The number on your Work board (e.g. 783×) is your leverage over the last 7
days: how big a real company it would take to ship what your AI team shipped,
divided by the time you personally put in.
It is one equation — the Tycoon ROI Equation:
leverage = Old Cost ÷ New Cost
Cost = Hours × Rate × Coordination Tax × Interruption Tax (+ AI Cost)
Old Cost is what a real company would burn to ship the same output. New Cost is what it actually cost you. The ratio is your leverage.
The one thing to understand: human cost multiplies, machine cost adds
Everything inside the parentheses is human cost, and it multiplies — double any one term and the whole cost doubles. The machine (tokens) is an addition, and in real knowledge work it's a rounding error next to the human terms. That's why leverage comes almost entirely from collapsing two human multipliers — Coordination Tax and Interruption Tax — not from AI writing faster.
How each term is measured from your workspace
Hours — the raw execution. We add up every finished, visible task in the window and give each one human-equivalent hours: how long a competent human would take to produce that deliverable (the value stamped at completion, or a role default for older tasks). Wall-clock isn't used — agents finish in minutes what takes humans hours.
Coordination Tax — the team friction you skipped. A real org shipping that much execution needs people, and people don't scale for free: n people have n(n−1)/2 sync channels to keep aligned (meetings, handoffs, reviews). We infer the team your output implies — the individual contributors plus the management layer above them — and price its coordination. A solo-equivalent workload pays 1× (no coordination); it climbs with team size and saturates around 4×, because real orgs break into small teams instead of keeping every channel live. This is where most of your leverage is: your version of that tax is 1× — one conversation with Astra.
Interruption Tax — how often the work stops for you. On your side, this is simply the time you spent: every message you send Astra and every approval you resolve is one interruption, priced in minutes. Agents that run in parallel and only surface at real decision points keep this tiny — which is exactly how one person supervises a whole team at once. A week where you're pinged constantly drives it up and your leverage down.
Rate and AI Cost — deliberately left out. Required expertise collapses (an operator can run work that used to need a senior specialist), but we can't observe real salaries, so Rate is held equal on both sides. And the AI bill is an addition so small it barely moves the number — so the headline leverage is a clean ratio of hours, not dollars.
Worked example
A company shipped ~2,585 human-hours of execution last week. The team that volume implies is large enough to hit the coordination ceiling, so the fully- loaded old-world cost is about 2,585 × 4 ≈ 10,300 hours — roughly a 260-person company. The chairman spent about 13 hours steering it.
leverage = 10,300 h ÷ 13 h ≈ 783×
Read it as: running a ~260-person company on ~13 hours of your week. A small workspace that shipped one project reads far lower and honestly so — little implied team means little coordination skipped.
What it deliberately does not do
- No AI judge of task value yet. Human-equivalent hours come from a transparent, role-based heuristic so the number is inspectable and hard to game; a future independent judge, capped per task, can replace it.
- No exact time tracking. Your invested hours are proxied from real messages and approvals; a precise activity signal can slot in later without changing the equation.
- It rewards building, not disengaging. The leaderboard ranks by leverage but every row also carries total output, so the way up is to ship more with the same attention — not to stop showing up.