Hi! Here is a closer look at how to complete tests successfully with the private AI assistant chimate.ai.
1. When it is useful
1.1 Completing tests and assignments on websites
Many websites where you take tests or complete assignments try to monitor user activity. In some situations, this can be a problem: if the website detects activity it considers suspicious, it may reject your results.
ChiMate solves this problem because it runs at the operating system level, where websites cannot monitor its activity. It captures data using AI-based OCR (text recognition from images), dividing it into blocks of text. This makes ChiMate independent of the data source: it captures the text visible on your screen. There are also modes for capturing an entire page, rather than just the visible area.
1.2 Assistance during video conferences
When you are on a call, sharing your screen, and discussing a task, code, document, or interface, you may need a quick answer, a hint, or a well-structured way to phrase something. But opening an AI chat on your main screen, where colleagues or people you barely know can see it, is not always convenient.
ChiMate helps you stay focused, work quickly, and maintain privacy: you carry on as usual while AI support stays in your personal workspace.
ChiMate lets you use a special semi-transparent window (the overlay) that is invisible during screen sharing. You can control the app with keyboard shortcuts or additional mouse buttons. You can also control it through the web interface from another device, allowing you to avoid taking any actions on the main computer running the conference.
2. General recommendations for completing tests and assignments
Let us start with some general recommendations for completing tests and assignments on websites that try to monitor users. There are a few precautions to follow:
The mouse pointer should stay on the page. Ideally, the page should be in full-screen mode.
Do not switch tabs.
Do not use browser extensions, as their activity may be visible to the active test page.
All of this is easy to manage with ChiMate, but there are also some less obvious points:
Do not type an answer too quickly. Think it through first; the process should be natural.
Do not copy and paste. Type your answer manually, even if there is a lot to write.
Do not grant the website additional access to your device. It will work without that access.
3. Ways to use ChiMate
3.1 Google Chrome CDP mode
This mode launches a separate instance of Google Chrome with the Chrome DevTools Protocol (CDP) debugging interface enabled. Through this interface, the app can capture the full content of the open page, rather than just the portion visible on screen. This is useful for long assignments that do not fit on one screen.
It is best to start this mode ahead of time, sign in to the services you need, and get familiar with the interface before starting the tests.
You can use the overlay to control the app. It is undetectable by the website because it runs at the operating system level, and the website does not know it is there. This is useful on websites that try to monitor user activity and reject results at the end if they detect anything suspicious.
The overlay has buttons to take a screenshot of the visible part of the page or of the entire page.
3.2 Using two monitors
Another option, if you have two monitors, is to keep the main ChiMate app window on one monitor and open the browser in CDP mode on the other, using keyboard shortcuts to capture screenshots.
3.3. Web Session mode
In this mode, the assistant runs on one computer, and you control it from your smartphone, tablet, or any other computer through the web panel in your account at https://chimate.ai/dashboard/web-session.
There is no need to move things around or fiddle with the target device: you control screenshot capture and receive answers directly on your smartphone screen.
4. Tips
How do you know which model tier is right for your task? Several models are available, and they use different amounts of credits. This raises the question of whether the lowest-tier model is enough for a task or whether you should use a more expensive one.
Answers are presented in this format: question topic (in this example: "AWS Direct Connect Security"), question category (in this example: "test"), answer option (in this example: "A"), and answer text (in this example: "A. AWS advertises its aggregate routes to the Internet allowing anyone on the Internet to reach the router.").
If you notice inconsistencies in these details, such as answer text that does not match the answer option number, there is a good chance that the model tier is insufficient. Send the request again using a more capable model.
If the task involves code, it is best to use at least the middle model tier from the outset.