gabriel.dev

0% read

on this page

introwhy we built itbuild logproduct detailswhat shippedwhat i'd change

meta

status live
stack react/vite
ai session telemetry
aiOS · 2026

Make AI-assisted work observable.

aiOS is the React observability cockpit inside the ai-squad repository. It turns recorded session artifacts into an attention board and detailed evidence views for phase, duration, cost, verification, and decisions. Screens shown use synthetic data.

aiOS cover: ivory observation rings around a terracotta core
AI work becomes easier to manage when cost, time, phase, verification, and required human attention are visible in one place.

React product inside the ai-squad monorepo

decisions.md (aios)

# challenge, architectural choices, and technical outcomes

focus interface · board

context: board

Organize sessions by the attention they need.

I designed a React board that reads ai-squad session artifacts and groups work into needs-you, in-progress, awaiting-deploy, and done states, with a separate archive. Cards expose phase, duration, cost or token data, and the reason attention is required.

focus details · drill-down

context: drill-down

Keep the evidence behind every status.

The session drawer connects the overview to task states, verification evidence, decisions, and the implementation trail. The server only exposes files from projects already known to the board.

focus overview · aggregates

context: aggregates

Add the portfolio-level view above the sessions.

The Overview aggregates attention, delivery, spend, and efficiency for today, seven days, or thirty days. It reports known-cost coverage honestly, includes P50 and P95 session costs, and connects summary cards and the feature table back to the board.

selected architecture · 1 deep dive

role / challenge / choices / outcomes

Portfolio-level triage backed by evidence recorded inside each AI-assisted work session.

My contribution: Product definition, information architecture, React implementation, server boundaries, and visual system.

product

Attention board

Session drill-down

Overview · synthetic data

Cockpit walkthrough

technical outcomes

4

Kanban states

3

overview windows

P50/P95

cost distribution

React

interactive cockpit

retrospective.md

The useful unit is a work session with a phase, evidence, cost, decisions, and a clear reason a person should look at it. The screenshots use synthetic data, and show the current local aiOS version inside the newer local ai-squad work.

github ↗

03 / 04