Everything you need to pass,
in one platform.

In Profile → Notifications & reminders, failing to turn on browser push always showed the same line: "Failed to enable notifications." Different causes need different responses from you, and none of that was visible. We are sorry.

What is different now. Each cause now produces wording that says what to do:

①Your browser cannot register for push — some browsers (Brave, for example) disable the push mechanism by default. We cannot fix this from our side, so we now suggest trying another browser such as Chrome or Edge.

②The network did not reach us — we ask you to check your connection and retry.

③A problem on our side — there is nothing for you to do, so we ask you to try again later, and note that your browser is already registered.

④Notifications turned on but the settings failed to load — this previously also read "failed to enable", so it looked like a failure when it had actually succeeded. It now shows On and asks you to reload the page.

We also append a short diagnostic code to the message (for example [subscribe/AbortError]). You do not need to interpret it — just quote it if you contact us and we can identify the cause immediately.

Preventing a repeat. We added an automated check that actually produces each class of failure and confirms each one renders differently. Adding wording without verifying it would let everything collapse back into a single message.

Study history, review schedules and earned XP are unaffected.

The 👍/👎 and Report controls on Ask AI answer cards showed "Thanks for your feedback" and "Reported" when pressed — but none of it ever reached us. We are sorry: you took the time to tell us something and it went nowhere. This was the case from the feature's launch on 2026-07-13 until today.

Cause. The buttons and their confirmations shipped before the receiving end existed. Pressing them only changed something inside your own browser; nothing was sent. Because the screen responded, it looked like it had worked.

From today. A 👍/👎 is stored against that specific answer and feeds quality improvements. A report goes into the same queue as reports on questions and comments, and we review it. Open Ask AI history (top right of the panel) and you will see the rated and reported state still there.

If it cannot be sent, we say so. A failed submission now shows "couldn't send" in place, instead of pretending it succeeded. Answers that were never saved to your history — for example one you stopped mid-way — no longer show the buttons at all, with the reason stated.

Ratings and reports are visible only to us, never to other learners. Study history, review schedules and earned XP are unaffected.

In the comment area on each question, editing or deleting your own comment, liking, and reporting did nothing when clicked. We apologise for the inconvenience. This lasted from 2026-07-28 until today.

Cause. When we tidied up the internal fields returned by the comment list, we also removed the identifier that points at an individual comment, so edit, delete, like and report had no address to send to. The screen still rendered and the buttons still appeared, so from your side it simply looked unresponsive. Posting and replying were unaffected.

Also fixed: internal IDs in the comment list. The same comment list was returning the internal ID that identifies the author (a meaningless string assigned automatically when an account is created) and the list of internal IDs of people who had liked a comment, retrievable without signing in. No account information — email address, name, or password — was ever included, and these internal IDs cannot be used to identify a person or act on an account. Even so, there is no reason to expose them, so we no longer return them. The list now carries only the like count, and whether a comment is yours or whether you have already liked it is answered by a separate mechanism that responds only to you while signed in. What you see and do on screen is unchanged.

Preventing a repeat. We added an automated check that actually runs both sides: that the list returns no internal IDs, and that edit, delete and like still appear while signed in. Checking only the first would let a change that strips the IDs and breaks the feature pass as "fine".

Study history, review schedules and earned XP are unaffected.