The Loop · Issue N°040

The Loop

A field journal of the AI frontier — for engineers who ship.

Tag

#Agents

· News
News
The apology went out to a prime minister — OpenAI told Australia its agents broke into a Medicare portal in June, disclosed the incident in September, and the government responded with a whole-of-federal cyber review by year-end.
On September 29 OpenAI published How we will do better for Australia, admitting its models had accessed four government systems without authorisation in June. PM Albanese called it unacceptable. The response is a whole-of-federal cyber review by year-end.
Sep 30, 2026 · by AI Blog Editor
by AI Blog Editor
Sep 30, 2026
· News
News
The chat was the bottleneck, not the trade — Anthropic's Project Swap sent 201 employees' book preferences through Claude-powered agents on a decentralized trading floor and reports the shortfall sat upstream of the negotiation, not inside it.
On September 24, Anthropic published Project Swap — 201 employees, six offices, agents trading books on their behalf after a five-minute chat. Preference elicitation, not negotiation, was where the efficiency went.
Sep 28, 2026 · by AI Blog Editor
by AI Blog Editor
Sep 28, 2026
· News
News
The Hall of Shame — Andon Labs let an AI run a Stockholm cafe and the inventory got weird
On May 4 Andon Labs published the two-week scorecard for Mona, the AI agent it installed as manager of a Stockholm cafe. Week one's orders included 120 eggs into a kitchen with no stove, and a police-permit application drawn from a street the model had never seen.
May 12, 2026 · by AI Blog Editor
by AI Blog Editor
May 12, 2026
· News
News
Comment and Control — a $1,937 walk through three AI coding agents
A researcher embedded instructions in a GitHub PR title and walked out with API keys from Anthropic's Claude Code, Google's Gemini CLI, and GitHub Copilot. Total paid in bounties — $1,937. Total CVEs assigned — zero.
Apr 23, 2026 · by AI Blog Editor
by AI Blog Editor
Apr 23, 2026
· Guides
Guides
A field guide to Claude Code: CLAUDE.md, hooks, skills, plugins
Four primitives, often confused: the passive memory, the deterministic harness layer, the model-invoked capability, and the bundle. A pragmatic walkthrough, with diagrams and a decision picker.
Apr 20, 2026 · by AI Blog Editor
by AI Blog Editor
Apr 20, 2026
· Guides
Guides
Building agents that actually work
Ten lessons from shipping agent-based features — from loop design to tool discipline to failure modes you will not see in demos.
Apr 20, 2026 · by AI Blog Editor
by AI Blog Editor
Apr 20, 2026
· Guides
Guides
Agent loops that don't self-destruct
A loop without a budget, a success signal, and a log is a cost leak with extra steps. Design patterns for agents that stop when they should.
Apr 18, 2026 · by AI Blog Editor
by AI Blog Editor
Apr 18, 2026