Quantum computing projects are full of forks. Simulator or real hardware? Which algorithm family for a given problem? How many repetitions of a circuit before you trust the statistics? Nobody has a settled rulebook yet, so you end up making judgment calls, and judgment calls that go unrecorded are quickly forgotten. In this how-to, we will set up a quantum experiment decision log, step by step. It is a small habit that keeps your reasoning visible and makes your next experiment smarter than your last.
You do not need a quantum computer to follow along. The log works for learning projects on a laptop simulator just as well as for research-grade work.
Step 1: Create a home for the log
Make a folder called quantum-decisions next to your experiment notebooks or code. Inside it, create one plain-text file per decision, named by date and topic, such as the date followed by “backend-choice”. Using plain text means you can search it, track changes with version control and open it anywhere. If you prefer a single running document, that is fine, as long as entries are dated and separated clearly.
Step 2: Decide what counts as log-worthy
Quantum work has many small steps, and you cannot record each one. Log a decision when it meets at least one of these conditions:
- It changes how results will be interpreted, such as choosing a noise model or an error mitigation approach.
- It would be costly to redo, for instance spending limited hardware access on a particular run.
- Someone else will build on it, like a teammate reusing your circuit design.
- You catch yourself thinking “I will remember why”.
Step 3: Use a five-line template
Copy this skeleton into every entry:
- Question: what are you trying to decide, and why now?
- Options: at least two realistic alternatives.
- Pick and reason: your choice, with the argument that tipped it.
- Prediction: what you expect to observe, in a form you can check.
- Review date: when you will compare prediction and result.
Step 4: Write a worked entry
Here is how a real-feeling entry might read. Question: should I test my small optimisation circuit on a noisy simulator before using any hardware time? Options: go straight to hardware; use an ideal simulator only; use a simulator with a noise model. Pick and reason: noisy simulator first, since it shows whether the circuit is even close to workable once noise is present, and costs nothing. Prediction: the circuit will give a clear preference for the right answer on the ideal simulator, and a visibly blurred preference with noise. Review date: after the first batch of runs.
If you are working on that type of problem, an earlier write-up on using quantum methods against tough optimisation tasks will help you frame the options. Notice that the entry does not claim any quantum advantage. It simply records what you believed and why.
Step 5: Make predictions you can check
This is where most beginners stumble. “The circuit will work” is not checkable. “The correct answer will appear in the largest share of measurement outcomes, and the gap to the second option will shrink when noise is added” is. Whenever you write a prediction, ask: could I tell tomorrow if this was wrong? If not, sharpen it. Where your circuit uses a transform, the explanation of the quantum Fourier transform as an engine inside many algorithms can suggest what output patterns to expect, and you can write those expectations into the entry.
Step 6: Set a review rhythm
Put each review date in your calendar the moment you write the entry. Add a standing monthly slot as a backstop to catch anything you missed. At review, read the original entry first, without looking at your results. Then open the results and add a note below: what happened, whether the reasoning held, and what you would do differently. Never rewrite the original, even if it embarrasses you.
Step 7: Separate luck from skill
In quantum experiments, randomness is built into the measurements, so lucky and unlucky runs are common. When you review, classify each decision in one of three ways: sound reasoning and good result, sound reasoning and poor result, or flawed reasoning. A poor result after sound reasoning is often just statistical bad luck and should not trigger a change of method. A good result after flawed reasoning is the dangerous one, since it tempts you to repeat the flaw. Your log lets you tell them apart.
A tangent worth a short paragraph
Written decisions are common beyond physics. The vehicle sharing marketplace MotoShareGo, a site where motorbikes and scooters change hands through rental and sharing, with sign-in via a central account, shows how widely everyday services rest on choices someone once weighed; quantum engineers benefit from the same discipline when they choose how to run their circuits.
Step 8: Share entries with collaborators
If you work in a group, put the folder where everyone can see it, and agree that entries are read during planning sessions. A teammate who sees your prediction can challenge the assumptions behind it before you spend time or hardware access on the run. It also makes handovers painless: a new person can read ten entries and understand why the code looks the way it does.
Step 9: Learn from the pattern
After a few months, read the whole log in one sitting. Look for recurring themes. Do you consistently expect cleaner results than noise allows? Do you favour hardware time too early? Do your estimates of runtime miss in the same direction? Write one paragraph describing what you notice then adjust your template so that it prompts you about it. For example, if you routinely forget to specify the number of measurement repetitions, add a field for it.
A note on confidence levels
Because so much in this field is uncertain, consider attaching a rough confidence to each prediction: low, medium or high, or a percentage if you like numbers. Over time you can check whether your high-confidence predictions come true more often than your low-confidence ones. If they do not, your confidence is not tracking reality, and that finding alone is worth the effort of keeping the log. It is a gentle way to build calibration without any special software, just a column in your template and a little patience at review time.
Troubleshooting your log
- Entries pile up unread: shorten them and add the review slot to a meeting you already attend.
- Predictions feel silly: that is normal while the field is uncertain; write your best guess with a rough confidence level.
- You forget to log: make the entry part of your experiment checklist, before the run, not after.
- The log gets rewritten: use version control and look at the history; the original text should never change.
Wrapping up
A quantum decision log is a folder, a template and a calendar reminder. It will not make your circuits more powerful, but it will make your thinking about them more honest and more cumulative. Start with your next experiment: write the question, the options, your pick, a checkable prediction and a date, then run it and see how well you called it.