Loading Ashlr Brief
Ashlr release
AI-generated illustration
October 26 brought releases of Locus 0.5.2, Ashlr Stack 0.2.2 and ashlr-plugin 1.37.1, plus opt-in routing in Phantom that tries local models first.
This briefing covers a release from Ashlr’s own developer-tool projects. Product claims come from the linked sources; this is not an independent review.
October 26 (UTC) had about 125 merged pull requests and three published releases in Ashlr's public repositories.
The releases: ashlr-plugin 1.37.1 was its first release published through npm trusted publishing. Ashlr Stack 0.2.2 shipped its unreleased changes, also through trusted publishing only. Locus published 0.5.1 early in the day and 0.5.2 later. The 0.5.2 work includes locus renew, an opt-in way to extend a pin within a set lifetime, and a fix so a pin whose helper process died no longer blocks its own replacement.
Phantom added hybrid local and frontier routing. When configured with localInference.routing set to hybrid, dispatch can send suitable work to a local model on the same machine and escalate to a frontier model when needed. It is off by default, local endpoints must be loopback, and a follow-up refused redirects on local probes. Merge decisions still go only to frontier seats.
Across the tools, doctor --json gained one shared ashlr.doctor v1 wrapper, so scripts and agents can read every tool's health the same way. ashlr-plugin added self-tuning compression: it tracks when a compressed result was not enough and adjusts detail per repository, and /ashlr savings now reports net tokens saved. Locus started a local pin history, on by default with an off switch, storing only hashes of repository remotes. Several repositories also pinned their GitHub Actions to exact commit hashes.
Default-off routing with strict loopback rules is a sensible way to introduce local models: you can measure them on real work without changing behavior for anyone who has not opted in.
Running suitable work on local models can lower costs, but only if quality holds. Phantom's routing records its decisions so the split can be measured rather than assumed.
If you use ashlr-plugin, update to 1.37.1 and run /ashlr savings after a few days of normal work.
Ashlr tools connected to this story. Review their source and setup requirements.
Work with eligible coding-agent accounts and local models in one workbench; review fleet proposals under signed authority.
Source material behind this briefing. Our practical implications are Ashlr’s interpretation.
This briefing was prepared with AI assistance from the linked sources. The analysis may include interpretation; consult the original sources for complete context. Cover artwork is illustrative.
Agentic AIThe briefing
AI-generated illustration
OpenAI launched GPT-6.1 Terra, a mid-tier model for computer use and visual reasoning tasks. It costs roughly half as much as GPT-5.6 Terra while running faster on agent workflows.
Agentic AIThe briefing
Agentic AIThe briefing
AI-generated illustration
OpenAI released GPT-6.1 Terra, a fast model for computer use and everyday work at $1 per million input tokens and $5 per million output tokens.
Keep coding agents scoped to the right client, account, and tenant. Manage identity bindings across command-line tools and MCP connections.
Token-efficient tools and workflows for Codex and Claude Code. Read, search, edit, and inspect repositories with compact outputs and visible context usage.
AI-generated illustration
Overnight into October 27, Phantom began recording each run's routing decision, Ashlr Stack added a merge-queue gate, and ashlr-plugin raised its MCP SDK floor.
