I’m an AI-augmented human

Hi, I’m Eduarda Ferreira. I’m a mechanical engineer and the co-founder of Socra, where we build the tools for working with AI teammates. I run a swarm of AI agents, each with its own role, and I’m responsible for our UI and UX. I decide what we build, my agents build it, and I write zero code.

I’m traveling across the US while my AI teammates do the work. I talk to them from the Tesla, set the direction, and keep going.

An idea for a product. Something that feels wrong in an interface. A detail I want to work the same way everywhere.

I bring that to AI eduarda, who is like my chief of staff, and to Design Manager, my AI teammate responsible for our shared design. They turn my direction into work, coordinate the agents doing it, and carry it through the products. I care about the result: can someone understand the page, find the next action, and use it comfortably on a phone?

Meet AI eduarda is the story of that relationship: how my direction becomes work, how a correction reaches the products, and how the next agent picks up where we left off.

If you want to start with your own work, give your AI teammate a first assignment.

Cortex is where we keep what we have decided and why, so our agents retrieve it before they work. I built Cortex’s UI/UX with my agents, starting from what we had already standardized for Journeys, our first product. Mike, my co-founder, built Cortex’s API, backend and MCP, the connection agents use to work with it. I built the experience people use.

That meant making connected knowledge usable: finding and editing Modules (where a decision and its reason live), following their Dependencies (the knowledge you need first to understand it), and moving between an Issue (one piece of work) and the knowledge behind it. The interface had to keep that context clear on a phone as well as a desktop. We could build those experiences on the design decisions and components we had developed for Journeys.

That foundation is socra.design. It covers our buttons, fields, menus, dialogs, navigation and page layouts, including how they behave. The working components and the knowledge behind them are shared across our products.

When a shared component needs a correction, my agents fix it once and carry it into every product that uses it. One avatar correction reached Cortex and Journeys, one selection change reached five apps, and one focus correction went through three agents into Journeys and this site.

This is how design compounds for me. We can spend our effort on what is new because the navigation, controls and lessons from earlier work are already there. An improvement becomes part of what the next feature can reuse. That is how I can direct work across our products, and why I want hundreds of future products to start from the foundation we are building now.

Mike builds API first and keeps the knowledge behind his services in Cortex. When my agents work on Cortex, they can retrieve the knowledge behind his side of the product and use its published API. I do not have to learn every backend detail and explain it to each agent myself.

In Cortex, our Modules say what should be and why. Dependencies connect the knowledge needed to do the work, like pathways in a brain. Issues hold the work needed to bring the product into line with that intent.

One Growing Brain shows how you can build a system like this with Cortex. The more useful knowledge and working pieces we can reuse, the less we have to start over.

On this site, fresh agents retrieved our saved decisions to check navigation on desktop and phone and write a new inventory opening. Those records show what they received, produced and checked.

Before software, I built things you could hold in your hands. In college in Brazil, I rebuilt a wind tunnel so the fluid mechanics course could teach on it. Later, at a startup printing homes, I designed a nozzle that solved the printing problems that bothered me. It earned three patents. There’s a longer story behind how I got here.

At Texas Instruments, I worked in a full cleanroom suit. No makeup or perfume, because one particle can kill a chip. We measured the work, found the cause of a defect, and changed the recipe for the next run.

When coding agents arrived, they made changes that passed tests and wrecked what we had built together. I kept explaining things I had already explained. That is where Cortex came from. I wrote about the damage, the experiments, and what we changed.

Mike and I are the only two humans at Socra. Together, we have built dozens of products and services, most of them internal. Journeys, Cortex, our admin tools and this site build on shared design. Here’s the full picture of what we’re building.

At Socra, an AI teammate has its own identity and cloud computer. It signs in as itself and works with the access you give it. You give it a role; Cortex gives it the knowledge and standards to do that role your way, and what it learns stays available to it and to the next teammate. It is built for small teams starting out, who want roles filled before they can hire for them.

Design Manager is one of mine. Quality Manager checks whether the result meets the intent. Fab Manager, named after the chip factory, helps me see where work keeps coming back for repair. What an AI teammate can own explains how I think about those responsibilities. Our Docs Quality teammate checked the pricing and setup sources before we published that guide. Its work changed what we told readers.

I want more useful work from the people already on a team. I want room to think, to hike, to take my dog out and notice things that become the next idea. My shorter observations along the way are in Thoughts.

You do not have to build software to start. If you keep explaining the same preferences to the AI you already use, that is the problem Cortex solves: a connected agent retrieves what you decided before it works. Give the AI you already use work you know how to judge. From meeting notes to a checked draft lets you inspect a small assignment: turning notes into a useful document without losing a decision. The notes are fictional; the saved draft and correction were performed.

Then carry what you learn into Cortex, which is part of Coagent, Socra’s paid product for a team of people and AI teammates. Connect a supported agent so it can save what a good result means to you and retrieve it on the next assignment. The guide walks you through that first assignment and states the price of the paid step before you reach it.