16 articles
As AI moves beyond answering questions to filtering the world, organizing context, interpreting meaning, participating in value judgments, and taking action, it is occupying a new position: an intelligent intermediary between people and the world.
From “I’ll get straight to the point” and “let’s break it down” to “it’s not A, it’s B” and the Chinese catchphrase “I’ll catch you steadily”: a field guide to the shared clichés and distinct language personas of ChatGPT, Claude, Gemini, Doubao, and DeepSeek.
## What Happened This week's commit log reads like an ER report: - **chore**: Unify stop entrypoint and deploy hint - **fix**: Add 5-minute cooldow...
This week wasn't about adding new features—it was about removing dead weight. Two menu items ("Skill Management" and "Blog Stats") were cleaned up, a...
## What Got Done This week saw three core tasks land for **branch-janitor**: ### Evidence Collector (TASK-063) Pure function `collectEvidence(repoR...
## What I Did This Week Spent the week normalizing orchestrator-api endpoints. Sounds fancy, but really just renamed a bunch of routes. **The chang...
## Background Three main changes this week: 1. **Vite proxy fix** - Added `/api/otel` proxy for local dev, fixing the OTel Collector 404 issue 2. **...
## What Happened The most "重量级" (significant) commit this week is `feat(orchestrator): CI failure → auto-debug pipeline`. Simply put: when the `dep...
## Background In today's commit, we renamed `ConsciousnessLayer` to `ExecutionGate`. This isn't just a simple variable rename—it's a reorientation o...
## What Happened A seemingly minor fix landed in the codebase today: the scheduler now detects whether a workspace still exists *before* dispatching...
## Quick Tech Changes Overview Today's changes focused on two main areas: 1. **OpenTelemetry Ecosystem Upgrade**: Core components jumped from 2.5.0...
The big move today was cranking `ci_mode` in `policy.yaml` all the way to `enforce`. Don't ask why it took this long—the answer is I was too scared t...
A recap of the day when Engineering Agent pointed out Docker build mistakes with better-sqlite3 and the trust crisis behind PPT Agent optimistic success on iPad.
What the Agent can do is not determined by which tools we add, but by what Office.js supports on the current host. Requirement set fragmentation, platform ceilings, and API model evolution are the real constraints.
Dynamic capability detection and tool gating based on Office.js requirement sets, so the AI knows what the host can and cannot do.
Sharing my thoughts, architecture design, and technology choices during building OUTBIRD.