Thirty accounts times seven is 210 a month, about 2,500 a year. Model the migration before you do it: assume you lose 10 to 15% of the migrated cohort, so call it four accounts, and you keep roughly 26 x 19 = 494 instead of 30 x 12 = 360. That is a gain of about 134 a month, and you have spent goodwill you cannot re-earn.
The version that has worked for me is a middle path: sixty days notice, and the offer to lock the old price for one more year by moving to annual. Most of the cohort took the annual lock, which gave me the cash up front and moved the awkward conversation twelve months out.
Tell your supervisor. I know. I said 'it's going fine' for eight months, and the conversation I was terrified of took eleven minutes and ended with a restructured plan and a monthly deadline. Supervisors have seen this many times. The version they haven't seen is the student who says nothing until the funding runs out, and that's the one that ends badly.
84% coverage on an inherited service with zero tests, in one go, was always going to be false. Real coverage on legacy code needs refactoring for testability first, and the model routed around that by mocking - which is exactly what a person under time pressure does too.
Can confirm on asking. I assumed the answer would be no and sat on it for a year. When I finally asked, I had a written 'sure, just do not sell it to our customers' back in four hours. The year of quietly worrying was the expensive part.
Regional variation is bigger than people expect. The sign an app teaches you may not be the one used locally, and unlearning it later is annoying. Local instruction gets you the local variant the first time, which is a real practical argument for the class over the app.
Unmasking at home was the single biggest lever for me, and it is not advice you can act on in one afternoon. Stimming freely, lights off, no talking after 7pm on bad days. I'd spent a decade doing 'recovery' that still involved performing for an audience of nobody.
Don't apologise for the state of the product and don't narrate the roadmap. New founders spend a third of the call pre-defending things nobody has complained about. If a screen is ugly they'll say so, and then you say "yeah, that's next month" and move on.
Depends heavily on the community. Some remove it on sight regardless of how useful the answer was, and getting the account restricted costs you the entire channel.
Also ask what number they need for the approval. Sometimes there's a threshold, like anything under 15k doesn't need sign-off, and knowing that shapes how you structure the whole thing.
Two years on it. The content style is genuinely different: full sentences and cultural notes rather than matching exercises. For Portuguese specifically, check whether your library's version carries both European and Brazilian, because they are separate courses and picking the wrong one is a mildly painful discovery at month three.