TriStar
Web app · SaaS · 2026

Orchestrator

A personal command center that keeps side projects moving instead of dying.

The Orchestrator Today inbox showing five ranked next-actions across active projects
Role
Design + Engineering
Timeline
Ongoing · 2026
Stack
Next.js 16 · NestJS 11 · tRPC · Supabase · Tailwind CSS 4
Problem

Side projects don't fail at the idea stage. They fail by going silent.

The pattern is predictable: an idea gets momentum for a week, then a distraction arrives, and the project never gets touched again. No explicit decision was made to kill it. It just drifted.

Orchestrator is a tool built for that problem. Every project moves through a fixed lifecycle (idea → consolidate → mock → pre-market → dev → marketing → live) and the app tracks where each one sits. A daily ranked inbox pulls the single most important next-action across all active projects to the top. An AI reviews each playbook item and adds a nudge explaining exactly why that item is blocking and what to do. Stage promotions require an AI gate review before a project can move forward; manual decisions (park, kill, pivot) are recorded in an audit trail so nothing disappears silently.

Portfolio grid showing eight projects at different lifecycle stages
Portfolio: eight projects, seven stages. Each tile shows the project's current stage, a progress track, a health dot, and days since last touch.
Decisions

The inbox ranks by staleness × stage weight, not by creation date.

The naive approach (show everything, let the user decide) is the failure mode this tool is trying to fix. The ranking formula multiplies how long a project has been untouched by a per-stage weight (late stages score higher, since a near-live project going cold costs more). The result surfaces the item where inaction is doing the most damage. AI nudges are fetched lazily on first render and cached by input hash against a hard daily spend ceiling of $0.17. The same question never costs twice, and AI features stop when the cap is hit rather than running up an unexpected bill.

Today inbox with five ranked next-actions, AI nudges, and Mark done buttons
Today: ranked inbox, AI nudge per row, mark done in one click.
Portfolio grid showing projects at idea, dev, pre-market, and live stages
Portfolio: eight projects across the full lifecycle.
Mobile view of the Today inbox with cycle banner and repo suggestion card
Mobile: same inbox, narrower column, same density.
Sign in screen with a Continue with GitHub button
Sign in: GitHub OAuth, one method, no passwords.
Outcome
MVP
Phases 0–6 shipped
7
Lifecycle stages
$0.17
Daily AI cap

Orchestrator exists because I kept losing projects to silence rather than explicit decisions. The MVP is complete through phase 6. Inbox ranking, playbook management, AI nudges and stage-gate reviews, 10-day cycles, and GitHub repo detection are all live. What remains before a production deploy is the stage-gate promotion UX in the web app, a Playwright E2E test for the golden flow, and the Vercel + Railway configuration. The AI cost ceiling and caching layer mean the app costs roughly $5 a month at daily use, a constraint that shaped the architecture as much as any feature requirement.