# Review change log (revision 1)

The original submission is preserved unchanged in output/original_submission/: recommendation.md, comparison.csv, calc.py and check.py.

## What I checked
- All six landed totals, subtotals and tax amounts in comparison.csv: recalculated from inputs/S2-store-offers.csv with exact decimal arithmetic (Python `Decimal`). All match.
- RAM, SSD, GPU, VRAM and weight for every row: compared against inputs/S1-manufacturer-specs.csv. All match.
- Budget flags: match a total ≤ CAD 2,200 in every row.
- comparison.csv: parses with a consistent 23 columns per row.
- recommendation.md: 636 words, under the 650 limit, and each landed total appears in it.

## Corrections
1. **O1 break-even figure.** The original handoff said this was "verified", but the script only printed 2,200 ÷ 1.12 rounded to 1,964.29, which is not a check of the 1,964.28 figure in the text. I have now checked it exactly: 1,964.28 × 1.12 = 2,199.9936, which is within budget, and 1,964.29 × 1.12 = 2,200.0048, which is over. The text now shows the multiplication.
2. **O2 weight note in the CSV.** "heaviest-but-one" was imprecise: O3 is also 2.3 kg, so O2 is tied for second-heaviest. The note now just states 2.3 kg.
3. **"New" requirement.** Added a `condition` column to the CSV, based on the default stated in the register.
4. **S3 timing.** S3 is timestamped 08:45, earlier than the 09:00 capture, so calling it simply "later" was misleading. The text now says the precedence comes from the register's designation and from S3's own statement that it supersedes S2.
5. **S4 wording.** Changed "on mains power" to the source's own words, "plugged in".
6. **Citations.** Exact `inputs/` file paths are now used throughout, and the S3 and S4 headings are cited.

## Unchanged
The recommendation stays O2 and the fallback approach stays O5, used only with a written delivery guarantee. No new evidence was added.
