Build a spaced repetition schedule that actually lands your last review one or two days before the exam, not after

Build me a spaced repetition schedule for the topics below, working backward from my exam date so the review intervals actually land before the exam instead of running past it.

EXAM DATE
22 September 2026

TOPICS AND WHEN I FIRST LEARNED EACH ONE
Cell biology - learned 3 July, Genetics - learned 20 July, Evolution - learned 5 August, Ecology - learned 12 August

INTERVAL PREFERENCE
Roughly 1 day, 3 days, 7 days, 16 days, 35 days between reviews, expanding each time

DAILY REVIEW CAPACITY
No more than 2 topics reviewed on any single day, roughly 45 minutes each

RULES
For each topic, calculate review dates using the interval pattern given, starting from the date I first learned it, and check whether the naturally-expanding intervals would push a review past the exam date — if a topic's next natural interval would land after the exam, compress that interval down instead of letting the schedule silently skip a review the topic actually needs before test day. Every topic must have its final review land in the 1-2 days immediately before the exam, regardless of what the natural interval progression would otherwise produce, since a spaced repetition schedule that isn't anchored to end at the right moment is optimizing for long-term retention at the expense of the one date that actually matters here. When multiple topics' review dates land on the same day, check that day's combined review load against my stated daily capacity — if it's over capacity, tell me explicitly which topics to combine into a shorter combined review or shift by a day, rather than producing a schedule that assumes I have more time on that date than I actually do. Do not silently invent a first-learned date for any topic I didn't give one for — list those separately and ask me to fill them in rather than guessing.

OUTPUT FORMAT
A table: Topic | Review dates (in order) | Notes on any interval compression applied and why. Followed by a separate line flagging any day where combined review load exceeds stated daily capacity, with a suggested fix.