Stephens AI · our own tooling · Stephens AI · v2 live 2026-03-01

The Bridge: the AI operating system our company runs on every day, where every new tool is tested before a client sees it.

We run our own company on it every day: calendar, inbox, messages, meetings and priorities on one screen, with the whole company's knowledge behind it. Anything we install for a client runs here first.

10agents each on a least-privilege tool list, plus 45 reusable skills, kept lean by a monthly audit that retires what stops earning its place

07:30 every morning: new meeting recordings become pages, priorities, decisions, and a text with the action items

The Bridge dashboard: a 3D graph of the vault beside today's calendar and inbox — The Bridge, the dashboard we run the company from. Two panels blurred.

What I built

The build

The Bridge, a dashboard over the whole company: a 3D map of every note, both calendars merged and editable, both inboxes, iMessage and Slack, health data, and Ask Charlie, a chat over the whole knowledge base that can read but never edit or send. Voice dictation in every box.
10 specialist agents, each on a least-privilege tool list (reviewers that read like a dentist, an optometrist or a physical therapist; research; meeting notes), and 45 reusable skills for proposals, research, QA, deploys and reporting.
Scheduled jobs: a daily 07:30 meeting sweep that writes the meeting page, updates the week's priorities, logs decisions, and texts the action items; a weekly vault health check. Every run leaves an inspectable log.
Doctrine, written down: verify the output, augment rather than replace, understand the why, loop humans in, stay transparent. What stops earning its place gets retired.

What changed

Before and after

Before

By hand

After a client callNotes typed up later, or not

Finding a factGrep and memory

A repeated jobRe-explained each time

Sending anythingWhatever the model decided

After

The system

After a client callPage, priorities, decisions, text by 07:30

Finding a factRetrieval search over the vault

A repeated jobA skill with the gotchas encoded

Sending anythingA human approval step, always

Where the numbers come from. Counts are from the system's own folders on 4 October 2026. They change as skills earn their place or get retired.

For engineers

How it's built

Architecture: A markdown vault with frontmatter as the database, a generator that rebuilds the index, sub-agents defined as files with tool allow-lists, skills as files with the gotchas at the top, a local dashboard server, and scheduled jobs on the Mac that deliver to iMessage.
Where it runs: On my Mac, with connectors into Google Workspace, Slack, Fathom, and the clients' tools.
Tests and evals: A weekly lint over the vault (contradictions, stale facts, orphans), a layout lint on every branded page, and a two-reviewer gate (adversarial fact-check plus a domain reviewer) before anything client-facing ships.
Cost and latency: One subscription plus about $10 a month of API credits.
Guardrails: Every agent runs on a least-privilege allow-list. Nothing is sent, deleted, or deployed without a fresh yes. Every scheduled job leaves a trace.
The tradeoff we chose: Why this way and not the other
Files over a database. A vault of markdown is slower to query than Postgres, but every agent, every human, and every future rebuild can read it, and the whole thing is in git.

In their words

“The Hermes agent… thank you so much. This is sick so far. I set it up… and it freaking set the whole thing up.”

Payton Bilodeau, AI educator, AI Lab