Domain Arrivals
You’re viewing the 2026-08-16 issue, not the current issue. Go to the current issue: 2026-09-09

Featured in Domain Arrivals · 2026-08-16

Aetius Treats AI-Agent Security Like Regression Testing

aetius.app ↗ · Developer Corner · score 84.0
○Open: public page gives enough substance to understand the site $Paid: pricing, booking, ecommerce, subscription, or paid access is visible ADAds: visible ad load or ad-supported content ✦Pretty: notable design, copy, craft, or presentation ◆Pro: polished, finished-feeling, serious, or operationally mature ◇Niche: specific audience, workflow, or unusually focused use case ◉Human: personal, local, community, civic, handmade, or real-person signal !Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern

A pre-launch system reconstructs a production AI agent from its traces, runs adversarial tests before and after a change, and reports a credible interval instead of a brittle pass/fail score.

Why it surfaced

Aetius focuses on a concrete failure mode: a prompt edit, tool addition, or model upgrade quietly making an agent easier to compromise. Its illustrative report distinguishes flagged, no-difference-detected, and inconclusive results, with an example using 2,400 tests per arm and explicit power calculations.

Read the complete issue Observed source ↗

Continue into the edition

The 240-Material Organ

August 16’s newborn domains include a 240-material perfume organ, a repeatedly saved JPEG cat, a ski tracker designed by a 15-year-old, and a searchable page claiming to contain 148,303 prescription records.

This is one of 1,000 discoveries in the 2026-08-16 Domain Arrivals issue.

Explore the complete issue Browse Developer Corner

More from Developer Corner

Three more arrivals

Laplace, the Pokémon bot that searches eight possible worlds trylaplaceai.com

Laplace is an open-source Pokémon Showdown AI that plays Gen 9 Random Battle using determinized Monte-Carlo Tree Search.

 Claimset, the stale-citation detector claimset.org

Claimset is an open format and command-line verifier for checking whether published claims still match their cited sources.

 Observability Arena Puts Agent-Observability APIs Through Their Paces agent-otel.tech

An executable benchmark ranks five OTEL-based agent-observability platforms by how effectively their read APIs return trace data.