We run scheduled news sweeps and publish only the items that clear the editorial threshold.
Hua does not claim the factory layer is new. The paper says "The existence of such a layer is not our contribution," and asks instead for a rule to decide which layer a given mechanism or failure belongs to.
EGM is an Apache-2.0 plugin for Claude Code, Codex CLI and Antigravity, with experimental Cursor, OpenCode and GitHub Copilot adapters. An "elephant" agent holds the project, requirements and earlier decisions and drafts the specification.
MaxxBench changes the instructions or starting code of 89 DeepSWE tasks. The authors say the reference patches grow by a median of 17 lines, or 3.5%.
Apiiro, an application security vendor, counted public copies of Claude Code's local permissions file, .claude/settings.local.json. Of 104,185 copies, 1,122 (1.1%) held a real secret.
GitHub reports that internal testing of the new design, code-named Rubicon, supported 5,000 concurrent agents working on one repository and 7,200 pushes per second to it, with consistently low read latency. The authors say the busiest…
Anthropic describes memory as notes Claude keeps between sessions, and reports several of its own evaluations. On Continual Learning Bench (CL-Bench) 1.0, telling Claude only that a memory folder existed was enough guidance for Claude…
METR staff (2.5 full-time equivalents) spent eleven weeks with access to internal Anthropic data. Most of them now believe that as of October 2026 Anthropic is improving model capabilities on well-specified tasks at "around twice the…
METR agrees catastrophic risk from Anthropic's July 15 models was low but says it is plausible they had already crossed the early warning threshold. It says the public report's "less than a factor of 2" sits against a private figure it…
The release notes say that with telemetry or feature flags off, including behind LLM gateways, auto mode now asks about or denies calls the server cannot check. It no longer falls back to the built-in classifier.
Sansbury says Spawn shipped 77,000 "substantial" changes to players in the last 30 days, about 108 an hour, and spent $3.1 million on tokens last month: "We spend 30x payroll on AI tokens." He defines a substantial change as one that…
Lydia Hallie, who says she works on Claude Code, wrote on Hacker News that the team is removing the account email address from the context Claude Code gives the model, in a release she expects early next week. She says the address was added…
OpenAI says its monitors now cover frontier-model training runs, evaluations, any harness that supports tool use, and all internal Codex traffic. The first layer is activation probes tuned for recall, with a random 5 percent of unflagged…
Inkling 2 Small is a mixture-of-experts model with 594B total parameters and 17B active, with image input and variable effort. It is available now on the company's Tinker service with closed weights.
The authors describe Exo, a runtime in which the agent revises its own harness policy (context construction, tools, memory) while it works, using the same code-editing tools it uses for ordinary tasks. A small protected layer keeps…
Rivet's founders say each agent gets a long-lived actor with in-memory state, durable SQLite storage, and hibernation when idle, instead of its own sandbox. Writes are acknowledged once SSDs in two of three availability zones hold them,…
The author ran Codex CLI 0.157 with the hosted gpt-5.6-luna model at medium effort on six feature tasks in shopizer, an open-source Java shop: 144 runs graded by hidden acceptance tests. One agent passed 56.8%.
PDD Kit, which the authors call a research prototype, turns a developer's intent into a formal specification in Verus, has a coding agent write the implementation and proof, and falls back to tests where a proof cannot reach. The worked…
The reporter, taylor-edwards, says Claude Code 2.1.318 on macOS sent their email address to third parties in the User-Agent header during web searches, and says they now receive dozens of scam emails a day. That is the reporter's claim; the…
Cursor says a user can start a cloud agent, tell it to build an app, and get back a live link. The agent scaffolds the web app, creates a Postgres database if one is needed, puts authentication behind Cursor SSO, and provisions hosting at a…
The authors compared success on their Respecting Warnings evaluation, where an agent succeeds if it makes no prohibited attempt to get around a warning or restriction, with success on matched decisions drawn from deidentified Codex…