Which listings count as cross-domain evaluation, and who can apply?
This page covers a narrow set of listings: the ones whose title names evaluation itself, or more than one field, instead of a single profession. On the date in Sources, micro1’s public portal showed a handful of them: an evaluation specialist role addressed to recent graduates, a research evaluation specialist role addressed to PhDs, researchers and professors, a multi-domain technical problem author, a STEM expert role, and an evaluator for an AI assistant working in email and calendar.
That is a small group, and it is not a catch-all. A listing that names a field — chart analysis, writing and editing, psychiatry, a software product — belongs to that field, and this page does not speak for it.
The full descriptions show that the bar is not the same across the group. The recent-graduate listing says it is open to people from any academic or professional background who can show research and writing ability. The research listing asks for a PhD, active candidacy or equivalent research experience. The problem-author listing asks for an advanced engineering degree or many years in industry, plus a record of writing exam or competition problems. The STEM listing asks for deep expertise in one scientific field and working Python, and says a degree is not strictly required if the expertise can be shown. The assistant-evaluator description names no degree or credential among its requirements — which is what the description leaves out, not a statement that none is needed: it asks for people in the United States who run their own life through email and calendar, and it lists the device and accounts needed. We have no data on how many people apply to these listings, so this page makes no claim about competition.
Check your country of residence before anything else. micro1 states it does not hire residents of: Afghanistan, Belarus, China, Cuba, Democratic Republic of the Congo, Hong Kong, Iran, Iraq, Libya, Macao, Myanmar, North Korea, the Russian Federation, Somalia, South Sudan, Sudan, Syria, Venezuela, Ukraine and Yemen. Snapshot from the date in Sources — check the current page.
What do the descriptions say the work is?
Three different things. Most of it is writing, not grading.
Writing questions a model gets wrong. The two evaluation-specialist listings describe creating original, difficult question-and-answer pairs; researching each answer across several sources and documenting it with citations; testing the questions against AI systems; and making them harder when the model answers too easily. The problem-author listing describes the engineering version: authoring advanced technical problems and assessing the AI’s responses to them.
Solving and checking in a field you know. The STEM listing describes solving hard problems in your own discipline, reviewing AI-generated solutions for correctness, and using Python for computation and validation. It describes itself as coding-heavy.
Testing an assistant on everyday tasks. The assistant-evaluator listing describes giving an AI assistant real-world tasks — email, scheduling, bookings — then verifying what it did and grading it against detailed guidelines.
Opens micro1. You apply and are vetted by micro1, not by this site.
What could I practice on, for each kind of work?
The examples below are illustrative exercises created for this guide, not published micro1 assignments or interview questions. They show the kind of judgment the work needs; they are not a preview of what a project will hand you.
The four exercises are this guide’s own. Each one names the kind of work, from the section above, that it is practice for. All of the material is made by you: your questions, your notes, your prompts. Where an exercise uses a chatbot, use one you are allowed to use, under its own terms, with prompts you wrote.
1. Write a question with a researched, cited answer. Practice for: writing questions a model gets wrong. Pick a topic you know well enough to check. Write one question whose answer needs two facts from two different published sources to be combined. Write the answer, and under it list the sources by title and place, with the sentence in each that supports the step. Cite; do not paste long passages. The discipline being practiced is that every step of your answer has a source somebody else can open.
2. Test it, and make it harder. Practice for: the same work — the descriptions list testing questions against AI systems and raising the difficulty. Put your question to a chatbot. If it answers correctly, change the question — remove a hint, add a constraint, ask for the second-order consequence — and try again. Keep a log: each version, what the model answered, what you changed and why. Stop when the model gets it wrong for a reason you can name, and check that your own cited answer still holds for the final wording.
3. Solve one, then check one. Practice for: solving and checking in a field you know. Write a problem in your own discipline at the level you would set for a strong student. Solve it by hand, then verify the result with a short Python script. Then ask a chatbot for its solution to the same problem and go through it line by line: mark the first line that is wrong, and say what kind of error it is — arithmetic, a wrong assumption, a step that does not follow.
4. Audit an assistant’s actions. Practice for: testing an assistant on everyday tasks. Invent a small calendar — four meetings, three made-up attendees — and a request: move one meeting and tell the attendees. Write down what a correct result looks like before you look at any output: the right meeting, the right time zone, nobody dropped from the invite, a message that says what changed, and nothing done that was not asked. Then check a response against that list item by item, each item marked met or missed with the evidence. Use an invented calendar, not your own inbox: other people’s names and messages are not yours to put into a practice tool.
A grid for a question you wrote
This grid is the blog’s own practice tool, not micro1’s criteria. It is for exercises 1 and 2 only — the question-writing work — and it does not grade the other two. Its levels are cumulative: Strong is Acceptable plus the thing named after “also”, and the level of a row is the highest one your final question meets completely. A row you have not checked is at Fail until you check it. A question with any row at Fail is not finished; the Fail cell says what to change, and Difficulty at Fail is simply the next round of exercise 2.
| Criterion | Fail | Acceptable | Strong |
|---|---|---|---|
| Sourcing | A step of the answer has no cited source, or its quoted sentence does not support the step. Correct the citation | Every step of the answer has a cited source whose quoted sentence supports it | Also: a second person reached the same answer from your citations alone |
| One answer | A second defensible answer to the wording has been found, by you or by a reader. Correct the wording | No second defensible answer to the wording has been found | Also: the assumption that rules out the nearest alternative is written into the question |
| Difficulty | The model you tested gave the correct answer to the final version. Correct the question and test again | The model you tested gave no correct answer to the final version — a wrong answer, a partial one, or none | Also: the log names the change that made the difference |
One limit: “has been found” is as far as this row goes — it cannot say that no second answer exists. That is what Sourcing’s Strong level, a second reader, is for.
What do I put forward when I apply?
Suggestions from this guide, tied to what the descriptions ask for:
- A field you can be checked on. A degree, a former job, a research area. Three of the five descriptions ask for one, at different levels.
- Research and writing. The recent-graduate description asks for the ability to show both. Short, sourced writing is what the question-writing work produces.
- Python, if you are looking at the STEM listing. The description calls the role coding-heavy.
- The device and accounts, if you are looking at the assistant-evaluator listing. The description lists them.
If you bring the exercises above, present them as what they are: practice you set yourself, not work done for a client or a lab.
What happens at the interview?
An AI interview — step two of the four micro1 lists, between applying and onboarding — plus, per micro1, human recruiters who review profiles and match experts to opportunities based on verified skills. The pages we read do not say what the interview asks or how it is weighed.
One suggestion from this guide: be concrete. Have one of your own practice questions ready to describe — the question, the sources behind the answer, how the model got it wrong, what you changed — and say that it is a practice exercise.
How is it paid?
The cards show an hourly range, and one description repeats an hourly range in its first lines. The compensation sections of the question-writing, problem-authoring and STEM descriptions then say:
Compensation is output-based; experts are paid per task that meets the project specifications. The time required to complete work may vary depending on the expert’s experience and workflow. Minimum submission requirements apply.
So the same listing can show a per-hour figure and a per-task pay structure, and it does not say how one becomes the other. For question-writing the difference matters: a question has to be researched, cited, tested against a model and rewritten if the model answers it, and the description ties payment to a task that meets the specification. Time spent on a question that needs several rounds may lower what the work pays per hour in effect; how that time is treated is for the contract to say. The assistant-evaluator description states no pay structure, so there the unit of payment is unknown.
Ask before accepting, in writing: what one accepted task pays; who decides a task meets the specification; whether a rejected task can be fixed and resubmitted; and how many tasks the weekly minimum is. Do not plan on being paid for all the time you put in until the contract says how time is counted.
This page gives no rates; that is the blog’s choice. A separate matter: micro1’s referral terms forbid anyone promoting the program from making promises about compensation, placement, interview success or acceptance rates.
Is it a foot in the door for specialist work? Nothing we read says so, and this page will not guess.
Does a referral change how the evaluation goes? We found no sign that it does. Vetting is micro1’s process and this site has no hand in it; micro1’s referral terms prohibit anyone promoting the program from promising placement, compensation or acceptance. Promises of an easier path through a referral are a warning sign.