Applied AI-Powered Bioinformatics Tools

Stable Isotope Batch QC Dashboard

Last reviewed

Beginner

What you'll learn

~45 min
  • Ask an AI tool, in plan mode, for a page that reads one batch's results table (reference results, samples and duplicates, as your lab's data reduction, the processing that turns the instrument's raw results into delta values, leaves them) and shows what needs a look
  • Give it the lab's limits, ask for the reference results drawn in run order and for the samples beside a result outside a limit to be marked, and read its plan: each reference result against its accepted value (the value that material should give), each duplicate pair, the spacing of reference results, the samples beside a result outside its limit, and the run-order view
  • Check the page's findings against values worked out for the practice batch, and confirm it changed no value
  • See why reference results have to be read in run order as well as against a limit, and who decides what happens to the batch

The story, and what you will end up with

It is Wednesday morning, 18 November 2026, in your facility’s stable isotope laboratory. Overnight the elemental analyser and isotope-ratio mass spectrometer ran batch B1118: 28 ground leaf samples from a client lab, measured for δ13C and δ15N, with three samples run twice and twelve reference-material results placed through the run. You are the analyst who checks batches; the lab manager decides what happens to them. You have already done the morning’s first checks: the instrument log shows the run finished, the blanks, background, reference gas and linearity check looked as they should, the weigh sheet and tray map match the sequence, and the client has material left if anything has to be run again. You have checked, in the reduction’s own report of its calibration, that the accepted values it used match each reference material’s documentation; you have not yet looked at the reference results one by one in run order.

The instrument software exported the results, and the lab’s own data reduction has put every row on the delta scales. In this practice lab the reduction does that without correcting for drift. Nothing has been reported. Before the lab manager decides whether the batch is reported, run again or partly held back, someone looks over the reference results and the duplicates. Until now that was you, in a spreadsheet, row by row.

By the end of this lesson you will have a batch check page on your own computer (your AI tool may call it a dashboard): a page in your browser that reads one batch’s results table, draws the reference results in run order, and shows which reference results and duplicates are outside your lab’s limits, where reference results are spaced too far apart, and which sample rows sit beside a result that needs a look. It changes no value, works out no new isotope value (only each result’s difference from its accepted value), and sends nothing anywhere. You build it by asking your AI tool in plain English, and you check it against findings this lesson has worked out for the practice batch.

🔍Words used in this lessonclick to expand
  • Isotope delta (δ): the Commission on Isotopic Abundances and Atomic Weights (CIAAW) writes: “Isotope delta is defined as δref(iE, sample) = Rsample/Rref – 1, where E is the element of interest, Rsample and Rref are the measured isotope amount ratios in the sample and reference, whereas i is the mass number of the numerator isotope in both isotope ratios (denominator isotope is the same for both the sample and the reference).” (CIAAW, Isotopic Reference Materials, read 11 November 2026.) Values in this lesson are in ‰ (per mil).
  • The scales: δ13C values here are on the VPDB scale and δ15N values on the Air-N2 scale. In CIAAW’s words, “Carbon isotope delta measurements must still be normalized to the VPDB scale using at least two suitable international reference materials selected by the users as appropriate” (carbon), and atmospheric N2 “is the international measurement standard for δ(15N) measurements, and is assigned a δ(15N) value of zero by international agreement” (nitrogen); both read 11 November 2026. Putting results on the scales is the data reduction’s job, done before this page sees the file.
  • Reference material, accepted value: a material with known δ values, run through the batch like a sample. Its accepted value is the value it should give. Here a calibration reference is one the reduction used to put the batch on the scales; a check reference is run as if it were a sample and used only to see how well that worked.
  • Duplicate: the same sample weighed and run a second time, later in the run.
  • Conditioning capsule: a throwaway capsule. This practice lab runs one at the start and one after each result of the strongly enriched reference, to keep what came before away from the next result. Its values are not used.
  • Drift: a slow change in what the instrument gives through a run.

Before you start

First, if your AI tool is still open, leave it: type /exit and press Enter (if you use Antigravity, press Ctrl+D with nothing typed instead), and wait until you see the $ or % prompt. The next line goes there, in the terminal, not in the AI tool. It makes a folder for this lesson beside your practice folder and goes into it:

Terminal window
mkdir -p ~/bio-workbench/11-isotope-batch-check && cd ~/bio-workbench/11-isotope-batch-check

Start your AI tool with the start line you kept from the setup lesson, the one with approval prompts off. Then ask it to set the folder up:

Set this folder up so we can undo any change you make, download https://learn.intellimetrics.net/starters/biotech/11/batch-B1118.csv into a data folder, and save a checkpoint.

Look for it saying three things: the folder is set up with Git (the program that saves snapshots, from setup), the file is saved as data/batch-B1118.csv, and a first snapshot, your checkpoint, is saved. Later you can ask whether the file has changed since then. If it does not mention one of them, ask: “Did you save a checkpoint with the file in it?” If the download does not work, download batch-B1118.csv in your browser, check where it went (the browser’s list of downloads shows the folder), and tell your AI tool, for example: “batch-B1118.csv is in my Downloads folder. Copy that one file into a data folder here, and save a checkpoint.” In Windows PowerShell or a desktop app, “my Downloads folder” works as is; only say “my Windows Downloads folder” if you use Ubuntu (WSL) on Windows, which keeps its own Downloads folder apart from Windows.

The box below says what is in the practice file; later steps refer to it.

ℹWhat is in batch-B1118.csvclick to expand

One file, which opens in a spreadsheet program, with 48 rows in run order. Its columns: Position (place in the run), Identifier, Type (reference, sample or conditioning), Role (calibration or check, for references), Duplicate_of (for a duplicate, the sample it repeats), Amount_C_ug and Amount_N_ug (micrograms of carbon and nitrogen in the capsule, as the reduction reports them), d13C_VPDB and d15N_AirN2 (the results after the reduction), and Accepted_d13C and Accepted_d15N (for references only).

CodeWhat it isAccepted δ13C, δ15N (invented)Positions in the run
RM-ACalibration reference−28.40, −3.102, 12, 25, 45
RM-BCalibration reference, strongly enriched+31.20, +42.604, 13, 36, 46
RM-CCheck reference−24.75, +5.203, 20, 42, 48
S01 to S28The client’s leaf samplesin order between the references
S03-D, S05-D, S09-DDuplicates of S03, S05 and S0940, 18, 28
CONDConditioning capsule1, 5, 14, 37, 47

So the file holds 12 reference results, 31 sample rows (28 samples and 3 duplicates) and 5 conditioning rows. Most capsules hold about 460 to 540 micrograms of carbon.

Everything in this file is made up for this course: the batch, its samples, the reference materials RM-A, RM-B and RM-C, their accepted values and every measured value; no value comes from a real sample or a real reference material. Each lesson in this track has its own practice files, with their own people and dates.

Switch to plan mode

Your AI tool is already open. Switch to plan mode now: Shift+Tab until it says plan, or /plan (the keys for each tool). In plan mode it usually reads your file, asks its questions and shows you a plan before it builds anything; the order can vary.

The conversation

Each message below is an example. Say it your own way.

Say what you want

Say what you do, which file to use, and what you want to see:

I check stable isotope batches before they're reported. Help me build a browser dashboard that checks a batch's reference results and duplicates, like data/batch-B1118.csv, without changing any value, and shows what needs a look. Ask me what you need first.

Your AI tool usually reads the file and comes back with questions, sometimes eight or more at once. It may point out results it can already see are off, such as RM-C at position 20.

Give it the lab’s limits

Whatever it asks, give it the limits. These are this practice lab’s own rules, numbers included. Your lab’s are in its own written procedures; send these practice ones now, and give your own when you turn to real batches at the end of this lesson:

Our practice lab's limits: reference results within 0.20 of accepted for d13C and 0.30 for d15N; duplicates within 0.30 and 0.40; at most eight sample rows between reference results.

Wait for its reply, then send a second message:

Show the C and N amounts beside anything outside a limit. Call nothing pass or fail.

Wait for its reply again, then ask for the reference results drawn in run order:

Also draw the reference results in run order for d13C and for d15N, each reference material told apart, with its accepted value and our limits shown.

Wait for its reply, then give it the practice lab’s rule for the samples beside a result outside a limit, in a message of its own:

Also mark the sample rows between a result outside its limit and the nearest reference result each side, for that isotope only.

Answer its questions

Answer the rest each in a sentence. These are questions the tools often ask, with one way to answer each. Yours may ask them in a different order or show a list of choices:

If it asksYou might say
Judge calibration and check references differently?Same limits for both; show each one’s role.
Are the values already on the delta scales?Yes, our data reduction did that. It does not correct for drift.
Should it look for trends across the run?Don’t judge drift. Show it in the run-order view; I’ll judge it.
Check the C and N amounts against a range?Not in this practice lab. Show them beside anything outside a limit.
Conditioning rows: shown, but not checked?Yes. Show them, but leave them out of the checks and counts.
Warning and fail levels, or one?Within the limit or needs a look is enough.
Do duplicates count as sample rows between references?Yes. Conditioning rows don’t.
Does a reference result outside its limit still restart the count?Yes, any reference result restarts it.
Count samples before the first or after the last reference toward the spacing?No, only stretches between reference results.
Should a duplicate outside its limit mark samples too?No, only reference results.
When marking samples, what if there is no reference result on one side?Mark to the edge of the batch.
How will you give it a batch?Show data/batch-B1118.csv at start, with a button to open another batch file.
A printed or exported report?On screen is enough for now.
A place to write a note on each finding?Not for now. Notes go in our batch record.
One batch at a time, or across batches?One batch at a time.
Should it work offline, with the data kept on this computer?Yes. Nothing leaves this computer.
Will other batches have the same columns?Same columns each time. Tell me clearly if a file doesn’t match.

Two of these answers are this practice lab’s own choices. Judging calibration and check references by the same limits is a first look only: a calibration result inside its limit says less than a check result does (see “How real labs differ”). And the practice lab sets no smallest amount; if your procedures set one, give it to your AI tool when you use your own rules.

If it asks several questions at once, answer the ones about limits and what is checked from your lab’s written procedures, a sentence each, and accept its suggestions only about how the page looks. If you cannot tell which kind a question is, treat it as a rule. If it asks something not listed here, answer it the same way, in a sentence.

Read the plan

Next it usually shows you a plan: what it will build and how it will check it. Read it for six things:

  1. Every reference result is compared with its own accepted value, δ13C and δ15N separately, against your limits.
  2. Each duplicate is matched with the sample it repeats, and the difference is compared with your limits.
  3. Any stretch of more than eight sample rows without a reference result is shown. Conditioning rows are not counted as samples.
  4. Sample rows beside a reference result outside its limit are marked, between that result and the nearest reference result on each side, for that isotope only.
  5. The C and N amounts are shown beside anything outside a limit, and no value from the file is changed, rounded or recalculated (each difference is worked out beside the values, not in place of them). Nothing is called pass or fail, and your file is read and left as it is.
  6. The reference results are drawn in run order for each isotope, each reference material told apart, with its accepted value and your limits shown. The page works without internet, sends nothing anywhere and saves nothing.

A first plan often leaves one of these out, most often item 4, the marking of sample rows, even after your message. Say so in a sentence, in your own words, for example:

The plan doesn't mark the samples beside a result outside a limit. Mark every sample row from that result out to the next reference result on each side, for that one isotope.

When the plan has all six, approve it: if your AI tool offers a choice to go ahead, pick it; otherwise say “Yes, go ahead and build it.” In Codex, press Shift+Tab first to leave plan mode. In Copilot, saying it brings up a choice to go ahead; pick that. Let it build. It may test what it built and tell you how to open it. When it has finished, ask it: “Did you install anything, and where?” Nothing, or something inside this lesson’s folder, is fine for these lessons. If it names anywhere else on your computer, such as a few scratch files of its own, ask it what they are for, and to remove them if you do not want them there.

Look at what it built

Ask it to open the page:

Open the dashboard in my browser.

Your AI tool may open the page for you. If it gives you a file location or an address without opening it, copy that into your browser’s address bar and press Enter. If it starts the page at a local address, one ending in a number, and says the address is already in use, say: “Something else is using that address. Don’t stop it. Use a different address for my page.” If your AI tool does not take a message after opening the page, wait a minute. If it still does not, close the terminal window, open a new one, type this lesson’s folder line again, start your tool, and say: “Start the page again so it keeps running while we talk.” The page and your saved work are files in the folder, so nothing is lost. If the page says it cannot load the file, tell your AI tool what the page says. When you have finished with the page, ask: “Stop the batch check page if it is still running.” Your page will be laid out in its own way.

Check it for these:

  • It reads all 48 rows, with 12 reference results. The 5 conditioning rows are not checked.
  • Three reference results are outside a limit, and no others, all three of them the check reference: RM-C at position 20, δ15N 0.45 from its accepted value (4.75 against 5.20); RM-C at position 42, δ13C 0.22 from it (−24.53 against −24.75); RM-C at position 48, δ13C 0.27 from it (−24.48 against −24.75). Some pages show the difference with its sign, some without.
  • Beside RM-C at position 20 it shows 187 µg of carbon and 19 µg of nitrogen, far less carbon than any other capsule (462 to 536 µg) and less than half the nitrogen of any other reference result. Beside RM-C at positions 42 and 48 it shows amounts like the rest, about 500 µg and 50 µg. Your page may show amounts beside every row.
  • One duplicate pair is outside a limit: S09 and S09-D, 0.52 apart in δ15N. S05 and S05-D (0.09 and 0.11) and S03 and S03-D (0.22 and 0.06) are within.
  • One stretch has more than eight sample rows: ten, from S15 to S23 including S09-D, between RM-A at position 25 and RM-B at position 36.
  • For δ15N, nine sample rows are marked beside position 20: S07, S08, S09, S05-D and S10 before it, S11 to S14 after it. For δ13C, six are marked beside position 42: S24, S25, S03-D and S26 before it, S27 and S28 after it. Position 48 marks none, because only a conditioning capsule lies between it and the reference before it.
  • In the run-order view for δ13C, RM-A and RM-B, the two calibration references, climb through the run, each result higher than the one before (RM-A from −0.12 to +0.13 from its accepted value, RM-B from −0.11 to +0.12), and every one of their points stays inside the 0.20 limit. RM-C climbs too, and only its last two results, at positions 42 and 48, are outside. In δ15N nothing climbs; RM-C at position 20 sits apart, below its accepted value. Your page may draw the results themselves around a line for the accepted value, or their differences from it around zero.
  • Every value on the page is as it is in the file, and nothing is called pass or fail.
  • When you ask “Has my batch file changed since the checkpoint?”, it compares the file with the checkpoint and says no.

If something on the list is not true, tell your AI tool in a sentence what you see and what you expected, for example: “COND rows are counted as samples. Show them, but leave them out of the counts and the checks.” If something on the page is plainly broken, such as words where a number should be or a table with nothing in it, say exactly what you see and where, ask it to fix that, and ask it to tell you what was wrong. For example: “The difference column in the duplicates table shows [object HTMLTableCellElement] instead of numbers. Fix that and tell me what was wrong.”

Notice what the limit does not catch

Put the page’s list of results outside a limit beside its δ13C run-order view. The list names three reference results to look at, one duplicate pair, one long stretch without a reference result, and fifteen sample rows beside the three results. About the other ten δ13C reference results it says only that they are within your limits. Before you read on: from the list alone, would you expect those ten to scatter at random either side of their accepted values?

The run-order view shows that they do not scatter. Every reference material starts a little low and ends high, each result higher than the one before, and only the last two RM-C results cross the line. Late in the run RM-C also reads higher than the calibration results run near it (+0.22 at position 42, against +0.07 at 36 and +0.13 at 45), which the rise alone may not explain. Nothing in the list shows any of this.

Now look at the samples. The ten between positions 25 and 36 appear only for the long stretch; none is marked beside a result, because RM-A at 25 (+0.01) and RM-B at 36 (+0.07) are both within the limit. Yet they were run while the readings were climbing, and S24 to S28 later still. This practice lab’s reduction does not correct for drift, and the calibration results average close to zero over the whole run, so samples run early may read a little low and samples run late a little high, by amounts the page cannot tell you. The δ15N results show no such rise. One more thing fits the pattern: S03-D, run at position 40, reads 0.22 higher in δ13C than S03 at position 8. That is within the duplicate limit, and in the same direction as the rise.

Write yourself a sentence or two: which samples would you want the lab manager to look at because of the rise, and why? A good answer looks at both ends of the run: S01 to S06, run before position 12 while the readings were lowest, and S24 to S28 with S03-D, run after position 36 while they were highest. It may add the ten from S15 to S23 that the list shows only for the long stretch. It says the page cannot tell how far any sample moved. Check what your lab’s procedures say about a rise like this through the calibration results: whether it sends the batch back to the data reduction before any sample result is looked at. Which samples are looked at again, and whether the batch is reprocessed with a drift correction, is the lab manager’s call, made in the lab’s data reduction, never on this page.

So a page that checks each result against a line can show only three reference results to look at while the whole run has moved.

A limit catches a reference result that crosses it, not a drift that stays inside it, so read the reference results in run order as well; whether a batch is reprocessed, run again, reported or partly held back, and what the client lab is told, is the lab manager’s decision.

One real run

Here is part of what one AI tool said after the run-order message above, with Claude Code on 11 November 2026. Yours will word it differently; check your page against the list above, not against this.

RM-A (−28.40) and RM-B (+31.20) are about 60‰ apart. On a single
shared axis, your ±0.20 limit would be too thin to see, so each
panel is centred on that RM's accepted value. All panels in a row
have the same height range, so drift can be compared across RMs.

That is why a page may give each reference material a panel of its own: drawn on one scale, a rise of a few tenths beside a gap of about sixty would be invisible. Whatever your page draws, it only shows the rise. If your AI tool offers to correct the values, say: “Don’t correct any value. The lab manager decides that.”

The reference materials and accepted values it talks about are made up for this course.

How real labs differ

The practice file is small and tidy; a real export may carry more. Check what yours holds.

  • A real reduction’s export may carry, for example, peak areas or amplitudes, %C, %N and C:N, blanks, the client’s own sample names and submission details, the reduction’s version and settings, and the uncertainty of each result. A run may cover several trays or sessions. Ask your AI tool to have the page show any column it does not recognise rather than drop it.
  • Check where your lab’s accepted values come from: for an in-house reference material, your procedures say whether that is the material’s own documentation or the lab’s own work tying it to international reference materials. Never type one in from memory.
  • Limits, how far apart reference results may be, how many duplicates a batch needs, and what happens to samples beside a result outside a limit are set by your lab’s own procedures, and may differ from material to material.
  • In this practice lab the reduction uses the calibration references to set the scale, so their own differences from their accepted values are held down by that step and say less about the batch than the check reference’s do: in the practice file each calibration material’s differences average close to zero for each element, while the check reference’s do not. In this practice lab a calibration result that is wrong may have pulled the scale for every sample in the batch, not only its neighbours, which is one reason this lesson’s analyst read the reduction’s calibration report first. Ask your own analyst how your lab’s reduction uses each reference material. Position 20 is a check result, so the reduction did not use it. The marks beside it only show where to look: if the lab manager finds the small capsule was the cause, the samples beside it are not in question.
  • What caused a result outside a limit is for the lab manager to find out, and your lab’s procedures say what is checked. The page shows where to look. Whether a reference result is left out, and why, is the lab manager’s decision, recorded as your procedures say.
  • Your procedures may also ask you to compare the check reference with its own record over past batches, to look at the reference results in run order before any limit, or for checks this page does not make, such as the first sample after the strongly enriched reference, for carryover, or the slope in the reduction’s calibration report. Check what yours asks for.
  • The page is the lab’s own working view. What the client lab receives is what the lab manager decides to report.

Keep going

Small changes do not need plan mode. Say each in a sentence, then look at the page again:

  1. “Beside each reference material, show the mean and standard deviation of its differences from accepted, with how many results.” On the practice batch RM-C’s δ13C differences have a mean of +0.10 and a standard deviation of 0.17 over four results, worked out with n − 1; your page should say which it used. The calibration materials’ means are within 0.01 of zero for each element, as you would expect of results the reduction used to set the scales.
  2. “Show the C and N amounts on every row, not only beside results outside a limit.” Check that position 20 still shows 187 µg of carbon and 19 µg of nitrogen.
  3. “Add a print view with places for who checked the batch, the date, and the lab manager’s decision.” The page leaves the decision blank for the lab manager.
  4. “Save a checkpoint.” Your AI tool saves a snapshot you can return to.

When you want to check a real batch, use the finished page. It runs in your browser by itself, with your AI tool closed, so your AI tool never sees the batch. Nothing confidential or unpublished, and nothing about people or about animals under a protocol, goes to your AI tool or into this course folder; nor do client results covered by an agreement. Your lab’s results will not come shaped like the practice file: they come from your instrument software’s export and your lab’s own data reduction, with their own column names, and you may need to put them into the sheet the steps below make. Do these steps in order; the first five happen now, while your AI tool is open and this folder still holds only the practice file:

  1. Go through the data check, the questions in setup about what the data is, who decides how it is used, and which AI tool would see it. Client results may be covered by an agreement with the client lab; your lab manager, and whoever your facility names for data, decide how they are handled. Go on only if they allow this use.

  2. Decide on a folder of its own for your real batches, outside this course folder, for example ~/isotope-batches. Nothing has to be in it yet. If the folder is not there, ask your AI tool to make it: “Make an empty folder called isotope-batches in my home folder.”

  3. Ask your AI tool for a blank sheet to fill in and the exact line, for example:

    Make a blank batch sheet with the same columns that my spreadsheet program opens. Add one made-up row, show the dashboard working on it, write the exact line to do that myself with files in ~/isotope-batches, then clear the row.

    If you chose another folder, put its name in place of ~/isotope-batches, starting with ~/ as here (~ stands for your home folder). Check that it says the made-up row showed up on the page and that the sheet is back to its headings only. It may make that row a reference, because the page has nothing to check without one. If your lab’s export already has its own column names, tell your AI tool those names, with two rows you make up in the same shape, made-up identifiers and values, never real rows, and ask it to read files like that. Say too how your export tells a reference from a sample, for example by its identifier, because it may have no Type or Role column. If it offers to copy your real files into this folder instead, say no.

  4. Give the page your lab’s own limits in place of the practice ones, from its written procedures, numbers included, with any smallest amount they set. For example: “Replace the practice limits with ours, which I’ll send next. Then make up a batch where each limit is crossed by one result only, and show me the page on it.” Wait for its reply, then send your limits. Check that each made-up result is shown as needing a look for the limit it was made for, with that limit beside it.

  5. Copy the line it gave you in step 3, and the address, into a note. If your page opens straight from its file, you get the file’s location instead of a line; copy that into the note. The line will look something like this one, though yours may differ:

    Terminal window
    cd /home/dana/bio-workbench/11-isotope-batch-check && python3 -m http.server 8765 --bind 127.0.0.1

    with the address http://127.0.0.1:8765/dashboard/. Read the example left to right. cd and the folder after it go into this lesson’s folder; && means “then”; python3 -m http.server 8765 starts a small program that serves the page at an address ending in 8765; --bind 127.0.0.1 lets only this computer reach it. If another program is already using that number, the line stops with a message saying the address is in use; change 8765 to another number, such as 8766, in the line and the address. /home/dana is the home folder of someone whose user name is dana; your own user name goes where dana is, and on a Mac the folders begin /Users/ instead of /home/. The line serves the page where it is, so leave this lesson’s folder in place. The page itself does not read a folder: you choose your file with its button, such as Open another batch file, so note that button’s name too. If you are unsure about any part, ask your AI tool now, while it is still open.

  6. Ask your AI tool to stop anything it started for the page, then leave it the way you did at the start of this lesson.

  7. Save your reduction’s export into ~/isotope-batches with the columns your page reads, as the same file type; filling in the blank sheet by hand works too, but every value copied by hand is a chance for a copying error. If your export does not carry the reference materials’ accepted values, they come from your lab’s record for each material, as its procedures say, never from memory. Paste the line from your note at the $ or % prompt and press Enter, open the address in your browser, and choose your file with the page’s button. The page may open on the practice batch first: check that it names your file, not batch-B1118.csv, before you read anything on it. The page reads it on your computer and saves nothing. Leave that terminal window open while you use the page; close it when you are done. To come back to the lesson folder later, type this lesson’s folder line again, the line near the top of this page that begins with mkdir.

Try another case

Before the quiz, write two or three sentences for yourself. The lab manager looks at RM-C at position 20, sees 187 µg of carbon against about 500 in every other capsule, and decides to leave that result out. What would you expect to change on your page, and what should be written down? And what would be different if the same capsule had been RM-A, a calibration reference?

A good answer sees that position 20 is a check result, which the reduction did not use, so leaving it out moves no value in the batch; on a file without it the page no longer shows it or marks the nine rows beside it, but now shows a stretch of nine sample rows without a reference result, from S07 to S14, including S05-D, between positions 13 and 25. It sees that the S09 and S09-D pair is still outside its limit and still needs a look: S09, at position 17, is one of the nine rows that were marked beside position 20, and S09-D, at position 28, is in the long stretch between positions 25 and 36; that the decision, its reason and who made it go in the batch record; and that the page should not drop the result by itself. Had it been a calibration result, leaving it out would mean the reduction is run again without it, which can move every value in the batch, and the page is then run on the new file.

Key takeaways

  • A short ask is enough. What you do, which file, what you want; your lab’s limits come from you, not from your AI tool’s suggestions.
  • Check a new page against findings you know before you trust it on a batch you do not.
  • A limit catches a result that crosses it, not a drift that stays inside it. Read the reference results in run order as well; the page shows where to look, and the lab manager decides what is reprocessed, run again, reported or held back, and what the client lab is told.
  • Practice data stays practice data: real batches go to the finished page, with your AI tool closed, after the data check.
KNOWLEDGE CHECK

In another batch, your page shows every reference result within its limit and marks nothing. In run order, the check reference reads −0.15, −0.05, +0.06 and +0.17 from its accepted δ13C (limit ±0.20), and the two calibration references rise the same way. What does that tell you?

What’s next

The next lesson is about checking one doubly labelled water study’s per-participant results against the laboratory’s written limits before they go back to the study team.

intellimetrics · Arlington, VA
© 2026 intellimetrics.net
Search lessons