All micro-internships
Micro-internship brief ≈10h across 3 days — forecast on Day 0, then a daily actuals check for 3 dayslow-code

Make a real forecast — then face the actuals

Anyone can analyse the past; forecasting makes you accountable to the future. Build a simple forecast with AI’s help — tomorrow’s temperature, AQI, petrol rate — write your predictions down BEFORE, then verify against reality for 3 days and score your error.

Data AnalysisResearch & Fact-checkingSpreadsheet AnalysisAnalytical ReasoningFinancial Modelling

Posted by The AI Internships

The work brief

  1. 01Pick something with public daily values and get 2+ weeks of history (weather sites, AQI portals, fuel price trackers).
  2. 02Ask AI to suggest simple forecast methods — moving average, same-day-last-week, trend line — pick one, build it, and make sure you can explain it with the chat closed.
  3. 03Write down predictions for the next 3 days WITH dates, before the days happen.
  4. 04Each day, record the actual value. Compute your error (predicted vs. actual) and compare against a naive baseline ("same as yesterday").

What you’ll produce

4 deliverables

Submission standard

Submit your method, the dated predictions, the actuals as CSV, and the error analysis vs. the naive baseline. Honest misses beat suspicious accuracy. Include your AI workflow — tools, best prompts, and what you changed. Publish your predictions somewhere timestamped (a version-history Google Sheet, dated post, or commit) before the forecast days, and link it.

  • What you forecast, the history you used, and your method explained in your own words

    Written responseRequired
  • Your 3 dated predictions, exactly as written before the days happened

    Your 3 dated predictions exactly as written before the days happened — these must match the timestamped record you link.

    Written responseRequired
  • CSV: date, predicted, actual, error — plus the naive baseline’s error

    CSV: date, predicted, actual, error, and the naive baseline's error — one row per forecast day; honest misses beat suspicious accuracy.

    CSV dataRequired
  • The analysis: your error vs. naive baseline, where you missed and your theory why

    Written responseRequired

You’ll complete these inside your private workspace.

What you must submit as proof

This brief requires evidence an AI can’t fabricate.

  • Link to your timestamped, pre-committed predictions (a version-history Google Sheet, a public post, or a dated commit) made BEFORE the days happened

    The timestamp must pre-date the forecast days — this is what proves you did not back-write the predictions to match reality.

    Public linkRequired

Submissions without this evidence cannot be submitted.

Protect other people in your proof. Blur faces, names, phone numbers and email addresses before you upload, and refer to anyone you worked with by role or number ("Listener 1", "the stall owner"). Your proof is only ever used to check your work — it is never published, never appears on your certificate, and is never shown in your public portfolio.

How your work is evaluated

The passing benchmark is 70/100.

Genuine pre-registration

22%

Predictions were clearly made before the actuals (dates, method consistency).

Verification rigor

22%

Actuals are real, errors computed correctly, baseline included.

Learning from misses

11%

The analysis treats misses as information, not embarrassment.

Timestamped pre-commitment

44%

A verifiable timestamp (version history, post date, or commit) must show the predictions existed before the actuals. Suspiciously perfect accuracy with no pre-commit timestamp is a fail; honest misses with a real timestamp pass.

How we grade your AI usage

30% of your score

Using AI is the point — it’s the skill this certificate proves. You’ll answer three short questions about how you used it: what you asked, what was wrong with its first answer, and what you changed. Specific, honest answers score high. “I pasted the brief and submitted the answer” scores near zero.