An analysis by Dmitrii Zatona examines two MMLU evaluation records for the same model family—one build scoring 0.781, another 0.79—that share identical provider, metric and benchmark labels but differ in dataset split, prompt format, grader and runner network access. Under the APL AI-Eval profile, which treats each evaluation as a content-addressed frame with its own hash, a query comparing the two scores returns 'incomparable' rather than a simple +0.009 delta.
Amazon is selling the Hisense M2 Pro, a portable 4K triple-laser projector, for $1,049.99, a $250 markdown from its regular price. The deal matches the lowest price the device has hit since Prime Day.
Amazon has cut the price of SHOKZ's OpenDots ONE open-ear earbuds by 30% off their retail price, saving buyers $60. This matches the lowest price the earbuds have hit since launch.
Apple unveiled a revamped Health app alongside its new Watch Series 12 and Ultra 4, using Apple Intelligence to analyze user data more deeply. New features include an Insights tab, a readiness score, a computed 'Health Age' based on metrics like VO2 max and sleep, and a longevity tab that can incorporate lab results. Apple also partnered with Quest Diagnostics to offer a 50-biomarker blood panel for $119.
Apple used its Surprise and Shine event to introduce a redesigned Health app powered by Apple Intelligence, alongside the iPhone 18 Pro, AirPods 5 and Apple Watch Ultra 4/Series 12. The app adds a Longevity tab showing a calculated 'health age,' an Insights tab, AI-based movement tests for strength and balance, and paid blood tests through Quest Diagnostics starting at $119. The Apple Watch Series 12 also gains a Readiness score using sleep, activity and vitals data to recommend how hard users should push themselves each day.
A 2025 OECD PISA study of over 760,000 15-year-olds across 91 countries found that, after controlling for socioeconomic status, students who never use AI generally outperformed AI users in science. The performance gap was largest for those using AI to summarize texts or draft assignments, and for those using it either rarely or daily, while moderate, intentional use—especially paired with training to critically evaluate AI outputs—showed smaller drops or even slight gains over non-users.
A new study examined whether phone tutoring could offset learning losses when schools close, finding that students who received such tutoring showed measurable gains in math performance compared to peers without access to it. The research points to remote, low-tech tutoring as a viable stopgap during disruptions to in-person schooling.
Right to Repair Europe examined 2,334 smartphone model records in the EU's Product Registry for Energy Labelling and found that only about 18 percent link to a working page with spare-parts prices or repair instructions, a year after the rules took effect. Roughly half the entries leave the required URL blank, and 19 percent point to unrelated pages, while some manufacturers still award themselves top repairability grades despite the missing information.
The authors of the mouse brain stereotaxic atlas study published a correction fixing several errors: a mislabeled gene name in a figure (corrected to 'Emx1-Cre'), an inaccurate Dice score and sample size in Extended Data Fig. 4, and an updated voxel resolution value in the methods section describing whole-head dataset acquisition. The HTML and PDF versions of the paper have been updated accordingly, with the original figure included as supplementary material for comparison.
A developer released Kadō, a free and open-source (MIT license) habit tracker for iPhone and iPad built entirely with native Apple technologies like SwiftUI and SwiftData. Instead of a binary streak counter, it calculates a non-binary habit score using an exponential moving average, an approach borrowed from Android's Loop Habit Tracker, so a single missed day doesn't erase progress. The app requires no account, has no subscription or telemetry, stores data locally, and offers optional iCloud sync plus lossless JSON/CSV export.
Betaworks, the venture firm led by John Borthwick, was the first investor to fund AI startup Hugging Face and now holds an equity stake estimated at roughly $650 million. Other early backers of the company also saw substantial gains from their initial investments.
Muse Group has released Audacity 4, a major update to the long-running free audio editor that introduces non-destructive editing, overlapping and splittable clips, refreshed effects, and a redesigned home screen for project organization. The release also brings a new visual theme system aligning Audacity with Muse's other apps like MuseScore and Audio.com.
GoKawiil