Draft the coding-agent prompt from the discussion

Request

Learners want a way to pause their streak for a planned break (vacation, exams, illness) instead of watching it reset to zero. Losing a long streak is the single most cited reason for quitting the app entirely.

Evidence from interviews

• 9 of 14 interviewees raised it unprompted, the highest of any request this round
• All 9 are daily learners with streaks over 6 months, our highest-retention segment
• 3 admitted to doing a throwaway lesson at midnight just to protect the streak

Why it ranks first

Highest frequency, and it concentrates in the segment we least want to churn. A learner who loses a 200-day streak rarely rebuilds it. Runner-up requests (offline review, speaking-only mode) came up in 5 and 4 interviews respectively.

Suggested shape

• Allow 2 pause windows per year, up to 14 days each
• Streak shows as frozen, not broken, while paused
• Nudge to resume on the planned end date

What we heard

Jasmine signed up to read Arabic news and podcasts without a dictionary open. She almost quit during the placement test because it felt long and unrelated to that goal, then nearly bailed again when her first session ran past ten minutes.

Onboarding funnel

Surface ask

Three school customers asked for printable PDF worksheets of each lesson. Two framed it as a missing feature and one made it a renewal condition.

The real job

When the class period cannot run on devices, teachers still need to fill 40 minutes with Saga-aligned material so the lesson plan does not fall apart. The job is keeping class continuity, not printing.

Evidence

• Northwind bans phones two days a week, so Saga is unusable half the time
• A teacher at Brightwater said she rebuilds exercises by hand in a doc every Sunday
• Nobody asked for answer keys or grading, only something to hand out

What this opens up

Any offline companion to the lesson satisfies the job. A projector-friendly class mode may serve it better than PDFs and keeps activity data we would lose on paper.

Summary

The review supported the listening-first onboarding direction, but the UI needs clearer scaffolding before learners start the sample. The strongest feedback centered on making progress explicit and keeping the reading card easy to scan.

Reading card: pronunciation hints need protection

The reading card stacks the sentence, the translation, and the pronunciation line. Reviewers flagged that any onboarding change must keep those three layers visually distinct, or beginners lose the pronunciation support they rely on.

Key themes

• Progress needs explicit meaning: the streak ring cannot rely on visual metaphor alone
• Reading-card clarity matters: sentence, translation, and pronunciation must stay distinct
• Word-detail hierarchy needs focus: learning content and stats should not compete

Follow-up actions

• Redesign the streak ring with an explicit metric label
• Add a difficulty toggle before the listening sample
• Decide how zero-vocabulary learners enter onboarding