The purpose of AI is to give you more say in how you spend your time. Vinci is our unified product direction, built around three simple ideas.
01
Freedom to delegate.
Give useful work a clear purpose. Our goal is to remove unnecessary supervision, so you can turn your attention elsewhere.
02
Clarity as work unfolds.
Understand what happened, what still needs attention, and what it cost. A reported success should be backed by a result you can inspect.
03
Your decisions stay yours.
Review outcomes, set limits, and decide what happens next. Human judgment and responsibility remain part of the process.
These principles guide what we build. Each release has its own documented capabilities and limits.
Next for Vinci · in development
Start the work. Make room for life.
We’re building Vinci for work you can hand over without handing away control. Start on your computer, let a useful task move forward independently, and check in from your phone.
A Vero illustration · conceptual scene
The experience we’re working toward
01
Give the work a clear goal.
Ask Vinci to investigate a bug and prepare a fix. Set the scope, the budget, and the decisions you want to keep.
02
Close the laptop. Go live your day.
The goal is for work to continue on its own machine, with a clear account of progress and a way to ask for your attention when it matters.
03
Pick up where the work leaves off.
Review the proposed change, its checks, and what it cost from your phone. Decide what happens next.
A developer-facing coding agent with orchestration capabilities. Bring your own provider credentials, read the documentation, and review the source. Vinci-managed authentication and inference are under maintenance.
Released open weights
Vinci Models
Explore MLE, Cyber, and our earlier model releases. Read each model card for its intended use, licence, evidence, and limitations.
Our open-source CLI helps our AI agents work on coding tasks and keep track of what’s done, what’s blocked, and what still needs checking. See the tool we use in our own work.
We study AI-model behaviour and the reliability of the tests used to judge it. Technical reports preserve methods, trade-offs, corrections, and negative findings.
SimpleDirect builds software, engineers AI systems, and publishes research. We want technology to demand less of your life—and leave more of it up to you.
