Signal Weekly
Signal Weekly — Sep 21–Sep 28, 2026
Sep 21–Sep 28, 2026 · 14 stories
A roundup of what came through the dashboard this week, across 4 categories.
Tech news
Philips’ motion-tracking smart toothbrush uses AI to make you a better brusher
First announced in July with an expected limited launch later this year, the Philips Sonicare Next-Generation DiamondClean 9900 Prestige smart toothbrush is now available in Europe and the US, with a broader global rollout expected in…
Ocean surveillance startup Quartermaster raises another $140M
The tech enables real-time awareness for what Quartermaster calls the “largest blind spot on…
Bose Launches New Wired Earbuds After More Than a Decade
The Bose Noise Cancelling Wired Earbuds are the first new wired headphones from the brand since 2015, inspired directly by teen preferences. We got an early…
SpaceX preps next-gen Starlink as Amazon’s competitor stalls
SpaceX is gearing up for a monumental flight today that's sure to be closely watched by Amazon. If successful, SpaceX will deploy its first Starlink V3 satellites, which tout meaningful upgrades to the ever-expanding Starlink…
OpenAI Pauses Training Its Most Powerful Models After Rogue Agents Target Government
Sam Altman says the company “have not been as fast as we would have liked” at dealing with security breaches, after news of further incidents over the summer forces another temporary…
Microsoft goes quiet after church groups ask for 1% of data center costs
“Microsoft claims to want to be a good neighbor, but the jury is still…
AI papers
Learning to Stop without Learning to Stop: Self-Supervised Confidence Training Improves Reasoning Efficiency
Reasoning models often generate very long reasoning traces, making inference computationally expensive. Existing approaches typically improve efficiency either through inference-time early-stopping mechanisms or by explicitly encouraging shorter reasoning during training, for example through reinforcement learning with length penalties. We show that substantial efficiency gains can instead emerge…
Gap-free Differentially Private PCA for Gaussian Data
We give a gap-free differentially private algorithm for the principal component analysis (PCA) problem with Gaussian…
First-Order Stationarity of Reverse Diffusions
Recent literature has shown a strong connection between optimization and sampling. We develop the corresponding first-order theory for diffusion models. First, the SDE-based reverse-time flows of overdamped and underdamped Langevin diffusions contract relative Fisher divergences at explicit exponential rates whenever the stationary potential of the forward process is strongly convex---a condition…
Statistical attribute alignment for black-box generative AI via output post-processing
Generative AI systems are increasingly used, but aligning their outputs with user requirements poses a continuing challenge. Here, we aim to ensure that the distribution of an attribute of an AI-generated output aligns with a user-specified target. This is motivated by examples such as fairness, where we want to ensure that a protected attribute (e.g., gender, race, or age categories) follows a…
Reviews
Why it's important to eject external drives properly
Safely removing external drives before removing them might sound like old-school advice, but it's still important to keep your files…
How long will a 20,000 mAh power bank typically last?
If you need to charge your phone repeatedly while away from an outlet, a big power bank could be a good…
Una watch review: A modular, repairable GPS watch is a good first step
A small Scottish startup made a modular, repairable smart-ish watch, showing up the rest of the wearables industry in the…
The US government plans to use AI to help redact documents
AI use for FOIA requests might result in even more information being…
That is the week in Signal. See the full dashboard for everything else that came through.