Product
Hyna is the desk, not the chatbot. You join a simulated company, take the assignment, and move forward only when the work you submit holds up.
NovaCart
Maya · Manager · AI
Sales were higher last week, but we are not sure which channel helped. Clean the export and explain what you changed.
| order_id | channel | status | note |
|---|---|---|---|
| 10482 | paid search | completed | keep |
| 10488 | refunded | keep | |
| 10491 | canceled | duplicate id |
Change log · the older copy of 10491 still needs a decision.
One sitting
Each mission is short enough to finish and specific enough to check. About 30–45 minutes on the first track.
Maya tells you what the store needs, what to turn in, and what “done” has to include. The company is fictional. The assignment is not vague.
The export is messy. After you have read the brief, Leo can share an approved note about a known source issue. He cannot invent a number that is not in the file.
On this track: a cleaned CSV and a change log, then KPI values with a written explanation, then a budget recommendation and a memo. The checks read what you uploaded.
If something fails, the review points at the row, field, or sentence. You fix that and submit again. Earlier attempts stay on the record.
Who you work with
Manager
Gives the assignment, clarifies what matters first, and chooses approved practice when you miss something. She does not reveal the answer key or offer you a job.
Operations
Explains the data dictionary and the operational constraints. He can describe a released source issue. He does not invent values or quietly change the dataset.
Stakeholder
Asks for a decision and brings an approved change, such as a smaller budget. Sam does not read hidden tests or another learner’s work.
The review
Code checks the objective parts. A separate review judges clarity against a fixed rubric and has to cite your submission. A passing mission needs both.
What is checked
Duplicate orders, required columns, canceled rows, budget arithmetic. These do not depend on how fluent the memo sounds.
Reasoning and clarity are reviewed with a quote from what you submitted. A fluent memo with the wrong figures does not pass.
Hints are tracked so the summary can tell assisted practice from independent work. Chatting more does not raise the score.
Boundaries
Questions
No. NovaCart, its customers, and Maya, Leo, and Sam are fictional. Hyna is a place to practice the work, and the product says so inside the simulation as well as here.
On the first track: a cleaned CSV and a change log, then KPI values with a written explanation, then a budget recommendation and a memo. The pilot does not ask you to upload PDFs, spreadsheets, or code.
No. Coworkers can clarify released facts and point you at the task. Credit depends on the artifact and the checks. Hints are tracked so the summary can tell assisted practice from independent work.
Code checks the objective parts, such as duplicate orders and budget arithmetic. A separate review judges reasoning and clarity against a fixed rubric and has to cite your submission. A passing mission needs both the numbers and the critical checks.
You can read, message, and review feedback on a phone. Editing the CSV is built for a desktop. The product will say that, rather than pretending a phone is the right desk.
The closed pilot is free and invite-only. Paid access is a later question, after the first track has been used. Writing to hello@hyna.io asks for a seat. It does not buy one.
No. Completion is not a job offer, an accredited qualification, or a promise that someone will hire you. It is a record of simulated work and the feedback on it.
The first cohort is small. Tell us what you want to practice. We read hello@hyna.io.
Request an invite