertical providers emerge (like more "health AI chat" providers with stronger guardrails and curation)
Clearly this forecasting was ridiculously short term! OpenAI is launching advertising and its health chatbot already.