Understand the tool before you depend on it.
Learn what a model can help with, what a prompt changes and where human judgement remains essential. The examples focus on clear requests, private information and verifiable results.
Understand generated answers as drafts to inspect
A model produces a response from its instructions and available context.
Updated 21 October 2026 · 2 min read →Every article on this topic
Begin a useful conversation with a model
Tell the assistant what you need and provide enough context to judge the result.
2 min read →Keep related conversations together
A folder can provide shared instructions without making every conversation public.
2 min read →Use the daily text allowance thoughtfully
Clear requests and relevant context help you get more useful work from the same allowance.
2 min read →Write a request you can evaluate
A good prompt names the goal, context and success criteria.
2 min read →Choose what information to share
Only provide the information needed for the task and selected provider.
2 min read →Check an answer that sounds convincing
Fluent wording is not evidence that every claim is correct.
2 min read →Know when you need a source and when you need a draft
Search retrieves material; generation helps organise or transform it.
2 min read →Work in Bengali with clear context
Use natural Bengali and check the output in the form you intend to share.
2 min read →Compare free tools by the work they let you finish
Look beyond a headline allowance to limits and export options.
2 min read →Understand a language model’s role in the workspace
The model helps produce content while the application manages records and actions.
2 min read →Use new tools to strengthen a working skill
Choose a task from your existing work and improve how reliably you complete it.
2 min read →Make freelance work reviewable and dependable
Agree on the deliverable and how it will be checked before starting.
2 min read →Turn a useful skill into a concrete service
Start with a customer need you can satisfy and verify.
2 min read →Keep track of the material you may use
Rights depend on the material and applicable terms, not merely on how it was edited.
2 min read →Plan around what the tools can do today
Make decisions from observed capability rather than predictions.
2 min read →Separate visible behavior from claims about a system
A convincing conversation does not settle questions about inner experience.
2 min read →A clearer brief makes a better beginning
Start with a person, a problem and a result you can check.
2 min read →Questions and answers
How much free text use is included?
Every account has 400,000 input and output text tokens each day, resetting at midnight in Asia/Dhaka. Separately priced services use wallet credits. A personal provider key uses its own billing.
Does a model search the web automatically?
Only a configured search or reference tool can bring in current external material. A generated answer by itself is not evidence that a live source was checked.
What happens when a request is interrupted?
An uncertain provider outcome may reserve or count an estimated amount to prevent duplicate paid work. Check usage and the saved result before submitting the same task again.
Step-by-step guides
The handbook explains the controls you can use in your workspace.