Updated daily175 verified stories
New frontier, open-weight, multimodal, reasoning, and specialist AI models—plus the evidence behind their capability claims.
Every story links to the strongest available evidence: original sources when available, otherwise clearly attributed reporting.
What happened, why it matters, and what to watch — no jargon tax.
When the signal is thin, we publish nothing rather than padding the feed.
A growing stream of verified perspectives for people who need to understand AI without chasing hype.
An arXiv study combines two vision models with open-weight multimodal language models to resolve conflicting evidence in plant-disease images. The approach improved PlantDoc accuracy from 63.9% to 68.5% with Gemma, while tests on two Cornell robot-acquired datasets reached 98.9% and 99.3% accuracy. The researchers…
A revised arXiv paper proposes TSDS, a framework that stops local reasoning when an action stabilizes and sends uncertain actions to a cloud model. The authors report 43%-65% lower per-episode thinking compute on three of four tested tasks while maintaining stated guarantees on expected reward and cloud-call rate.
Apple Machine Learning Research describes an integrated pipeline that pretrains an enlarged language model before structurally pruning and recovering it, reporting stronger results when compressing 2.8B-parameter models to 1.3B parameters.
Google says Gemini Live can now handle multi-step tasks across Gmail, Calendar, Docs, Sheets, Drive and other connected services through voice commands, with some features requiring paid Google AI subscriptions.
TechRepublic reports that Google Cloud launched Gemini Enterprise for Legal, a preview platform that uses AI agents for contract review, legal research, regulatory monitoring and related legal workflows.
Google Research says GlucoFM improved prediction across metabolic-health tasks and post-meal glucose responses, while acknowledging that its training population remains modest and its real-time performance is untested.
A Thai research team adapted Ai2’s open data-curation toolkit to create Mangosteen, a 47-billion-token corpus that the team says improved Thai-language model performance while removing much of the underlying web data.
Google is introducing spending caps, savings plans and a forthcoming usage-based option for some Gemini Enterprise customers as agentic workloads make AI costs harder to predict, MediaPost reports.
FourWeekMBA reports that Alibaba’s Qwen and Z.ai published new open-weight models on Hugging Face on August 26. The article says both use sparse mixture-of-experts designs, while performance and pricing comparisons remain vendor claims that have not been independently verified.
Google’s developer documentation details how Gemini 3.5 Transcribe handles audio files, including automatic language detection, speaker labeling, word timestamps and cleaned-up “smart” transcripts. The page also documents practical limits, including shorter maximum audio duration when diarization or timestamps are…
TechCrunch reports that robotics companies are improving physical capabilities faster than they are developing reliable, commercially useful robot intelligence. Developers are focusing on better data, simulation, reinforcement learning and task-specific deployments while searching for a breakthrough comparable to…
A new preprint studies agents that can transfer control to a stronger language model during reasoning, combining a Bayesian stopping rule with theoretical guarantees and a limited Qwen2.5-Coder validation.
One useful briefing each week
Get the week’s verified AI news, original data, useful tools, learning picks, and fresh AI jobs.
Hiring an AI professional or launching a useful AI product? Put it in front of people who came here to learn and act.
Post an AI job Submit an AI tool