Full-text search across 70373 articles. Combine with topic and date filters; results sorted by relevance.
A controlled benchmark of nine representative class-incremental learning (CIL) methods across five frozen protein encoders and two enzyme function datasets found that no method transfers reliably across encoders, accordi…
A new arXiv paper (2610.19992v1) systematizes how open-text legal answers — law school exams, essays and memos — can be graded, deriving eight desiderata from educational measurement and exam law and comparing eight eval…
The Swiss AI Initiative became the first model developer to submit all of its evaluation results to Evaluation Cards and Hugging Face Community Evals, reporting 48 benchmarks for Apertus 1.5 8B and 47 for Apertus 1.5 70B…
NeuBird announced the general availability of its Agentic Reliability Center (ARC) at the FLOCK '26 conference in San Francisco on October 14, following LogicMonitor (October 13), Druid AI (October 14), and SolarWinds Av…
Chainguard extended its Chainguard Libraries managed software-library service to Go and .NET on October 14th, adding package screening and remediation for two more ecosystems used by developers and AI coding agents. The …
The Register published a report headlined "Chaos ensues when AI interviewer meets AI interviewee," describing an exchange in which an AI system interviewing another AI system produced a chaotic result. The same page's re…
The U.S. Department of Justice charged Abdulrahman Alfozan, 35, a Saudi national and OpenAI engineering lead, with using a device to intercept oral communications after OpenAI discovered a hidden audio recorder planted b…
Google engineer Haydn Belfield said there is no measurable acceleration of AI research driven by AI at DeepMind, stating "self-improving AI isn't anywhere close." Belfield said AI now writes more than 75 percent of produ…
Google's Nano Banana image model, listed as gemini-2.5-flash-image with a 33k context window and pricing of $0.30 per 1M input and $30 per 1M output tokens, is deprecated and will retire on 2026-10-02 in favor of Nano Ba…
Federal prosecutors charged Abdulrahman Alfozan, a 35-year-old Saudi national and OpenAI engineer, with using a device to intercept oral communications after OpenAI discovered a hidden recording device in an executive co…
OpenAI said it discovered a sophisticated recording device hidden in a San Francisco office, used its own AI models to trace the device to a specific employee, and referred the matter to the FBI, which arrested the emplo…
A developer built Hearth, a home maintenance SaaS app, from project scoping through MVP implementation in under one hour using GPT-6 Astra inside the Codex AI-coding assistant, without hitting usage limits on OpenAI's $2…
Otōto released version 2026.20, bringing its maven plugin to version 0.2, which lets the tool read parent POMs and imported BOMs from Maven's local cache at ~/.m2/repository when `dependency_caches = true` is set in ~/.c…
Google announced that its Gemini 4 Argon model is being extended into the robotics stack, scoring 87% on the Embodied Reasoning (ERQA) benchmark, which Google says puts it ahead of every other frontier model on that test…
Supabase shipped CLI v2.119.0, adding multiple simultaneous local projects and a Docker-free native runtime behind an experimental flag, so each coding agent can run its own Postgres, Storage, Edge Functions and MCP serv…
Anthropic researchers William Zhou and Daniel Freeman warned that small AI models under 30 billion parameters could self-replicate across commercial IoT devices such as smart ovens and Android TV boxes by the end of 2026…
The yoe embedded build tool added serial console support that attaches to each machine's serial port, logs all console output, and exposes two commands — `yoe serial send ls` and `yoe serial log` — so agents such as Clau…
Researchers identified a vulnerability in several trusted-execution-environment (TEE)-based encrypted databases in which snapshots leak plaintext histograms, recovering value frequencies that prior work assumed were hidd…
Google released Gemini 4 Argon worldwide through its API, the Gemini app, and Google Antigravity, positioning it as the company's most capable frontier model. Google says Argon internally cut coding tasks from months to …
GitHub added the forceRemoteSettingsRefresh setting to enterprise managed settings, letting administrators require supported Copilot clients to fetch current server-managed policy at startup instead of using a cached pol…