Software built with AI, reviewed by people.
We develop with the latest AI models and work full-stack, in whatever stack each project needs: from the user interface to the server. On client projects, every change goes through continuous integration (linting, type checks and unit tests) and two automated reviews: CodeRabbit and a second AI review.
And people always review: no change we make enters a project’s code until a person has read and approved it, and on client projects every change of ours needs two people’s approval before it reaches the main branch. That is why what we deliver is solid.
And that is why we move fast: we are a few people working with the newest tools on the market, keeping them up to date and learning from every project. We build web applications, MVPs and digital products for US startups and for companies in Barcelona and the rest of Spain, and we offer it to public bodies too.
How we work
-
1 · One agent per task
We split every project into tasks. Each task is done by an AI agent on its own branch and ends in a pull request that can be reviewed.
-
2 · Continuous integration
Linting, type checks and unit tests on every change, with each stack’s own tools. When a check fails, the agent fixes it from the log.
-
3 · Automated reviews
CodeRabbit reviews every pull request, and the agent fixes or answers each comment. A second AI review looks for what it misses (design rules, accessibility, the project’s conventions) and checks that the pull request carries proof of the result: screenshots or test output.
-
4 · Human approval
No change we make enters a project’s code until a person has read and approved it. On client projects, it takes two before it reaches the main branch.
-
5 · Testing every release
One person on the team works on quality assurance and tests every release before it reaches users.
-
6 · Lessons written down
When a project teaches us something, we write it down as a rule the agents follow from then on.
Always current
We work with the latest AI models and tools on the market, and pick the model for the job. When a better one comes out, we adopt it, and what we learn on one project changes how we work on the next.
Why we are fast
Because the process is solid, a small team can move fast. Several agents work at the same time on different tasks, and the person you talk to is the one who directs them and reviews the code: no middlemen, no hand-offs. Automated checks catch mistakes when the change is made, not weeks later.
More capacity
When a project needs more hands, we bring in trusted developers we have worked with before, for that project only.
What we do
-
Web and application development
Custom web applications and sites, from the user interface to the server.
-
MVP and product
A minimum viable product to test an idea with real users, and the step from MVP to platform.
-
Maintenance and support
Ongoing development, fixes, deployment and support for applications in production.
-
Quality and testing
Functional testing of every release before it reaches users.
Who for
-
US startups
First product versions and platforms in production. We work with US companies continuously, in English.
-
Companies and public bodies
In Barcelona and the rest of Spain. We write proposals and documentation in Catalan or Spanish.
Clients
-
Luma Health
AI platform for patient access and healthcare operations. San Francisco Bay Area, California.
-
Bento for Business
Spending cards and expense control for small businesses, now part of U.S. Bank. San Francisco and Chicago.
-
LogZilla
Log management and analysis of network and system events. Austin, Texas.
-
Reflector
Decentralized price oracle for DeFi on the Stellar network. Menlo Park, California.
Experience
-
Agents and bots
Our recent projects for Luma Health and Bento for Business have been mostly AI: introducing agents, building bots and tuning the agents’ evaluations.
-
MVPs
Mostly first product versions for US startups.
-
Continuous work
Platforms in production that keep evolving year after year, for US companies.
-
Sectors
Digital health, business finance, online video, public broadcasting, insurance and online education.
Team
We are a small team in Barcelona: one person develops and directs the agents, another runs operations, deployment and support, and a third handles quality assurance and testing.
Contact
Barcelona.