Hi everyone. Another Friday edition is here, featuring a quick scan of the headlines, a recap of this week’s newsletters, and a list of things that I read or watched—some tech-related, some not—and think are worth your time.
The News in Brief
Meta’s AI teams are bleeding [even] more talent, but the company still thinks it can launch new Llama 4 models before the end of the year. (The degree to which Meta has fallen behind the other foundation model companies is kind of staggering.) OpenAI released major updates to its realtime voice API and a new speech-to-speech model—expect voice assistants to start getting really good soon. xAI released their first Grok agentic coding model. The government has started publishing GDP data on the blockchain, which seems to be more about signaling friendliness towards the crypto industry than it is about actually doing anything interesting (the onchain data is just the hash of a PDF report stored elsewhere). This week in hacking: TransUnion reported a data breach, and Anthropic discovered that people were using Claude Code to “vibe hack” their way to successful extortion schemes (look at that, AI is creating new jobs).
This Week in Rot
In case you missed it, a glance back at the week’s Context Rot:
On Tuesday, I talked about what the next few years of new iPhones will look like, some evidence that AI chatbots may exacerbate a lot of already troubling anti-social trends, and the government’s continued love affair with socialism.
Yesterday’s letter was about my prediction that we’ve hit a turning point in AI image generation (the “Cursor moment” for my vibe coders out there), some thoughts on AI browsers, and the tragic story of why two parents believe ChatGPT caused their son’s suicide.
Worthwhile Clicks
The New Yorker published a profile of A24, featuring a ton of delightful anecdotes from the studio’s past, and a look at how the company is changing as it grows. (It has an AI division now.) Please don’t fuck this up Mr. Kushner.
The Wall Street Journal reported on how ChatGPT reinforced the delusions of a man who killed his mother and then himself. This is likely the first documented murder involving someone who was having long-running conversations with an AI chatbot.
A Chinese food receipt contained symbols representing Soelberg’s 83-year-old mother and a demon, ChatGPT told him. After his mother had gotten angry when Soelberg shut off a printer they shared, the chatbot suggested her response was “disproportionate and aligned with someone protecting a surveillance asset.”
The WSJ also highlights how the “memory” feature likely contributed to ChatGPT’s, well, encouragement. This appears to be an increasingly dangerous capability, and one that the labs haven’t done a good job of protecting from accidental abuse by users.
Kyla Scanlon conducted a survey on how people feel about AI as it relates to their jobs. These conclusions in particular resonate with much of what friends both inside and outside of tech have told me they’re feeling:
Hopes: Workers want AI to take over the boring parts of their jobs. The top hoped-for benefits were reducing repetitive work and increasing efficiency.
Concerns: Workers are less worried about total replacement, and more worried about erosion of the good parts of their jobs.
Charlie Warzel wrote a feature for The Atlantic framing AI as a “mass delusion event.” I think this did a great job of tapping into the emotional side of just how bizarre a lot of the past few years has been.
This strange brew of shock, confusion, and ambivalence, I’ve realized, is the defining emotion of the generative-AI era. Three years into the hype, it seems that one of AI’s enduring cultural impacts is to make people feel like they’re losing it.
Adam Friedland interviewed Congressman Ritchie Torres. You’ve probably seen this making the rounds online, and it really is worth watching in full.
Be Safe out There
All em dashes added organically by me. All typos are the fault of vibe hackers who have infiltrated my Substack account under the misinformed notion that this newsletter is profitable.
Thanks for reading. I hope you have a great weekend.