👀 The witness may not take the stand
A New Mexico lawyer used ChatGPT to summarize a murder trial, then filed its version without checking it. Unfortunately, the AI had invented four witnesses, fabricated police testimony and changed important details. The state Supreme Court held him in contempt and imposed a $5,000 fine. Today’s reminder: AI can draft the work, but apparently it cannot feel embarrassed on your behalf.
📱 We all want an assistant
Meta’s personal AI agent has been downloaded over 83,000 times on iOS, ahead of Threads, WhatsApp, and Facebook, according to US data from Sensor Tower. It now ranks #2 in the App Store’s free apps section, only behind ChatGPT. Chief AI officer Alexandr Wang credited the agent’s performance so far to Meta’s newest family of models, also called Muse.
📵 To algorithm or not to algorithm
Australia has proposed rules requiring social platforms to let users choose between an algorithmic feed and posts only from people and creators they follow. The legislation is still a proposal, but it addresses a strangely radical idea: perhaps users should decide what appears on their own screens (what a silly notion…). The EU already requires a similar choice, although platforms have not exactly placed the button under a flashing neon sign.
😴 But did I really sleep?
A study involving more than 2,000 people found that wearable data explained only a small portion of how participants rated their sleep quality. Trackers can still reveal useful patterns in duration and consistency, but they may mistake lying still for sleeping - and perform less reliably for people with insomnia or health conditions. Treat the nightly score as a clue, not a medical verdict. Personally, I tracked my sleep for a week, realized I felt better every morning than what my watch told me, so I ditched it.
📺 They don’t call them smart for nothing
Researchers found LG TVs scanning home networks and logging more information than many owners would expect, including data connected to voice features and viewing habits. LG disputes parts of the investigation, saying ambient conversations aren’t continuously recorded and features such as voice recognition and Automatic Content Recognition require consent (uhm, default opt-in?). Either way, it’s worth reviewing your TV’s privacy settings. Because apparently watching television is now a group activity, and the television is in the group.
⌚ Actually, you did say that
Apple’s new Watch Series 12 includes Live Rewind, which can show a transcript of the previous 15 seconds of audio, and Siri Recap, which creates summaries of conversations. Apple says the processing happens on-device and both features must be activated; they’ll launch in beta later this year. It could be genuinely useful when you miss a detail. Or mildly alarming when it remembers the meeting better than you do.
♾️ Two hours should do it
As part of an up to $18 billion settlement with 29 US states, Meta agreed to a two hour daily cap on Instagram and Facebook, cumulative across both apps and only liftable by a parent, plus a midnight to 6am Night Mode and a School Mode that mutes notifications during class hours. The consent judgment was entered on 26 August, but as of the most recent status update no feature has launched and Meta hasn't announced start dates, only deadlines: the first, a periodic break prompt, lands around late December 2026, with the two hour limit and Night Mode required by around late February 2027. Meta is also daring TikTok and YouTube to match it (though probably not for altruistic reasons).
Know someone who would find this information useful, nay, indispensable? Forward the email to them! If someone forwarded this to you, subscribe at tooproductive.com
Till next time,
Jana

