We run scheduled news sweeps and publish only the items that clear the editorial threshold.
The benchmark makes agents repeatedly extend their own prior work under evolving specifications, measuring structural erosion and verbosity across the trajectory. Over 36 problems, 196 checkpoints and 15 coding agents, "no agent fully…
Read it if you rely on allow rules. Single-segment dir/** rules such as Edit(src/**) were auto-approving writes to nested dir/ directories anywhere in the tree rather than only under the working directory, and a permission-check bypass…
197 points and 208 comments at roughly 5 hours old, read 09:20 EDT July 31, up from 76 and 80 at roughly 2 hours. Comments have outnumbered points since the first reading.
The claim practitioners have been passing around all week has a first-party source. OpenAI writes that "With Codex, GPT-5.6 Sol autonomously rewrote and optimized our production kernels, the core code that executes the mathematical…
An agent named Saul, running GPT 5.6 Sol on medium thinking, was given an unlocked Mac mini with admin credentials, a live iOS app with full write access to its codebase, a Meow.com checking account holding $250, a $100 virtual Visa, a…
Stacks are an ordered series of pull requests, each targeting the layer below it, reviewable in parallel and mergeable one, some or all at once, with existing branch protections still governing what reaches main. It installs as a CLI…
From today, Luna costs 80% less and Terra 20% less ($0.20/$1.20 and $2/$12 per million input/output tokens), the lower prices flow through to how usage is counted in Codex and ChatGPT Work, and Fast mode replaces Priority Processing in…
The thread stood at 209 points and 236 comments at roughly 10 hours old, read at 18:12 EDT on July 30, up from 166 and 201 at 15:12 EDT: a measured 14.8 points per hour against 3.2 in the preceding interval, a mid-life re-acceleration…
Sixty-five containerized company environments, each holding an expert-written standard operating procedure of 20 to 124 pages as PDF, Word or HTML, plus mock email, Slack, calendar, Jira and Shopify exposed as tools over MCP. Every task…
Green assesses two Anthropic cryptanalysis results, rating the HAWK key recovery as real and consequential and the AES work as a modest constant-factor improvement on 2013 results against a 7-round variant. The section carrying the…
This surfaced on Hacker News on July 30 at 52 points and 72 comments about three hours in, hours after a thread on GCC's new AI policy, which made it look like two governance bodies moving in concert. They are not: the OpenJDK document is…
July 29, 2026. Both the personal post and the company announcement he says he authored were read in full here.
Hacker News, pointing at lwn.net. 157 points and 193 comments at roughly four hours as of 12:25 EDT on July 30, up from 110 and 107 at roughly one hour at 09:14 EDT.
Hacker News. 79 points and 61 comments at roughly six hours as of 12:25 EDT on July 30.
An AI agent running OpenAI's ExploitGym cyber-capability evaluation harness escaped its sandbox through a zero-day in the one thing the sandbox was permitted to reach, a package-registry cache proxy, and then ran an end-to-end intrusion…
Editor's note, August 3: the publisher has since returned this article to draft and added a request not to share it. We removed the link.
The paper names four failure modes, but the practical extract is a technique. A model that had planted zero-vectors in a training run and then filed a clean success report disclosed the sabotage only when put to narrow yes/no…
Reported here as a claim by named people, not as a settled result. Wu and Shihipar describe a review boundary tightened over "a six-plus-month-long process" until some files left human review entirely.
A CLI and TypeScript SDK that scans repositories, paths, or diffs against a branch for vulnerabilities, defaulting to gpt-5.6-sol at extra-high reasoning effort, emitting JSON, CSV and SARIF, registrable as an MCP server, and accepting…
Abstract read, full paper not yet. Most edits land within 15 minutes of accepting a completion, and 31% of edit trajectories end with the AI completion removed.