R$99 a month for more than 15 AI models under one login is the claim behind the Adapta AI platform. The pitch only pays off if you actually route different tasks to different models, so the real test is a routine built from observation, redesign, division, execution and maintenance.
What the Adapta AI platform actually offers
The Adapta AI platform is a Brazilian subscription that gathers more than 15 AI models behind a single login, and the creator of the source video reports paying R$99 a month for unlimited access to them. It is a reseller-style interface, not a model builder, so the models inside it are the familiar ones.
That distinction matters. The video presents Adapta as a way to stop juggling separate subscriptions for ChatGPT, Gemini, Claude and Perplexity, and the demo shows a chat selector plus a small library of courses on using AI for images, law and medicine. Adapta itself is the wrapper around those models, and the pricing, model list and course catalogue are commercial terms that change, so treat any figure here as the state described in the May 2026 video.
One caveat from the transcript is easy to miss. The presenter calls the service a platform that brings together the best AIs in one place, and that is an accurate description of what she uses it for. It is not evidence that Adapta has a partnership with each model vendor, and the video does not show any vendor confirmation. If vendor-official access matters for your work, check Adapta's own documentation before you commit.
How the account works in practice
Signing up is account creation, then a single interface. A left-hand chat area lists the available models; selecting one opens a chat with that model. The courses sit in a separate tab that unlocks when you subscribe. The demo generates images inside the same account, from models dedicated to image work rather than from the text models.
The presenter compares the value to hiring an assistant that works 24 hours a day, which is sales framing rather than a measurement. What the demo actually shows is task switching: one question to one model, the next question to a different model, without logging out or paying twice.
Which model fits which step of the routine
The presenter assigns a different model to each step of her method based on how that model behaves, which is the practical core of the whole workflow. Her mapping is first-hand experience from a single creator, not a published benchmark, so treat it as a starting heuristic rather than a measured ranking.
| Step of the routine | Model | Why that model |
|---|---|---|
| Observation and diagnosis | ChatGPT | Direct answers; good at structured diagnostics and long descriptive prompts |
| Redesign | Claude | Asks clarifying questions before answering; analytical tone |
| Division by area | ChatGPT | Content creation, writing, idea generation, fluid back-and-forth |
| Execution | Perplexity-time research with cited sources | |
| Maintenance | Gemini | Summarising and analysing a week of reports |
Two models the presenter names sit outside the five steps. DeepSeek is described as the most logical of the set, and Grok as the most direct. Adapta's own catalogue and any current model list should be checked separately, since the models inside a reseller interface can change.
The ORDEM method: observation first, not new habits
ORDEM stands for Observação, Redesenho, Divisão por área, Execução and Manutenção, a five-step routine method the presenter teaches before touching any AI tool. The first step is deliberately passive: you map where the chaos actually is instead of asking a model to invent a new schedule.
Here is the sequence in order, because the later prompts only work if the earlier output exists:
- Observação (observation). Describe your current routine as facts and let a model name the three biggest areas of chaos and rank which one damages the others.
- Redesenho (redesign). Feed the diagnosis back and ask for the routine rebuilt from scratch, not adjusted.
- Divisão por área (division by area). Repeat the redesign one life area at a time, choosing a model per area.
- Execução (execution). Use research tools with real sources for the points where the plan collides with reality.
- Manutenção (maintenance). Review weekly: what you managed, what blocked you, what the plan needs to change.
The reasoning behind that order is worth keeping even if you never subscribe to Adapta. When people reflect on their own routine they filter it through what they want to see, which hides the real bottleneck. A model reading a written description has no such filter, and the presenter uses that asymmetry to get a blunt diagnosis.
The method then rebuilds rather than patches: redesign the routine from scratch, split it by life area, execute against obstacles, and review weekly. Each of those four later steps needs a different kind of model behaviour, which is the specific thing the Adapta bundle is being sold to support.
Building the observation prompt that produces a usable diagnosis
The observation step asks the model to name the three biggest areas of chaos in your life and then rank which one damages the others. The exact prompt used in the video:
Descreva como está a sua rotina atual. Atue como uma consultora de organização pessoal. Identifique as três maiores áreas de caos na minha vida com base no que eu descrever e me diga qual delas está gerando mais impacto negativo nas outras. Aqui está a minha rotina: [descreva].
The presenter runs this in ChatGPT, pastes a short sketch of her week, and gets a structured response split across time, energy and attention. The demo splits the diagnosis into 10 questions across those three categories, and the answers become the input for the next step.
Three details make the prompt work better than a vague request for advice. First, you describe facts about your week rather than your interpretation of it. Second, you ask the model to identify causes rather than list symptoms. Third, you keep the output, because it becomes the input for the redesign step.
She also asks the model to export the result as a PDF, prints it, and annotates it by hand. That habit is not decoration: the observation output is the reference document every later step argues with, and a printed copy survives the week better than a chat thread you never reopen. In the Adapta interface the export is a button: ask for a PDF, download the document, choose the format, then write on it with a pen.
Redesign and division: rebuilding the week, then splitting it by area
The redesign step feeds the observation output back with an explicit instruction to rebuild the routine from scratch instead of adjusting the old one. The presenter prefers Claude here because it asks clarifying questions before answering, which suits a task where the model needs details you did not think to include.
Division by area then repeats the redesign at a smaller scale, one life area per model. The transcript walks through three worked prompts, and the pattern is the same in each: state the habit you want, state the constraint you actually have, and ask for a dated plan with checkpoints you can verify.
| Area | Prompt pattern | Constraint stated |
|---|---|---|
| Study | Learn topic X consistently, building memorisation techniques, spaced reviews and weekly targets | One hour a day available |
| Food | Improve diet without a restrictive diet, with a simple weekly menu using easy-to-find ingredients | Little time to cook during the week |
| Training | Return to training after a break, with a four-week progressive plan | One hour a day free, plus gym equipment available |
Each returns a dated plan. The study prompt asks for 30 days of memorisation techniques, spaced reviews and weekly goals you can track. The training prompt asks for a four-week progression, and the demo output is a concrete session list with retraining tips rather than a general intention. The food plan can be pushed further: ask for the ingredient list, then take that list straight to the shop.
Execution and maintenance: where most plans collapse
Execution is the step where a finished plan meets a real week, and the presenter routes research questions to Perplexity because it returns cited sources rather than unsourced assertions. A typical execution prompt asks for more information on one specific part of the routine, such as sleep, with reliable sources. The demo returns an article on techniques for sleeping better.
Maintenance is a weekly review run in Gemini, where you report what worked and what blocked you and ask what needs adjusting in the plan. The transcript's maintenance prompt is short: describe what you managed, describe what stopped you, and ask the model to identify what sabotaged the week. That produces a concrete adjustment instead of a fresh plan every Monday.
The weekly review is the part most people skip, and it is also the cheapest to run. Two limits apply to both steps. Perplexity citations are a starting point for reading, not a substitute for checking the underlying study or guidance, especially for health questions such as intolerances or dietary restrictions. And a routine plan generated by any model still has to survive your actual calendar, which no prompt can see.
Does the bundled subscription make sense for you?
The bundle makes sense if you already pay for two or more AI subscriptions and regularly switch between them for different kinds of work. The presenter's own case for it is that no single model is best at everything, and the demonstration supports that claim at the level of her workflow rather than as a general rule.
The case against it is equally simple. One login and one bill is convenient, but a reseller interface adds a layer between you and the model vendor, and convenience is the only thing that layer provides. If you mostly use one model, the switching value is close to zero.
A useful test before subscribing: list the models you would actually route work to this month. If the list is three or more, the bundle is doing real work for you. If it is one, you are paying R$99 for an interface. If you prefer the presenter's own habit of picking the most recent model in each chat, check whether the reseller exposes the same versions as the vendors do.
FAQ
- What is the Adapta AI platform? Adapta is a Brazilian subscription service that gives one login access to more than 15 AI models for a reported R$99 a month, according to the creator who demonstrated it in a May 2026 video on the channel Mariana Santos. It works as an interface layer over models such as ChatGPT, Claude, Gemini and Perplexity rather than as a model built by Adapta.
- Does Adapta include source-cited search? The demo shows Perplexity available inside the platform, and Perplexity returns cited sources. The citations still need checking against the original study or guidance before you rely on them for health, legal or financial decisions.
- Is R$99 a month for 15+ AI models accurate? That was the price and model count stated in the source video published in May 2026. Bundled pricing and included models change, so confirm the current figure on Adapta's own site before subscribing.
- Which model should I use for planning? The presenter assigns Claude the analytical planning step, ChatGPT the observation and content step, Perplexity the research step and Gemini the weekly review. She also names DeepSeek as the most logical of the set and Grok as the most direct. Treat that as one creator's working heuristic rather than a benchmarked ranking.
- Can the ORDEM method work without Adapta? Yes. The method is five prompts and a weekly review, and you can run each prompt in a separate free account if you prefer not to pay for a bundle.
From one recorded method to a written one
The ORDEM method only works because the reasoning survives the week. The video's transcripts, prompt sketches and model choices are exactly the kind of knowledge that disappears into a chat thread unless someone writes it down.
If you have explanations, interviews, lessons or opinions sitting inside a YouTube video, Skalablog can turn that recording into a structured article: paste the video URL, let it transcribe the audio, and get back a written piece built from what you actually said. That is how a method becomes something readers can find, quote and use.
Fork this article
Start a new branch from the same video, shaped your way. You keep the credit; the original keeps the attribution.
A fork in another language is filed as a translation of this article, so the two pages point at each other. You can unlink it later from the editor.
0/240
You are creating
- Format
- For
- Language
- Source
- Your angle
You will be asked to sign in before it is generated.
Buy credits