Calculate the numbers.
Rates come from the selected records. Changing the station changes the answer and its evidence together.
Independent demonstration · Built for this portfolio
A useful answer
shows its work.
An AI interface should help someone understand what changed, inspect the evidence, and recognize when more judgment is needed.
This public walkthrough explores that experience using fictional manufacturing records. It is separate from my work supporting Apple as a contractor through Flexon Technologies.
Follow the evidence.
Synthetic data · No live model04 / Calculated from the selected records
Yield fell by 5.0 percentage points.
The previous shift passed 96.0% of tested units; the current shift passed 91.0%. Both use the same sample size. All 50 additional failed units are recorded at Station B.
Previous shift96.0%
Current shift91.0%
Change-5.0 pp
Inspect the source records 4 records + 1 note ↓
| Source | Station | Shift | Tested | Passed |
|---|---|---|---|---|
| QA-01 | A | Previous shift | 500 | 490 |
| QA-02 | B | Previous shift | 500 | 470 |
| QA-03 | A | Current shift | 500 | 490 |
| QA-04 | B | Current shift | 500 | 420 |
LOG-01 / Station B, current shift, 08:30
Camera calibration updated. Follow-up inspection pending.
This is a deterministic interaction demo. Calculations use the records shown above; explanations are authored examples. It does not connect to an LLM or a client system.
Three decisions behind this demonstration
Keep the sources close.
Someone should be able to inspect the underlying records without losing the question they came to answer.
Make uncertainty useful.
When evidence runs out, explain what is missing and who should make the next decision.