Why 30 seconds before, 30 seconds after

MeetPing snapshots the past 30 seconds and the next 30 seconds of confirmed transcript whenever a keyword fires. People ask why those numbers — why not 60s for more context, or 10s for less noise. This post is the reasoning. It's mostly opinion and observation; we are not psychologists.

Working memory caps out around half a minute

The conversational thread you can keep "in your head" — the one you can rejoin without rewinding — runs about 20 to 40 seconds in attentive listeners. When you're half-tuned-out, which is the case MeetPing is built for, it's much shorter: you have a vague sense of the topic but no specific sentences. The popover is filling in the gap between "I know we were talking about churn" and "I can quote what was just said about churn." Anything beyond that gap is reading, not recall, and reading takes time you don't have.

↳ pull quote

Below the working-memory window you're rewinding the meeting. Above it you're reading rather than catching up.

Twenty seconds is too few sentences

A typical conversational pace is 130-180 words per minute, which means 20 seconds = 45-60 words = roughly 3-5 sentences. That is sometimes a single thought — an introduction with no payoff. We tested 20s in early builds. Several times the past column was just "okay so we should think about" and the actual subject was lost behind the cutoff.

Sixty seconds is too many to scan

Sixty seconds is around 150 words. 150 words is a paragraph. A paragraph takes 12-15 seconds to read for comprehension. We do not have 12 seconds in a meeting where someone just said our name. Everything past about 35-40 seconds becomes the wrong shape for the use case — too much to scan, easy to miss the relevant bit, demoralising rather than helpful.

The future window has the opposite shape: it streams in as tokens are confirmed, so you read it as it appears. 30 seconds of streaming text reads more naturally than 30 seconds of static — your eye tracks the tokens being added. We tried longer here too. At 45s you started feeling like you should keep watching the popover and stop watching the meeting. Bad.

Why symmetric (30 + 30) and not skewed

A reasonable counter-proposal: weight the past column heavier (say, 45s past, 15s future) since the future is also visible in the meeting itself, just as audio. We tested it. The result was that people read the past column for too long and missed what was being said live. The symmetric 30+30 forced the eye back to the meeting after a few sentences, which is correct: the popover is supposed to be a glance, not a habitat.

The overemployed user is the canary

The "overemployed" Reddit subculture — people running two or more concurrent jobs — has extreme requirements for a tool like this. Often not on camera, the meeting in a corner of one monitor, context-switching in five seconds, not fifteen. For that audience, anything longer than 30s of past context is a non-starter — there's no time to read it.

We didn't design MeetPing for that audience specifically, but they're the canary for a window that's too generous. If 30s works for someone with 0.4 of their attention to spare, it works for a normal user with 0.8. The reverse is not true.

What we'd change later

v0.1 ships 30+30 fixed. We get one issue every couple of weeks asking for 60s past — usually from product reviewers who want more architectural context, not less. That's a real signal. v0.2 might add a per-profile window override (engineering profile = 30+30, product-review profile = 60+30). We won't add it before we've shipped v0.1 to a wide audience and confirmed the default doesn't work for a non-trivial slice of users.

The take

Most product designs that hide behind "user-configurable" are dodging the calibration question. Sometimes the right answer is a slider, but more often it's an opinion. The 30-second window is an opinion. We think it's the right one. If you've used MeetPing and it's wrong for you, please open an issue — that's the signal we care about.

See the past-30s replay feature page for what the popover actually looks like.