Glossary, no jargon.

The practice of making a product work for everyone, including people with disabilities, following the W3C's WCAG guidelines.

Making the product work for everyone, including people with disabilities (WCAG guidelines).

Color contrast, keyboard navigation, alt text on images and labeled fields are the basics. It widens your audience, lowers legal risk and improves SEO, because search engines and AI read a page the way a screen reader does. It's cheaper to plan for it in design than to fix it later.

A program that uses a language model to carry out tasks on its own, step by step, using tools and data from your system.

A program with a language model that carries out tasks on its own, using the system's tools and data.

A chatbot answers questions. An agent acts: it reads data, runs checks, opens tasks, writes reports. For a founder, the payoff is in repetitive routines that today depend on someone remembering to do them. The watch-out is limiting what it can access and checking what it delivers.

A set of rules that lets two systems exchange data with no person in the middle.

An API is how your system takes payments, issues invoices, sends WhatsApp messages or talks to an AI model. A well-documented API lowers the cost of every new integration. Before picking a vendor, check that their API covers what you need and what it costs per call.

The part of the system that runs on the server: business rules, database, authentication and integrations.

Users don't see the back end, but it's where the data and the rules that make the product work live. Security issues, slowness and inconsistent data almost always start here. In a project rescue, it's the first part we audit.

The percentage of customers who cancel or stop using the product in a given period, usually measured monthly.

In a SaaS, high churn wipes out the effect of any acquisition campaign. The most common causes are confusing onboarding, recurring bugs and a product that stops evolving. Measuring churn by cancellation reason shows where to invest first.

An Anthropic tool where the Claude model reads, writes and tests code right inside the project, from text instructions.

It speeds up the build a lot when the person using it knows how to review what the AI writes. Without review, it turns into vibe coding and piles up technical debt. At the Studio we use Claude Code alongside Figma in design and build, always with human review.

Old code that's still in use but hard to understand, test or change, often with no documentation and no original author.

Old code still in use that's hard to understand, test or change.

Every product that succeeds ends up with some legacy code. The trouble starts when every small change takes weeks or breaks something else. The fix is rarely rewriting everything at once: it's mapping, stabilizing and replacing piece by piece.

When a visitor takes the action you want: fills out a form, subscribes, buys. The conversion rate is that action divided by visits.

When a visitor takes the action you want: fills out a form, subscribes, buys.

Doubling conversion usually costs less than doubling traffic. A clear offer, page speed and a short form matter more than button color. Pick one main conversion per page and measure only that at first.

Three Google metrics that measure the experience of loading and using a page: LCP (loading), INP (responsiveness) and CLS (stability).

Google metrics for loading (LCP), responsiveness (INP) and stability (CLS).

They're a Google ranking signal and directly affect conversion: a slow page loses visitors before it shows the offer. You can check yours in PageSpeed Insights or Search Console. On the sites we ship, these metrics are part of the deliverable.

Standardized code, usually JSON-LD, that tells search engines and AI what each thing on the page is: company, article, product, question.

Code (usually JSON-LD) that tells search engines and AI what each thing on the page is.

It helps Google show rich results, like FAQs and reviews, and helps AI cite your company with the right details. Visitors don't see it, but machines read it. It's one of the first things we set up in an SEO and GEO project.

Pushing a new version of the code to the server where the system runs, making the change available.

On mature teams, a deploy is a minutes-long routine that happens several times a week. If every deploy is a tense, manual, slow event, that's a sign of a fragile process. Ask to see how the team ships before you hire.

A set of components, visual rules and tokens (colors, type, spacing) that the whole product uses, in design and in code.

With a design system, a new screen ships faster and matches the rest. Without one, everyone invents a different button and the product looks patched together. It doesn't have to start big: begin with the 10 to 20 components you use most.

The first step, where we understand the problem, the product and the business before defining scope, price and timeline.

Without discovery, any quote is a guess. It's where we separate what needs to ship now from what can wait. At the Studio, the fixed price and timeline only come after discovery, and the deadline is written into the contract.

The research and definition phase where the team figures out what to build: problem, users, hypotheses and priorities.

Discovery keeps you from spending months building something nobody uses. It can include interviews, competitor analysis, prototypes and quick tests. The risk is that it never ends: it has to finish with decisions and a scope.

The future cost of shortcuts in the code: every quick fix today charges interest tomorrow in slowness and bugs.

A little debt is normal, even healthy, when you need to ship fast. The problem is when nobody tracks it or pays it down. Clear signs: everything takes longer than before, the same bug keeps coming back and the team is afraid to touch certain parts.

The list of what a project will deliver: features, screens, integrations and what's left out.

Vague scope is the most common cause of delays and fights between client and vendor. A good scope says what's in and, just as clearly, what's out. If it isn't written down, assume it isn't included.

A contract model where scope, price and timeline are set before work starts and don't change without agreement.

It gives the buyer predictable costs. It only works when the scope was well defined first, which is why it depends on serious discovery. Changes midway come in as an amendment, with their own price and timeline.

Ongoing work to improve the product after it ships: fixes, new features and adjustments based on real usage.

Monthly work to improve the live product: fixes, new features and adjustments.

Going live is the beginning, not the end. Usage data shows what needs to change, and a dedicated team responds faster than a new project for every request. It's one of the Studio's services for companies that already have a product running.

The part of the system that runs in the browser or app: the screens, buttons and everything users touch.

A good front end is fast, true to the design and works on any screen size. It relies on the back end to fetch and save data. When a design looks great in Figma and off in the browser, the problem is usually in the handoff or the front end.

Generative Engine Optimization: optimizing your content to be found and cited by generative AI like ChatGPT, Claude, Perplexity and Gemini.

Generative Engine Optimization: getting found and cited by AI like ChatGPT and Gemini.

More and more people ask an AI instead of searching Google. To get cited, your site needs clear answers, structured data, visible authorship and content the AI can read. GEO doesn't replace SEO: the two go together, and the Studio covers both in the same service.

The transfer from design to development: the files, specs and behaviors the dev team needs to build.

The transfer from design to code: files, specs and behaviors.

A bad handoff causes rework and screens that differ from what was approved. It helps to have a design system, a clickable prototype and documented states (error, empty, loading). When design and code sit on the same team, the handoff almost disappears.

A connection between your system and another service, like payments, ERP, CRM or WhatsApp, so they exchange data automatically.

Connecting your system to payments, ERP, CRM or WhatsApp so data flows on its own.

Integrations tend to be the least predictable part of a project, because they depend on third-party docs and limits. List them all during discovery. Each one needs error handling: what happens when the other system goes down?

A page with a single goal, like capturing a lead or selling a product, with no distractions pulling the visitor off that path.

A page with a single goal, like capturing a lead or making a sale, with no distractions.

It works best when it speaks to one audience with one offer. Speed, clarity in the first block and proof (numbers, case studies) matter more than visual effects. It's the go-to way to test an idea before building the product.

Brazil's General Data Protection Law (Law No. 13,709/2018), which governs how companies collect, use and store personal data in Brazil.

Law No. 13,709/2018, Brazil's data protection law, which governs how companies use personal data.

It applies to any system that stores someone's name, email or CPF (Brazilian taxpayer ID). In practice: collect only what's needed, have a legal basis for each use, protect the data and respond to data subject requests. It's cheaper to plan for it in the architecture than to retrofit it later.

Large Language Model: an AI model trained on huge volumes of text to understand and generate language, like Claude, GPT and Gemini.

Large Language Model: AI trained on massive amounts of text, like Claude, GPT and Gemini.

It's the engine behind chatbots, agents and AI coding tools. Each LLM has different cost, speed and quality, and the right pick depends on the task. What weighs most on the result is usually the context and data you give it.

A text file at the root of the site that sums up, in Markdown, what the company does and which pages matter, so AI can read it easily.

A file at the site root that sums up, in Markdown, what the company does, for AI to read.

It's a proposed standard, not yet officially adopted by every AI, but cheap to publish. It works as a short site map for language models. It's part of the GEO package, along with structured data and answer-style content.

Platforms that assemble systems from visual blocks and let you write code in specific parts.

They speed up internal systems and simple workflows. They get expensive or limiting when the product grows, needs its own rules or serious scale. Before choosing, check whether you can export your data and what happens if the platform changes its pricing.

Minimum Viable Product: the smallest version of the product that solves the core problem and can go live for real users.

Minimum Viable Product: the smallest version that solves the core problem and goes live.

The point of an MVP is to learn from real usage while spending little. Minimum doesn't mean sloppy: whatever goes in has to work well. The most common mistake is treating the MVP as the full version and going months without shipping anything.

Hiring a development team in a nearby country with a similar time zone, instead of a local or very distant team.

US and European companies hire Latin American teams for the compatible time zone and the cost. For you, the point is to weigh communication, language and meeting hours before price. The Studio, based in Fortaleza, works with clients in Brazil and other countries.

Tools that build websites, apps and automations without writing code, using only a visual interface.

Great for validating an idea or automating internal tasks. The limits show up with complex business rules, user volume and custom integrations. Migrating later is expensive, so think about the next step before you start.

The text instruction you give an AI model, with the request, the context and the expected response format.

A clear prompt, with examples and constraints, produces more predictable results. In a product, prompts become part of the system and need to be versioned and tested like code. A prompt alone won't fix missing data or business rules.

A clickable version of the screens, made before the code, to test the flow and validate the idea with users and partners.

Changing a prototype takes minutes; changing a finished system takes weeks. A good prototype shows the user's main path with real content. It also serves as a clear reference for whoever builds it.

Putting the product or a new version live for users, including domain, server, app stores and monitoring.

Deploy is the technical step; launch is everything that has to be ready for users to arrive. For apps, it includes App Store and Google Play review, which has its own timelines and rules. Plan the launch from the start, not in the last week.

Taking over a system that's stalled, late or badly broken, left behind by another team, and shipping it or getting it back on track.

Taking over a stalled or late system left by another team and shipping it.

A rescue starts with an audit: what exists, what works and what needs to be rebuilt. It isn't always worth salvaging everything, and the call should be made with numbers. It's one of the Studio's services for founders who've already spent money and still have no product live.

A plan that shows what the product will get and in what order, tied to business goals.

A useful roadmap prioritizes problems to solve, not a feature list with fixed dates. It changes as real usage shows what matters. Review it often and be clear about what's been pushed back.

Software as a Service: a system accessed over the internet with recurring billing, like a monthly or annual subscription.

The model runs on recurring revenue, so churn and onboarding matter as much as features. A SaaS needs billing, plans, permissions and, almost always, multiple companies using the same system. These belong in the scope from day one.

Search Engine Optimization: content and technical work so your site ranks better in search engines like Google.

Search Engine Optimization: tuning your site to rank better on Google.

It covers speed, page structure, titles, internal links and content that answers what people search for. Results take months, not days, but bring visits without paying per click. Today SEO work goes hand in hand with GEO.

A company that builds custom software for other companies, usually with development teams assigned per project.

A company that builds custom software, with teams assigned per project.

They tend to be strong on code and weaker on product and design, though that varies a lot. Before hiring, ask who decides what gets built, how the timeline is set and what happens when it slips. We compare the models on a dedicated page.

A short work cycle, usually 1 to 2 weeks, with a defined set of tasks and a deliverable at the end.

Sprints set a rhythm and let you course-correct often. For you, what matters is seeing something working at the end of each cycle, not a timesheet. If several sprints go by with nothing live, something is wrong with the scope or the team.

User Interface: the visual side of the product, like screens, colors, typography, icons and components.

Good UI makes it clear what's clickable, what's important and what just happened. It depends on good UX: a beautiful screen with a confusing flow is still confusing. A design system keeps the UI consistent as the product grows.

User Experience: what it's like for someone to use the product, whether they can do what they need, with how much effort and how much confidence.

User Experience: what using the product feels like, and how much effort and confidence it takes.

UX spans flow, copy, response time, error messages and even support. Better UX usually cuts churn and support tickets. It starts with understanding the user's task, not picking colors.

Building software by asking AI to generate the code and accepting the result without reading or reviewing it closely.

It works for prototypes and quick tests. In production, it tends to create security holes, duplicate code and technical debt nobody understands. Using AI in the build is great; what changes the outcome is having someone who reviews the code and answers for it.

A simple sketch of a screen, with no color or polish, that shows the structure and content hierarchy.

It's for discussing what goes on each screen before spending time on visuals. It's quick to make and quick to throw away. With a design system in place, many teams skip the wireframe and go straight to a prototype with real components.