An Agentic Robotics Fable
Agentic policies that write and iterate on their own Python skills reached a 34% to 65% overall task success rate across ten bimanual tasks, beating end-to-end policies trained on 30 times more demonstrations on 8 of 10 …
Agentic policies that write and iterate on their own Python skills reached a 34% to 65% overall task success rate across ten bimanual tasks, beating end-to-end policies trained on 30 times more demonstrations on 8 of 10 …
Apple released the Siri AI Developer Preview in the 27.3 beta releases of iOS, iPadOS, and macOS, letting developers integrate apps with the assistant through App Intents. The preview adds personal context understanding,…
AMD engineers Jeremy Fowers and Ben Vanik used AI coding agents to port and optimize 51 trending large language models across 19 architectures to Ryzen AI Max "Strix Halo" client GPUs in roughly four days, between Novemb…
Replit gave its Replit Agent two new capabilities: agentic commerce, letting the agent buy business supplies through Stripe's Link network (more than 300 million users globally) and Visa with founder approval before any …
OpenAI launched ChatGPT Wallet, a payment feature that lets its "dots" agents make purchases on a user's behalf without exposing actual card details, in partnership with American Express, MasterCard and Visa. The wallet …
A paper submitted to arXiv on 12 November 2026 introduces CoReMem, a cost-internalizing memory framework for coding agents that raises a 27B actor's resolve rate on the ChainSWE benchmark from 21.0% to 25.7%. CoReMem use…
A failure in Exchange Online that began shortly before 1 p.m. ET on Thursday, November 12, and lasted more than 20 hours spread to every Microsoft 365 feature that reads mail, calendars or labels, including Copilot, Team…
Google released Gemini 4 Hydrogen on November 12, 2026, priced at $1.50 per million input tokens and $7.50 per million output tokens, and Artificial Analysis measured it at $3.16 per task versus $4.97 for Gemini 4 Argon …
A paper submitted to arXiv on 12 Nov 2026 presents a PLPG-inspired continuous-action neurosymbolic reinforcement learning method that cuts violations of a six-rule regime-conditional investment mandate to 4.8%, down from…
A paper submitted to arXiv on 12 Nov 2026 introduces Agentic Semi-Supervised Learning (AgentSSL), framing semi-supervised learning as a program search problem for machine-learning engineering agents. Under weak feedback …
A developer behind the Nacre project described building and then removing an ingest-time permission-caching scheme for MCP search, in which each chunk carried hashes of the principals allowed to read it, because the copi…
Anthropic resolved an outage that degraded claude.ai, the Claude API and Claude Code on Nov 13, with the incident first flagged at 13:09 UTC and marked resolved at 15:38 UTC. During recovery, Anthropic said existing conn…
Deepnoodle-ai released decide v0.5.0, an open-source decision-model CLI that splits codebases, diffs, JSONL, CSV, Markdown and folders of images or PDFs into individual items and asks a decision model a yes/no question a…
OpenAI and Databricks outlined three ways enterprises can run OpenAI's frontier models on data governed in Databricks, including a ChatGPT plugin that queries Databricks Genie via the Genie MCP and a Unity Gateway that l…
Microsoft's Exchange Online began rejecting connections at 16:17 UTC on November 12, an incident logged as EX1550543, with outage trackers recording reports from 74 countries within roughly an hour and Downdetector loggi…
Anthropic's Claude Microsoft 365 connector and Claude for Outlook were failing for most requests after a Microsoft 365 service incident that began around 17:30 UTC on November 12, Anthropic said in a 22:23 UTC status upd…
Inkdrop developer Takuya moved 39 repositories into a single pnpm monorepo to remove the release bottleneck that slowed his agentic coding workflow with Claude Code. Takuya used git filter-repo to import each repository …
Nvidia joined AMD, Intel, Google, Meta, and Cloudflare as an inaugural member of the SGLang Consortium, a project from nonprofit LMSYS Org built around its full-stack SGLang inference engine for frontier AI models. Conso…
Sunny Chokshi, a PhD student at the University of the Cumberlands, released CommentInject, a dataset of 30 vulnerable Python functions spanning 14 CWE categories paired with 12 adversarial comment strategies across four …
A developer writing under the AgentRisk project argues that teams deploying AI agents with refund authority should cap total refunds per order rather than per action, since multiple agents or repeated runs on the same or…