I keep faking a plane crash in the first 30 seconds of every Google Meet
I've started joining every video call in my company 30 seconds early.
When you join early, everyone's virtual note takers join early too.
I spend the first few seconds yelling that both engines are out, we're going down over the Atlantic, I love my family, and someone needs to tell my story.
I then do the rest of the meeting normally.
When the meeting ends, everyone gets an emailed transcript where the AI summary reads:
"Attendee faces catastrophic engine failure, wishes his story to be told, and Q2 revenue updates"
_______
PS: If you're interested in more hacks like this follow me on X!
Haha...would be crazy if Q2 revenue is down, and the story tells "...and down it goes....Q2 crash-landed"
this is too good. granola not gonna know how to deal with this one
LOL love this
ohhhhmmmmggg this is absolutely hilarious. This sounds like a typical RD joke but retold for the AI era
hello
Achei o approach ousado pra prender atenção. Curioso sobre a taxa de conversão real desse tipo de hook — você mediu se as pessoas que ficam até o fim do vídeo convertem melhor do que quem entra "a frio"? Tenho testado hooks mais sóbrios nos meus projetos e queria comparar.
Please upvote me , i cant create posts yet
has anyone in a meeting ever actually noticed and asked about it afterward, or does everyone just skim past the transcript?
wow thats great. That can actually remove the hussle getting a dozen of customers orders you cannot handle as a starter
So, AI can never surpass humans — it'll always just follow the rules and do its job, haha.
This is chaotic in a very calculated way
You basically found a gap between how humans interpret context and how AI logs it. The model has no sense of “this is a joke before the meeting starts,” it just treats it as signal and folds it into the summary
Also highlights a bigger issue: a lot of teams are starting to trust these summaries as if they’re clean ground truth when they’re really just compressed transcripts with zero judgment
Would be funny until someone forwards one of those summaries to an exec or a client and suddenly you’re explaining why there was a “catastrophic failure over the Atlantic” in a revenue call
Low key though, this is a perfect demo of why prompt boundaries, filtering, and timestamps actually matter in production systems
So, AI can never surpass humans — it'll always just follow the rules and do its job, haha.
The real lesson buried in the joke: AI transcripts are now the system of record, and nobody reviews them. I have seen action items sourced straight from a meeting summary no human read twice. Whoever controls the first 30 seconds controls the minutes.
That really stood out to me too. As more teams rely on AI-generated meeting notes, the transcript often becomes the "truth" people work from. A small misunderstanding early in the conversation can ripple through action items and decisions if no one reviews the summary. AI saves time, but it still benefits from a quick human sanity check.
That's great 👍
thats funny and evil too :D
lol , this is pure gold! You’re basically testing the boundaries of how AI agents summarize human interactions. It makes you realize how fragile 'AI-generated insights' really are. I’ve been obsessed with the inverse of this lately: how can SaaS tools make themselves more readable and 'citable' by these same agents so they don't get filtered out as noise? Brilliant experiment, Matt."
Yes thanks for the input I'm trying to get a product published here.
This is inspirational. Thank you for improving my day.
Love the accessory your adding to AI transcripts. It's the fun we have a long the way that makes the journey.