The shortest route to the goal.
Work reaches its goal in fewer steps, with fewer model calls along the way.
Planning and decision support. It runs On request.
The shortest route on every run, with no model tokens.
In a world of rooms with a loop and a dead end, Duplex found the exact shortest route on every run, using no model tokens. GPT-5.5 matched it, spending tokens on every attempt.
On every run, Duplex found the exact shortest route to the goal.
The route stays short as your process changes.
Duplex learns from the work itself how the steps of your process connect. When the process changes, the shortest route changes with it.
The first contract takes the long way, and Duplex sees how the steps connect.
Duplex plans the route. Your controls run each step.
A route is a proposal. Every real step on it, from sending the contract to countersigning it, still goes through your tools and approvals.
Sending the contract to legal is routine, so it goes through your document tool.
Sees what an action will do.
Duplex learns how your systems respond, so it can tell what a step will do and find the shortest route to a goal. Real steps still run through your controls.
- What an action will do, before it runsYour team sees what a change will do to your systems before anyone makes it.
- A sequence of steps, played forward firstA plan is tried out first, so a dead end is found before it costs anything.
- The shortest route to the goalWork reaches its goal in fewer steps, with fewer model calls along the way.
- The next step, from plans already provenIn a familiar situation, the next step comes at once, from a plan that has worked before.
See it on your own work.
From your own assistant to every agent your company runs, a pilot puts Duplex on your own work, with the measures agreed before it begins.