New Fitbit Air looks like every advertiser's dream. Connect advertising with heart rate and sleep quality to know the exact moment you need to be shown something, and you won't have any chance to resist that impulse
@jackfriks A team of 4-5 friends is perfect even for a smaller mrr. I just found it's too boring to build something alone. The ability to speak with friends and do what you all like is priceless.
You can't just stop sending invoices to customers without any notice to save on Stripe fees ๐ @perplexity_ai
Now I have to email them to get mine. Feels like a good moment to cancel tbh.
After five months of Bluey on loop across nine countries, I'm back in London and would quite like to meet someone whose opening line isn't 'how old is yours?'
Wow, @perplexity_ai Personal Computer is incredible. I honestly like how it works, even with a slightly buggy interface.
But I need it with local models. I asked it to clean up and structure my reading list, and it cost me 1400 credits or $14.
Anything else I can try?
I like @cursor_ai cloud agents, but it makes no sense while we have $200/m plans in Claude code and Codex.
I just spent $1600, just because I forgot to check usage ๐ซ
Day to day using @claudeai while travelling and in the end they just banned me without explanation. My best guess is it's because I was in China at the time. The dates are really close.
The differences between Shanghai and London:
* People smoke cigarettes instead of weed.
* It is clean here.
* People follow the traffic lights.
* People on electric scooters want to kill me, not people on Boris bikes.
For full agentic AI coding, the most important thing is having e2e tests. Only in that way is it possible to work on full features in parallel without falling into micromanaging each agent.
Plan -> Code -> PR -> e2e pass -> Go to prod
Each agent owns a full cycle independently. If e2e passes, it ships.
@cursor_ai +1 hour and + $35 :D
It's finally in dead loop,
But the funny thing is, CC completed that task while I was checking how that agent felt ๐
The @cursor_ai long-running agent has been going for 5 hours now, and it's literally just updating the scratchpad for at least the last hour. ๐
Definitely setting a time limit next time...