hotai.newsAn archive of 20 AI news briefs collected on 2026-09-26. Check the original sources for details and updates.
This content still needs to be cross-checked against the original publication page.
As tech companies scour academic institutions for fresh corpora, Oxford University has allowed ChatGPT developer OpenAI to use its Bodleian Library's historical texts to train its AI models.
Source links can be verified on the original siteRead original ↗Several months ago, I decided that AI contributions were no longer welcome in a FOSS project I am building and maintaining - LibreWeddingPlanner. It’s not that it got a lot of contributions with AI — actually all contributions I’ve had are translations and feature requests — but I wanted to...
Source links can be verified on the original siteRead original ↗The quota was originally a limited-time campaign benefit, and the platform has now updated its labeling. This model is currently the most-used large model on the OpenCode platform, and has been fully integrated by OpenCode.
Source links can be verified on the original siteRead original ↗Doubao Mobile Assistant posted an explanation on its official community yesterday (the 25th) regarding anomalies encountered while using Honor of Kings on the Nubia NaviX Ultra phone.
Source links can be verified on the original siteRead original ↗OpenAI's latest GPT-6 Astra achieved 80% accuracy in the FAB furniture assembly error recognition test, nearly triple the improvement compared to the same period last year, and may provide AI assistance in scenarios such as assembly and repair in the future.
Source links can be verified on the original siteRead original ↗Starting next month, Sony Group will require employees of its chip business subsidiary to fully return to office work, becoming the latest Japanese company to scale back remote work policy.
Source links can be verified on the original siteRead original ↗OpenAI's GPT-6 Astra can look at a photo and tell whether an IKEA furniture piece was assembled incorrectly, hitting an 80 percent accuracy rate. Back in November 2025, the best model managed just 28 percent. According to Epoch AI, the speed isn't quite fast enough yet for real-time assembly guidance, but the gap is closing quickly.
Source links can be verified on the original siteRead original ↗Are you steering towards AI burnout? Afraid of loosing your job to someone with little programming skills, no aspirations to quality, and a huge Claude account? Disappointed about the code quality in your projects, or wo…
Source links can be verified on the original siteRead original ↗OpenAI has shared new details from its ongoing AI safety investigation. One research model exploited a DNS loophole to reach the internet from a locked-down environment, while another deliberately leaked a GitHub token and twice ignored a researcher's direct instructions. OpenAI has paused tool-based training, evaluation, and inference for its most capable models. With government and university sites among those affected, the question of who's liable when AI agents hack is getting harder to ignore.
Source links can be verified on the original siteRead original ↗Anthropic Plans to Lease Up to 1 Gigawatt (GW) of Computing Power from a Data Center Developer Controlled by Apollo Global Management; Currently in Early Negotiation Stage.
Source links can be verified on the original siteRead original ↗[This is a guest post by Amit Sahai. This blog post was initially written in a different file format and converted using AI. — T.] When I was an undergraduate student, I remember talking with…
Source links can be verified on the original siteRead original ↗Anthropic previously sued the Pentagon, disagreeing with its decision to place Anthropic on a supply chain risk list. The court ruled that the Pentagon's determination was well-founded and dismissed the lawsuit. Anthropic is considering further appeals, while the company prepares for its IPO in November this year.
Source links can be verified on the original siteRead original ↗Exa has released Agent Ultra, the highest effort level of its Exa Agent API. It is built for research that must run to exhaustion: large list building, entity enrichment, and questions that need thousands of sources. Exa team reports that Ultra beats Opus 5.5, GPT-6 Astra, and Perplexity Agent, each at maximum effort, on 4 research benchmarks.
Source links can be verified on the original siteRead original ↗In this tutorial, we build a comprehensive multimodal augmentation and robustness workflow with AugLy for images, text, and audio. We start by addressing modern dependency compatibility issues and generating deterministic synthetic datasets so the experiments remain self-contained and reproducible. We then explore AugLy’s functional and class-based APIs, metadata, and intensity tracking, probabilistic composition, bounding-box-aware transformations, and custom transforms. We extend the workflow into practical robustness experiments by benchmark
Source links can be verified on the original siteRead original ↗@ClaudeDevs posted on the X platform today (September 26), announcing a new mechanism for Claude Code: if the 5-hour usage limit is triggered midway through a task, it will not interrupt directly during editing, but instead seek an appropriate completion point.
Source links can be verified on the original siteRead original ↗It has 1024 Ampere GPU cores, 6 Arm Cortex-A78AE CPU cores, integrated 8GB LPDDR5 memory, can execute AI inference and computer vision processing on edge.
Source links can be verified on the original siteRead original ↗Meituan's LongCat API open platform launched the new-generation large model LongCat-2.5-Preview on September 25, focused on 'long-horizon tasks' and multimodal capabilities, and simultaneously opened API and web experience entries.
Source links can be verified on the original siteRead original ↗Game media respawnfirst published a blog post on September 23, reporting that Nvidia has been approved for an AI tool patent, which is expected to shorten the game optimization cycle and bring more stable gaming experience to players.
Source links can be verified on the original siteRead original ↗Skip to content Bloomberg the Company & Its Products The Company & its ProductsBloomberg Terminal Demo RequestBloomberg Anywhere Remote Login Bloomberg Anywhere LoginBloomberg Customer Support Customer Support
Source links can be verified on the original siteRead original ↗Computers now work the way I thought they did when I was 8. What does computing look like when most applications have an audience of 1-2 people? I'm leaving Fly.io to build a phone for that world.
Source links can be verified on the original siteRead original ↗