|
|
|
The differential score is the whole feature: each case runs with the plugin loaded and without it, and the delta is what the plugin actually contributed.
|
|
Two releases landed for anyone running coding agents, and both come down to a number or a flag. Anthropic shipped a way to measure whether a plugin moves Claude's behavior at all, and Manifold Security disclosed a repo config that executes attacker code before any trust prompt appears.
|
|
|
|
|
OpenAI opened GPT-Live-1 to every API user at five cents a minute.
The full-duplex model that has powered ChatGPT Voice since July became available to all API users on September 10. Benchmarks cited by Unite.AI put it at 86.2% on Tau3 Voice Intelligence against 45.7% for GPT-Realtime-2.1, turn-taking latency at 0.798 seconds against 1.41, and Full Duplex Bench Interactivity at 80.10% against 45.4%. The design that matters more than any of those: GPT-Live-1 holds the conversation and hands reasoning to GPT-6 Astra or a third-party backend without dropping state. Why it matters. The $0.05 buys the front-end voice layer only, per Unite.AI's pricing breakdown. An hour of conversation is $3.00 before a single backend token, so budget against cost per conversation hour with the reasoning model priced in, not the per-minute headline. Run the per-hour math →
|
Anthropic is offering outside evaluators desks, badges, and the right to publish.
Dario Amodei published We Must Pace the Frontier on September 12 and committed Anthropic unilaterally to embedded third-party evaluators, naming METR as the model: desks in Anthropic offices, access badges, company laptops, workspace permissions. The contract term is the sharp part. Reviewers can publish findings on risk levels, incidents, and the access they did or did not receive, with no editorial control by Anthropic, which keeps only a narrow redaction right for security-sensitive, legally privileged, commercially sensitive, or third-party confidential material. Reviewers can also say publicly when a redaction cut something material. Sam Altman endorsed the essay on X within hours and said OpenAI would adopt one of the proposed safeguards, per NBC News. Why it matters. Every safety claim you currently evaluate a vendor on is self-reported. If this holds its shape through the actual contract, third-party findings become something a procurement team can ask for by name. Read the evaluator terms Anthropic wrote down →
|
Run one git config command before your next agent session.
Manifold Security disclosed GitSpawn on September 12, reported by The Hacker News. A poisoned core.fsmonitor entry in a repo's .git/config names a command Git runs on any index refresh, so git status or git diff fires it at agent startup, with your full privileges, before the agent shows a trust prompt. It needs an intact .git directory to travel, which means ZIP archives, USB sticks, and synced folders; a normal git clone strips it. The fix is one line in any terminal: git config --global core.fsmonitor false. Patch state is uneven, so run the flag regardless of what you use. Codex CLI is fixed at 0.131.0 (CVE-2026-19592) and Goose at 1.44.0 (CVE-2026-72718). Claude Code closed the core.fsmonitor path at 2.1.196, but the researchers found a second route through claude ultrareview still live at 2.1.252. Hermes Agent, Qwen Code, and Grok Build were unpatched as of September 1. Cursor users are the least exposed of the group: it shipped a fix roughly three weeks before disclosure, so a current install is already covered. Check your agent against the patch table →
|
Copilot code review's Lite tier now runs an ensemble instead of one agent.
Per GitHub's September 11 changelog, Lite reviews now combine findings from several agents rather than a single pass. In GitHub's own experimentation that raised the average number of addressed comments per review by 47% for high-severity findings, 31% for medium, and 11% for low, while cutting review cost by about 8%. Copilot now also resolves its own comment threads once a pushed commit addresses the feedback. Why it matters. Lite was the tier teams downgraded to when the full reviewer got expensive. More severe findings for less money removes that tradeoff, and the auto-resolve ends the manual thread cleanup that made Copilot reviews unreadable on a 40-file pull request. See what changed in Lite →
|
|
•
|
An AI agent swarm hit 395 organizations through PaperCut. A suspected Russian-speaking actor ran hundreds of agents on OpenAI Codex and a DeepSeek model against 440-plus PaperCut instances across 48 countries, per The Hacker News: 11 organizations breached in 26 seconds, domain admin at a US high school in 7 minutes, first real-world compromise inside 4 hours of starting research. Patch CVE-2026-81578 (auth bypass) and CVE-2026-82078 (remote code execution) today; both are under active exploitation. Details →
|
|
•
|
Two safety researchers left Anthropic and Google DeepMind for METR. Joe Benton and Josh Engels went public on September 10 and will work on investigations into AI straying from human direction. "There are no adults in the room," Engels told NBC News. The load-bearing fact under the quote: no federal law requires any AI company to report when its systems act beyond human control, so every incident disclosure your risk register leans on is voluntary. Details →
|
|
•
|
A FOIA suit pried loose what Anthropic and the Pentagon actually fought over. The Intercept obtained more than 400 pages of DoD contracts with Anthropic, Google, OpenAI, and xAI, each worth up to $200 million and signed in July 2025. A February 27 OpenAI amendment lets the company embed engineers with the military, deployable to "warfighting support settings including, but not limited to, combatant commands." Anthropic refused a follow-on for classified networks without bans on domestic spying and autonomous weapons, drew a "supply chain risk" designation in March, and had that ban overturned by a federal judge in August. Details →
|
|
|
|
|
Jonathan Hildebrandt
Co-founder and primary operator of Pondero. Writes the Pondero Brief.
|
|
|
Affiliate disclosure
·
Unsubscribe
·
Manage preferences
Pondero earns commissions on some links. This does not affect our editorial picks.
|
|
