System 1.0: releases

Softwarethat runson your machine.

I'm João Pedro Schimidt Mantovani. I build and ship my own desktop applications. Vidi finds any scene inside your video archive; CompanIA is an assistant that lives on your screen. I also develop custom systems for companies running on spreadsheets, notebooks and WhatsApp. Every project is published with its real stage of development on display.

  • Vidi · internal beta
  • CompanIA · pre-alpha
  • Computer Engineering · Mauá
Apps of my ownCustom systemsVidiCompanIAWindows · macOSSpec before codeMeasured evidenceBuilt in publicApps of my ownCustom systemsVidiCompanIAWindows · macOSSpec before codeMeasured evidenceBuilt in public
01Releases

The apps

Each with its own page, its real stage declared, and a separate waitlist. You join only the one that interests you.

Custom systems

I also build your company's system

Finances, scheduling, team and clients in a single system, shaped around the way your operation already works. The same method that builds the apps built the ERP running the operation of Brazil's largest character-entertainment company, in daily use by the people who work there.

About the service
02Method

How one person delivers this

You follow products that get finished, not demos that disappear. What sustains the pace isn't typing fast. AI types fast and produces bad code at scale just as easily. What sustains it is the process.

  • 01

    A specification before the first line

    Every feature starts from a specification written before the code: what it is, who it's for, what behaviour, what acceptance criteria. Its real job is to reveal early that an idea doesn't work. That is far cheaper than finding out after it's built.

  • 02

    A constitution per repository

    Each project has rules even I can't break: local privacy, user media read-only, tests before implementation, versioned contracts with mandatory migration. A written rule is the only kind that survives a rushed late night.

  • 03

    Nothing is "done" by impression

    A milestone only closes against an objective exit criterion with evidence attached. Evidence that didn't come from a committed tree doesn't count. I've invalidated one of my own runs over that rule, with a good number on the screen.

  • 04

    AI accelerates. It doesn't sign

    AI doesn't choose the data model, doesn't define business rules and doesn't carry risk. It accelerates the repetitive work inside those constraints. Architecture and judgement remain engineering, and Computer Engineering at Instituto Mauá de Tecnologia is where I go for the rest.

03Background

Who is on the other side

You know who to hold accountable, and that changes what you can demand. It's no coincidence that the first application is about video archives: I've been on both sides of that pain, the studio floor and the person keeping the system up.

  1. 00

    First job, at fifteen

    2015

    Six months at my father's company, Manto Produções: helpdesk, printers, Photoshop compositing, photo and video editing, the beginnings of motion design. It was also my first contact with programming. Basic, but enough to know it would come back.

  2. 01

    Technical degree in Mechatronics

    2017

    ETEC Getúlio Vargas, taken alongside secondary school. Electronics, mechanics and logic before turning 18. Engineering from early on.

  3. 02

    Prime Talk: the IT sustaining the production house

    2019

    I joined the production company of TV presenter Dony De Nuccio and economist Samy Dana, both well known in Brazil, as the person responsible for IT: helpdesk, servers, NAS, networking. In a video production house, when Premiere crashes, that's my problem. I grew fast because things broke often.

  4. 03

    The PTZs took me into the studio

    2019–2024

    When the PTZ cameras arrived, I was the only one who could make them work. On a 3 a.m. shoot, someone was missing to record audio. I recorded it. I never left the studio again: lighting, sound, the switcher, vMix before vMix was popular, sets built with my own hands. The technical half working next to the artistic one.

  5. 04

    The broadcasts

    2019–2024

    Through the production house: daily YouTube shows for Easynvest, which became NuInvest; agribusiness content for Bayer with international coverage trips; shoots at B3, the Brazilian stock exchange. And the high point: taking part in the broadcast to Brazil of Nubank's IPO, live from the New York Stock Exchange.

  6. 05

    Back at Manto, now in management

    2024

    I returned to the family company, Brazil's largest character-entertainment company for events, to take over general management: process, standards, an engineering mindset in an operation that ran by hand. And, from time to time, a hand-picked freelance job: I ran the technical side of a live broadcast at TV host Ana Maria Braga's home, and went to Hawaii as outsourced crew for Globo and Disney at the Moana 2 launch.

  7. 06

    Engineering, the ERP and the launches

    2025–today

    I started Computer Engineering at Mauá and built from scratch the ERP that runs Manto's entire operation, used every day by the people who work there. The focus turned fully to software, for the scalability and because it's what I like doing. The applications on this page come from here.

About this developer

João Pedro Mantovani

Software developer · applications and systems

In the field
since 2015
Based in
São Paulo, Brasil
Education
Engenharia de Computação · Mauá
Status
Taking on projects

Allocation

  • Software88%
  • Audiovisual72%
  • Infrastructure64%
  • Management48%

Education

Computer Engineering

Instituto Mauá de Tecnologia

Started in 2025, expected to finish in 2030, at one of the best private universities in Brazil. Before that, a year and a half of Science & Technology at UFABC. I left to commit to the work that became all of this. Data structures, systems architecture and network theory are what separate an application that can grow from one that breaks at the first spike.

Frequently asked
A solo developer. Will this still exist in a year?
It's the first question I would ask. The answer I can give today isn't a promise, it's a method: each project has a milestone roadmap with objective exit criteria, and each page shows what's done, what's open, and what's blocked. If something stalls, you read it here before finding out on your own.
Is this just another AI wrapper?
Vidi sends nothing anywhere: once the model pack is installed, indexing and search run offline on your machine, and your media is opened read-only. CompanIA uses Claude and says so plainly. The work is in the desktop layer: an always-visible transparent window, memory isolated per character, local tools. Neither is a text box on top of an API.
If it's built with AI, is it built carelessly?
The opposite, and it's verifiable: a versioned specification per feature, tests written before implementation, contracts that don't change without a migration, and an evidence rule that fails closed. What AI gives me is speed inside those constraints, not permission to skip any of them.
04Contact

Let's talk

Do you work with video and want to test Vidi? Found a problem, have an idea, or want to follow the projects? I read and answer every message myself. There is no support form in between.

Contact

Based in

São Paulo, Brasil