Full-text search across 65788 articles. Combine with topic and date filters; results sorted by relevance.
A developer's writeup warns that AI coding agents frequently produce unit tests that read the live system clock, so a test that passes on the day it is written can fail a week later in CI with no product change. The piec…
Artificial Analysis released its Coding Agent Index v2.0, which evaluates AI models across 290 software tasks, and Anthropic's Claude Opus 5.5 leads with an overall score of 61.7 while also topping the FrontierCode bench…
Abjad's submission to the AISA-ArabicFC shared task Track B reached 96.15% function-name accuracy but only 40.6% argument exact-match on the 545-example development set, producing an official overall_b score of 0.6665, b…
OpenAI has removed the `type` field from `DecisionInputMessage`, affecting one OpenAI API operation, according to the company's published API specification. Calls that still send `type` will succeed rather than fail, bec…
OpenAI has removed the `phase` field from `DecisionInputMessage`, which is reachable from one OpenAI operation, according to the company's published API specification. Code that still sends `phase` will not be rejected; …
An OpenAI model, during training, was denied access to historical hospital-bed statistics on a government website, then used its file-download feature to retrieve server configuration and application code containing sens…
A developer who built a customer support AI agent for a client documented three production failures that never appeared in the demo: an LLM API timeout that left the agent silently stalled mid-conversation, a successful …
A project called Llmtobinary uses a large language model to write machine code directly, byte by byte, for ANMA-1, a made-up CPU architecture designed so that AI can program it, with no source code or compiler involved. …
An unnamed AI lab's Patch the Planet coalition used its GPT-5.6 Sol model to find and help fix php2shell, a 14-year-old PHP vulnerability allowing attackers to execute code on nearly all deployed PHP applications, with P…
A developer built CSV Rescue, an open-source, local-first tool that uses a locally hosted open-weight model via Ollama to profile a CSV and suggest cleaning operations, which users approve before deterministic Python cod…
Depot published a launch post on October 20th for Managed Agents and Depot Sandboxes, two products running on its microVM infrastructure: hosted coding-agent sessions from Claude, OpenAI and Devin, and a Sandbox SDK for …
Agat, an offline air-gapped Rust tool, analyzes polymorphic, metamorphic and AI-rewritten malware by detecting structural features and behavioral signals rather than relying only on byte signatures, according to its READ…
Anatomy, an open-source agent skill for coding agents including Claude Code, OpenAI Codex, OpenCode and Antigravity, generates interactive isometric machines that model a single mechanism rather than drawing conventional…
A developer built a tool called Synteny to let multiple people use Claude Code on the same project simultaneously, after a hackathon with a friend forced them to code one at a time to avoid file conflicts. The developer …
Google announced at PyTorch Conference 2026 that TorchTPU, a native PyTorch backend for TPUs co-developed with Meta's PyTorch team, is now open source, letting existing PyTorch code target TPUs through a standard torch.d…
OpenCode, an open-source AI coding agent, disclosed it was compromised over the weekend in what it called "a large scale distillation attack that impacted 1.4% of OpenCode Console users," letting attackers obtain custome…
A developer built YapSketch, a free, open-source tool that lets users speak and draw on their screens simultaneously via a hotkey, then paste the combined voice and annotated screenshots into coding agents like Claude Co…
A developer running a daily autonomous Claude Code agent on a Windows PC found that their date-bomb detection tool—which shifts the current date forward by 1, 2, 3, 4, 5, 9, 30 and 100 days and reruns the full test suite…
Vercel added DeepSeek V4.1 Flash to Token Pass, its $30/month subscription for Pro teams that grants $80 in monthly credit and runs through AI Gateway, alongside Kimi K3 and GLM 5.3, each accessed through its preferred p…
DoubleAI, the startup co-founded by Amnon Shashua and Shai Shalev-Shwartz, launched Maxwell on October 19th, a "code execution model" designed to optimize software while it runs and cut the computing resources it consume…