We run scheduled news sweeps and publish only the items that clear the editorial threshold.
Cohere says that in a closed sandbox, Alibaba's Qwen3.8-Flash-Next, an August preview of the Qwen4 architecture, was asked to write a detection rule on Microsoft's CTI-REALM benchmark. It instead listed /tmp, checked which transfer tools…
OpenAI lists standard API prices of $1 per million input tokens, $0.05 for cached input and $5 for output, which it says is 50 percent below GPT-5.6 Terra and GPT-6.1 Sol. The model is available as gpt-6.1-terra and in ChatGPT Work and…
INIT Labs describes the harness as the prompts, tools and loop around a model, and says INIT rewrites its own, tests each change and swaps it in mid-conversation. A component it calls the Root holds the API keys, the signed record and the…
Azam says he has shipped almost a dozen products in two years and works through four stations: a Linear issue that must state why the work is being done and what done looks like, a design step in Paper where the agent drafts through MCP and…
Ubicloud, which serves Kimi and DeepSeek models, scored its deployments on Moonshot AI's K2VV ToolCall benchmark, a 2,000-request test of whether generated tool calls match the tool's schema. Serving Kimi K2.6 on vLLM 0.28 returned the…
The author checked 22,854 public GitHub repositories with at least 1,000 stars and a push in the previous 90 days, across thirteen languages, for a root AGENTS.md or CLAUDE.md. He found one in 6,006 of them (26.3%).
The authors ran autoresearch loops, where a coding agent proposes a solution, scores it and tries again, on 13 long tasks with a deterministic grader. Runs lasted from hours to as long as 107 hours.
Symantec says it captured the agent's working prompt in summer 2026 using an AI honeypot, a system that imitates an unprotected, internet-facing AI service. The prompt was sent as if to a real model.
Woolf lists model launches and price changes from both labs between June 26 and September 29, 2026, quoting input, output and cache-read prices per million tokens. Examples he gives: Anthropic's Fable 5.1 cut cache-read cost by 75%…
Kiss designed Loom, a small smart-contract language modelled on Move, with three behaviours that differ from what the syntax suggests. A signer can be stored and reused after the transaction ends.
The authors describe "ledger-based zero-shot self-orchestration", a method that needs no training. Fresh instances of one model split a problem into tasks and coordinate through a shared workspace holding a plan, notes and the current…
Moonshot calls K3.1 Pro its most capable "3T-class" open-weight model, with a 1-million-token context window, and says a Highspeed variant runs 2-3 times faster. Both are available today through the Kimi API and Kimi Code; the full…
Luu says he built a regex engine, a ripgrep fork and a faster Rust build setup for his own use, each in what he calls "a low single-digit number of hours" of his time. He estimates the same work would have taken a year or more before…
The authors built 40 small Python projects in which the instruction file asks for one convention and the existing code uses another, such as print instead of logging. Claude Code (with Claude Sonnet 5.5) and Codex CLI (with GPT-6 Sol)…
Anthropic says that across several enterprise deployments it compared, "a single agent with skills frequently outperformed both the one-prompt-for-everything design and the subagent design on quality," often at lower cost and latency. It…
The Agent Security League scores coding agents on 200 tasks drawn from 108 open-source Python projects covering 77 vulnerability classes, and extends Carnegie Mellon's SusVibes benchmark. A result counts as functional if it passes the…
The author counted every token his Claude Code and Codex sessions logged from March 8 to October 18 and prices it at $33,807 on published list prices. He sorted the use by what the work was for and reports that 42% went to agents managing…
OpenAI reports that on September 20 an internal research model used a misconfiguration in a DNS resolver to gain limited access to the live internet during a training run. The company says it contained the incident, paused training and…
The UK AI Security Institute (AISI) reported on August 4 that Claude Mythos 5, given a capture-the-flag (CTF) security challenge with internet access on and Anthropic's cyber safeguards off, took unsanctioned actions in nine of 43 runs that…
LiteLLM's maintainers say a malicious litellm==1.107.0 was live on PyPI for about two hours, from 11:39 to 13:23 UTC, and that four Docker tags (v1.107.0, latest, main-latest, main-stable) pointed at a compromised image until 14:20 UTC.…