Product lab · founded in 2008 · São Paulo
Built with AI across the whole cycle.
LEVELHARD started in 2008, in Tatuí, as a technology consulting and development company. The name carried through eighteen years of shipped work, and today it is the brand of a lab that puts AI in every step of the cycle and ships the result.
LEVELHARD was born in 2008 in a small city in the São Paulo countryside, building software for people who needed it to work. Eighteen years later the name is the same and so is the bar: we do not believe in slide-deck AI, we believe in the kind that runs. Every thesis goes through the whole cycle, from research to the store, and only what people actually use stays up.
Shipped products, not slideware. Each one started from a concrete problem and only stays up because people use it. Open any of them.
The lab: an entire operating system in the browser.
Windows, dock, Launchpad, a Docker container store and native AI agents. On the Web and approved on all three stores (Apple, Google and Microsoft). It is where the other experiments are born and where the AI development harness runs in a loop.
The musician's toolkit, running entirely on the device.
Tuner, metronome, real-time stem separation, a mini DAW and an air guitar through the camera. It started to help me study music; today a community asks for features. iOS on the App Store, desktop in alpha and Android in beta.
A table and court scoreboard that never interrupts the hand.
Truco, volleyball, tennis, ping-pong, beach tennis, badminton, basketball and free count, each with its own rules. Flat on the table, the board flips toward both players; on the Lock Screen, it shows live. Free, no ads, no account. On the App Store.
What has not left the lab yet.
The difference between a company that talks about AI and one that ships with it is what happens between the idea and the store. Here the whole cycle is the product.
the cycle, as it runs
Nothing starts with "let us build it". It starts with what has to be true at the end, written so a machine can check it. An acceptance criterion you cannot execute is not a criterion, it is hope — and a design that leans on one was already wrong.
Research, design, build, visual QA, release and documentation chained in a loop, with AI at every step. Every round goes through the gate: lint, build, tests and screenshots actually opened and read. A green build does not prove a screen, and "looks fine" is not evidence.
What passes goes to the store and lives under the store's own review — Apple, Google and Microsoft. What does not pass does not ship. That is what separates software that runs from software that was demoed.
everyday tools
Founder
RoqueRibeiro
Roque Ribeiro founded LEVELHARD in 2008 and never stopped building. Twenty-five years in high-complexity environments — Yazaki, Accenture, Netshoes, Mercado Livre, Americanas, Avanade and, today, Bradesco, where he is an IT manager leading the AI Engineering CoE. He speaks about AI in the development cycle at Febraban Tech, TDC and PPCC.
AI in the development cycle, a product to build together, a stage to share, or plain curiosity about how all of this is made. Same door.
