Alp SuleimanGrowth Partner

Growth claim check

AI made us X percent faster. In real terms?

Pick the claim you heard in the meeting and answer five questions. At the end you see whether the number is ready for a real reading, and a next step for every gap that still needs closing.

Runs in your browser · answers are written to the link, the claim sentence is not · about 2 minutes

Claim

Which sentence did you hear?

Optional. If you pick one, each question gets a note specific to that claim.

Which sentence did you hear?

Optional. This field is not written to the link; it is cleared when you close the page.

Answered 0 / 5

Five questions

  1. 01 Task and person: In which task, and for whom, is the gain?

    Well-defined work from scratch and context-heavy work inside an existing system don't give the same result. Neither do someone new to the area and someone who knows it by heart.

  2. 02 Measurement: What was measured: perception, an intermediate metric, or an outcome?

    "We believe", "we feel", "most of us think we're more productive" are measurements of perception. They are valuable, but they don't replace time and outcome.

  3. 03 Stability and quality: Is there a quality metric next to the speed or the lift?

    In growth this can be refunds, support tickets, cancellations or the reproducibility of an experiment result. If speed rises while stability falls, the system is being amplified, not fixed.

  4. 04 Counterfactual: What would have happened without AI?

    A randomly assigned control group, an older cohort, a similar team over the same period. The more the tool spreads, the harder the counterfactual is to build; build it early.

  5. 05 Nominal or real?: Were the period's other effects taken out of the number?

    Seasonality, a price change, a campaign, the market's overall growth and new users who will try the product and leave (AI tourists) can all inflate the number. Take them out, the way you would adjust for inflation.

Answered 0 / 5

Answer all five questions to see the reading.

Method

The way you adjust for inflation

Felt speed is the nominal number. To reach the real one you have to take out expectation, task selection and the rework that comes after. This tool ties that adjustment to five questions.

The questions come from the essay The Real Growth of AI, gathered here into five steps: task and person in one step; the counterfactual split in two, the comparison and the adjustment for the period's other effects.

The reading is one of three states, and it is not a score. If the measurement rests on perception, the reading stays at perception whatever the other answers are; a single gap keeps the number nominal.

For the reasoning behind the five questions and the research they rest on: The Real Growth of AI →

FAQ

Does the tool give a number or a score?

No. It doesn't estimate, correct or benchmark the number. It only shows which questions the claim answered on its way in; the corrected number comes from your own measurement.

Why does a perception measurement decide the reading on its own?

The other four questions adjust a measurement. If the measurement is a survey answer, there is no time or outcome to adjust. Perception is valuable, but it doesn't replace an outcome metric.

What if we can't build a control group?

Compare with older cohorts at the same age, or roll the tool out in stages; the team or segment whose turn hasn't come yet becomes a temporary comparison. If neither is possible, write that next to the number.

Where are my answers kept?

There is no sign-up; the reading runs in your browser and no separate request is made to send your answers. So the reading can be shared, the answers are written to the link as a short code; if you share the link, whoever opens it sees them too. The claim sentence is not written to the link.

Source

  1. [1]METR, "Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity", 10 July 2025 (perceived and measured speed; early-2025 tools, 16 developers. In February 2026 METR published new data and wrote that this result no longer reflects the current impact).