Martijn Smit
@smitmartijn
Building productivity tools. Flying dutchman. Stroopwaffles 📊 http://WhatPulse.org 📞 http://MuteDeck.com Improving & mapping productivity is my sweet spot.
sometimes I need to switch focus to something entirely to clear my head usually, that turns into building something fun and quick... and today that was Typing Sniper - a typing game where you need to 'snipe' words quickly 😊 typetest.io/games/typing-s…
Antrophic adding rate limits to their paid plans..this caught my eye: "Most Pro users can expect 40-80 hours of Sonnet 4 within their weekly rate limits." meaning the reason they're adding limits is because people are letting it run MORE than 80 hours a week?!
nothing like going into your public knowledge base to update your docs for a new feature, only to discover that article was still mentioning a feature you removed 2 years ago 🫠
when architecting a solution: if it feels too complex, it probably is.
running multiple AI agents is like herding cats: sometimes they behave and things are organized but sometimes they're gremlins and start ffing things up, infecting each other with mischief
I absolutely love the subgenre of youtube creators that are able to work for free because they're getting incoming from their videos - so much heart like SB mowing, the Rudan brothers, showcasing the best side of the creator economy ❤️
adding an identifier to websites where you sign up, is still one of the best ways to see who is selling your data
starting your Monday with a customer message like below hits different 🫠 "I don't want to pay for your product, but here's a list of features I want"
another thing to add to your system prompts: "only 100% test success rate is acceptable." otherwise you get 💩 like this:
my kingdom for select drop downs that are actually sorted alphabetically
if you do need to scroll make sure it's the Cats! feed you're scrolling (or Dogs! if you're the other kind of person)
once you get to a certain number of tests in your app, it pays to check if you're not seeding too much data per each test I had a general seeder that tests in the beginning all used data from, but as time went on - less so. I removed the general seeder and this was the result:
