Independent career guide

Physicists and AI Training: Solve, Audit, Adjudicate

What micro1 physics listings describe: solving research-level problems, auditing derivations and adjudicating between them, by subfield and seniority.

On this page

The physics listings on micro1’s portal ask for more than final answers. They ask for derivations — assumptions and approximations written out — and for reviews of other people’s and models’ derivations. This page reports what the descriptions say, then follows one practice problem from the prompt to the grade.

On the date in Sources, the physics listings on micro1’s public portal were not generic. They were posted by subfield — AMO and quantum information, condensed matter and magnetic materials, biophysics and statistical physics, high energy and mathematical physics, gravitation, cosmology and astrophysics — and by seniority: PhD, postdoc, junior professor, principal investigator.

Browse all expert opportunities

Opens micro1. You apply and are vetted by micro1, not by this site.

What the listings describe: three assignments

The full descriptions name three assignments — Solver, Auditor and Adjudicator — and the listings split along seniority in the same way.

Solving. The PhD and postdoc listing describes solving advanced physics problems in your specialization and delivering rigorous, documented derivations: assumptions, approximations and final results written in LaTeX, with SymPy, Python and Jupyter used for symbolic or numerical verification. It also asks the solver to point out subtleties in the problem statement — special cases, boundary conditions, dimensional consistency — to flag ambiguities, and to revise solutions after reviewer feedback.

Auditing. The postdoc and junior-professor listing describes critically reviewing solutions, derivations and theoretical arguments produced by researchers or by AI platforms: finding errors, unjustified steps, missing assumptions and dimensional inconsistencies, separating substantive problems from cosmetic ones, and writing down the reasoning behind each flaw.

Adjudicating. The professor and principal-investigator listing describes judging between competing arguments or solutions: which approach is better, under what assumptions and in which regimes, where an approximation breaks, and when the honest answer is that the field has not settled the question.

The subfield listings describe research-level benchmarks on named topics, with the assignment depending on subfield fit, seniority and hands-on experience with the methods. The descriptions ask for a PhD or advanced candidacy and recent publications with arXiv or DOI records, and one is focused on the US, Canada and the UK.

The case below is the blog’s own illustration. It is practice for the adjudicating assignment — two derivations, one verdict — and uses the auditing method, line by line, to get there. It does not practice the solving assignment; a suggestion for that is further down.

One problem, two derivations

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 problem, the two abbreviated derivations and the grading are this guide’s practice material. The Debye limits and the quantum oscillator’s equipartition condition have separate references in Sources. The worked comparison applies those results to the stated regime; it is not a specialist-reviewed research result.

The prompt, written by the expert: find how the lattice heat capacity of an insulating crystal depends on temperature, at a temperature far below its Debye temperature. The stated condition is the point of the problem. It is a regime-validity probe: the classical approach is wrong because of one clause, and the pass criterion — the answer must scale as the cube of the temperature and say why — is written before the model runs.

Two generated derivations come back, and the task is to decide which is correct and where the other diverged.

Solution A, and why it survives a quick check

Solution A treats each atom as three independent oscillators, assigns each quadratic degree of freedom an average energy of k_B T/2 by equipartition, sums, and differentiates. It arrives at 3Nk_B, the Dulong–Petit value.

A quick read passes it. Every algebraic line follows from the one before, the units are those of heat capacity, and the result is the classical value. Solution B is longer: a Debye density of states, an integral, a limit.

Where the review stops reading

At the second line of A. For a quantum harmonic oscillator, the classical equipartition limit requires k_B T to be large compared with the spacing between adjacent energy levels, hbar omega (Fitzpatrick, equations 5.390–5.395, in Sources). The spacing is not the oscillator’s total energy. At low temperature, modes with spacing much larger than k_B T contribute little to heat capacity; their ground-state energy does not vary with temperature. The Debye module explains that only the low-frequency acoustic modes are appreciably excited far below the Debye temperature. Applying the classical value to every mode, as A does, is therefore outside its regime.

The order-of-magnitude check confirms it independently of the algebra. A’s answer does not depend on temperature at all, so it overstates the heat capacity by a factor that grows as the crystal gets colder. B’s Debye treatment gives the cube-law scaling the problem asks for.

B is not beyond comment either. It extends the upper limit of the Debye integral to infinity, which is justified in this low-temperature limit — and B never says so. That is a valid step that needs care, and the review should note it.

The adjudication, in one sentence: A applies equipartition to modes whose level spacing is far above k_B T at the stated temperature, so it overstates the heat capacity; B’s low-temperature Debye limit is the valid result.

Grading line by line

Step-level grading marks each line and classifies the first one that fails. Here that is A’s equipartition step, classified as an approximation used outside its regime. A grade for the whole answer would not say where A went wrong.

The grid below is this blog’s own practice aid, not micro1’s rubric or a hiring criterion. Its levels are cumulative: Strong is Acceptable plus the thing named after “also”, and a row’s level is the highest one the derivation meets in full. If the problem states no regime, the Approximation row cannot be read: clarify the problem first. A derivation with any row at Fail is rejected, with the action its Fail cell names.

CriterionFailAcceptableStrong
AlgebraAt least one line does not follow from the one before, even granting the approximations it invokes. Correct from that lineEvery line follows from the one before, granting the approximations it invokesAlso: flags the step that needs care, or says none does
ApproximationAn approximation is used outside the regime the problem states. Redo from the step that uses itEvery approximation is valid in the stated regimeAlso: the regime and its boundary are stated
ResultThe temperature dependence or the dimensions are wrong. Re-assess the steps that lead to itBoth the temperature dependence and the dimensions are rightAlso: the result is checked against a limit or a known case

Applied to the two derivations: A is Acceptable on Algebra — once the equipartition step is granted, each line follows — and fails Approximation and Result, so it is rejected. Its action is Redo from the second line: everything after the equipartition step rests on it, and fixing the Result row alone would leave the cause in place. B is Acceptable on all three rows and reaches Strong on none: it never flags the infinite upper limit as the step needing care, never states the regime boundary, and checks no limit. B is accepted, with those three notes.

A practice version for your own subfield: generate a derivation with a model you are allowed to use, on a problem you wrote, and grade it with a grid like this one. If you show it to anyone, say it is a practice exercise.

What the work looks like beyond this case

For the solving assignment, which this case does not practice: write a problem at the edge of your subfield and solve it the way the description asks — assumptions, approximations and result written out in LaTeX, with a symbolic or numerical check — then list the special cases and boundary conditions the problem statement leaves open.

Is this for you

Physics PhDs, postdocs and faculty — and advanced PhD candidates, whom the solving listing names explicitly. The auditing listing asks for postdoctoral or junior-faculty experience and a record of reviewing others’ work; the adjudicating listing asks for a professorship or group leadership. Read the qualifications of the one that fits your career stage.

The listings are posted by subfield and by seniority, so state both: “condensed matter, quantum information, six years post-PhD” says which posting you fit. Whether this contributes to automating research is a fair thing to weigh; this page will not argue you out of it.

Institution, unpublished work, rates

  • Check your institution’s outside-work policy. It may require disclosure or prior approval; read it before you accept, not after.
  • Do not contribute unpublished results. Material from an unpublished manuscript or a collaboration’s internal work is not yours to hand over. The case above is built from textbook physics for that reason.
  • Ask about attribution and ownership; do not assume. None of the descriptions we read says who owns the problems, solutions or reviews you write, or whether your name is attached to them. If publication record is what your career is measured on, read the contract for what it says about intellectual property and confidentiality, look at the concrete activity, and check your institution’s and your field’s rules on outside work and on reusing material.

On pay, the listings say less than they seem to. The cards on the opportunities page show an hourly range. The descriptions give time, not money: a few state a weekly schedule, and one adds

Availability to engage with the project consistently over an 8–10 week period (approx. 10 hours/week).

None of the physics descriptions we read says whether pay is by the hour or per accepted solution or review, so the unit of payment is unknown: the hourly figure on a card does not settle it either way. A research-level derivation can take far longer than planned; if pay turns out to be per accepted piece of work, that extra time may lower what it pays per hour in effect. Before accepting, ask in writing what the unit of payment is, who accepts a solution and on what criteria, what happens to work sent back for revision, and whether the stated weekly hours are a cap, a minimum or an estimate. Do not count on every hour being paid until the contract says so.

That this page names no rate is our decision. What the referral terms do is something else: they forbid anyone promoting the program from making promises about compensation, placement, interview success or acceptance rates.

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 against project requirements based on verified skills. One subfield description mentions adaptive, spoken AI interviews answered one question at a time, with concrete examples from research you are legally free to share.

  1. Lead with subfield and seniority, in the vocabulary the listings use.
  2. Have one derivation you can narrate in ninety seconds: setup, the step that matters, the check you ran at the end. Published work of yours, or a practice problem — said to be one.
  3. Practice describing a mistake you caught — in your own work, or in something published. Leave out anything you saw as a referee or inside a collaboration.
  4. Be precise about what you would not grade. The descriptions assign work by subfield fit.

Countries, and the referral question

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 list before applying.

Is a referred physicist assessed any differently? We found nothing that says so. The vetting is micro1’s and this site has no part in it; the referral terms forbid anyone promoting the program from promising placement, compensation or acceptance. Read any page that promises a referral advantage as a warning sign.

Browse all expert opportunities

Opens micro1. You apply and are vetted by micro1, not by this site.

Sources

Programme terms, eligibility rules and reference pages change. Each claim above is dated; verify it on the page it cites before acting on it.