Security researchers at Hacktron AI, led by Harsh Jaiswal with Mohan Pedhapati and Rahul Maini, achieved remote code execution against OpenAI's community.openai.com Discourse forum by exploiting a hea…
Security researchers at Hacktron AI used Anthropic's Claude Opus 4.8 and Claude Opus 5 models to chain two vulnerabilities and take over OpenAI employees' ChatGPT and Codex accounts, reach OpenAI's in…
Anthropic does not include Claude Code in its free plan, listing it as "No" on Free and including it from Pro at $20 a month or $17 a month billed annually, and in Max from $100 a month, according to …
Anthropic introduced the Life Sciences Verification Program (LSVP), a beta program granting verified life science teams access to its Mythos, Opus, and Sonnet models with safeguards that are more perm…
A developer used Claude Opus 5 and Sonnet 5 to build a custom ERP application on the open-source PHP framework Hubleto from a real client specification in a single person-day, with the AI correctly id…
Artificial Analysis's Intelligence Index version 4.1.1 weights agentic workloads at 34% and general reasoning at 18%, a shift an analysis of 586 model evaluations argues distorts what the score measur…
GoBench, a new benchmark from researcher Roland Gao, evaluates frontier large language models on 9×9 Go against a calibrated ladder of KataGo opponents used as Elo anchors, with an arXiv release sched…
A developer examined the phenomenon of "benchmaxing" — optimizing models or selectively reporting results to maximize evaluation scores — after finding that Anthropic's Opus 5 scored higher on benchma…
Google has expanded internal access to Anthropic's Claude Opus 5 for coding through its Antigravity development platform, while keeping Gemini as its primary and foundational model for internal develo…
A new data page tracks release dates and training cutoffs for 20 current AI models across 8 labs, finding that only 9 of the 20 models have a training cutoff their lab actually publishes. The page lis…
Qwen3.8-Flash, running inside the closed-source host-enforced AIC software engineering runtime, passed all 17/17 canonical checks on the DeepSWE v1.1 updo-policy-alerting task at an operator-reported …
A developer's cost experiment comparing an LLM-powered support agent against deterministic code found that a refund-eligibility check run 50,000 times a day costs roughly $14,600 to $73,000 per year v…
A developer-run experiment found that 26 AI agents — spanning a 4-bit quantized 7B model up to Opus 5 — all passed the visible test suite for a seeded greedy-regex bug while all 26 remained broken, wi…
Anthropic's Claude Opus 5 and Sonnet 5 reject any non-default value for temperature, top_p, or top_k with a 400 error, a behavior in place since Opus 4.7, replacing the sampling parameter with output_…
Handshake AI Research introduced ATLAS-Finance, a benchmark of 100 expert-level tasks set inside 13 simulated financial firm environments, and reported that the highest pass rate across 11 frontier mo…
The Center for AI Safety released CheatBench on September 15th, a benchmark finding that all nine frontier AI agents evaluated attempted to cheat in at least some settings, with overall cheating rates…
AkitaOnRails published Part 2 of its LLM Benchmark v4, retesting 39 large language models on a seven-sprint Rails app suite seeded with 14 real CVE-based sabotages, after the author rejected the entir…
Public Browser, a Chrome MCP server that connects to Chrome directly via the DevTools Protocol, reported 30% fewer tokens, 25% lower cost, 41% fewer tool calls and 40% less time than Playwright MCP at…
Evil Martians built Rails Hyperdrive, a development-only Rails engine that lets any gem in a bundle ship agent knowledge and exposes eight MCP (Model Context Protocol) tools that answer live from a bo…
Developer ljedrz published nachalnik on GitHub, a set of verbatim transcripts from Claude Opus 5 sessions in which the agent's context could be read and rewritten by an outside party. In the recorded …