Senior LATAM product engineers, working inside an instrumented AI delivery framework.
The nearshore market is saturated with "we use AI." Nobody can show you what that means inside a session: which models get used, whether tests run before a commit, how often AI output gets challenged before it ships.
We built a framework that proves it. Every Eversynced engineer operates inside it, contributes to it, and is measured by it. We then open-sourced the measurement tool so any team can run the same check on itself.
Engineering company first. Nearshore second.
pheebs - Local-first telemetry for AI coding agents
Installs as lifecycle hooks in Claude Code, Cursor, and Codex, plus OpenTelemetry export where the agent supports it. Writes one JSONL line per event to your own machine and sends to a backend you control, if implemented and configured.
npm install -g pheebs # Install
pheebs init # Interactive setup, all three agents
pheebs doctor # Check the wiring
ai-proficiency-model.md - Proficiency = repertoire + quality signals
A versioned document describing the practices and signals that matter when working with AI agents. It gives structure to what would otherwise be a stream of raw events, and it is published on its own so it can evolve in the open and be adopted beyond Pheebs.
- Layer 1, repertoire. Twenty-four practices across six competencies: Models, Artifacts, MCP, Evals, Context management, Orchestration. Each is either Unobserved, Adopted, or Recurring.
- Layer 2, quality signals. Verification coverage, pushback rate, refinement-to-repair ratio, and wholesale-accept rate. What happens to AI output before it ships.
Backend contract — Five routes, and a reference implementation in the repo
Pheebs holds a base URL and a token, nothing else. Stand up your own backend and telemetry goes from your developers' machines to your infrastructure. Pheebs never sees it.
