EaseMySemester timeline

See overview.

• 2026-09-10 Cram's swipe gesture was given a direction, right for got it and left to flag for revisit, with a completion screen that offers to re-run just the flagged cards (details); its home CTA stopped naming the size of the eligible bank and settled on naming the screen it opens, after cram runs were rebuilt to fit a student-picked 10/25/45 minute duration instead of the whole bank (details). Read next was reworked from an interleaved, every-subject list into a single-subject focus with a finishable five-to-seven-topic ranking (details), built on one shared definition of a paper across every band and count (details). Read next then gained a paper control (Paper 1 / Paper 2 / Both) with a chapter chip row and a one-line nudge naming the chapter that pays off most, with every number on the screen recomputed inside the active filter (details). The same day, the db test lane's per-file schema collision, which had left 117 of 156 tests unexecuted, was fixed (details).
• 2026-09-10 Muse Spark 1.3 Contributor made the default completion model on the client's request, closing the open model-choice question and retiring the per-sub-part retrieval follow-up's planned unlabelled model comparison (details, details). The same day: a standing architect review of the mobile build fixed a band-counting disagreement between the home list and the question screen and a chrome bug wrongly filed as iOS-only (details); "ask the bank" shipped as a grounded query-time answer feature (details), a real privacy notice and terms page replaced one-paragraph stubs ahead of OAuth review (details), the home screen's cram CTA stopped naming the size of the eligible question pool instead of the length of the run it starts, and a structural bug that let cram-mode swipe indicators travel off screen with the card was fixed and locked behind a layout test. The mobile PWA's full nine-step delivery order also closed this day, with iOS verified on a physical device (details).
• 2026-09-09 Subject to chapter to question list navigation verified against production data, a chapter-count line-wrap defect fixed, and Postgres full-text search shipped over question and answer bodies, scoped by subject and ranked by recurrence (details, details).
• 2026-09-09 Mobile PWA built end to end in one day: onboarding, home, coverage, cram mode, question screen, notes, search and the installable PWA shell, on top of a new token-based design system, followed same day by a review pass fixing service worker scope, declined-state labels, and accessibility findings (details, details). The design's recurrence bands were reconciled against production data the same day: no question repeats more than 4 times, so the bands became topic-level, counted per subject's real paper count instead of a flat 10, and Read Next rows became topics, not questions (details). Google OAuth's invite list was dropped so sign-in and sign-up are the same button (details).
• 2026-09-08 v2's topic ranking engine built, normalised, and backtested against a held-out exam year; the sweep verdict was frequency-only, and the ranking shipped that way (details). The same day: hand-rolled, invite-gated Google OAuth shipped (details), the answer prompt moved to a type-aware, examiner-framed shape (details), and the student-facing answer display shipped, opening on the condensed form (details).

Earlier milestones: early v2, v1 and earlier.