Agent Orchestration Goes Managed, Webhooks Turn AI Into Event-Driven Teammates, and Roast Mode Ships Zero False Positives
Published on 05.03.2026
Gas Town by Kilo: Managed Agent Orchestration for Multi-Agent Coding Workflows
TLDR: Steve Yegge's Gas Town agent orchestrator -- which coordinates 20-30 AI coding agents simultaneously with roles, merge queues, and patrol loops -- is getting a fully managed version on Kilo's cloud infrastructure. It removes the ops overhead of self-hosting while providing elastic scaling and access to 500+ models through a single gateway.
Cloud Agents + Webhooks: Turning AI Coding Assistants Into Event-Driven Automation
TLDR: Kilo's Cloud Agents now support webhook triggers, allowing external systems to kick off AI coding sessions via HTTP requests. This enables automated workflows like issue-to-PR pipelines, dependency updates, documentation sync, security patching, and tech debt cleanup -- all triggered by events in your development ecosystem.
Will It Roast? Testing AI Code Review's Brutal Honesty Mode on 5 Levels of Terrible Code
TLDR: Kilo tested its Code Reviewer's Roast Mode on five PRs of escalating code horror -- from sloppy variable names to eval() on user input. It found 39 real issues with zero false positives, escalated severity appropriately from "address before merge" to "do NOT merge," and consistently prioritized real bugs over easy jokes.
Will It Roast? 5 Levels of Terrible Code
Community and Ecosystem Notes
The newsletter highlights 35+ community contributions merged in a single week from 27 contributors, spanning new providers, model support, bug fixes, and UX improvements. The one-click review suggestion feature -- where Kilo suggests starting a review after completing a task in Code or Orchestrator mode -- is a small but thoughtful workflow optimization. Reducing the friction between "I finished writing code" and "I should review what I wrote" catches more issues at the point where developers still have full context. The Slack integration for editing code directly from chat conversations is worth watching, though the practical value will depend heavily on how well it handles the inevitable context limitations of chat-based interfaces.