@dannypostma The ask before anything risky part is the real feature. An agent that emails customers needs a hard approval gate in code, not a prompt telling it to be careful.
@dannypostma Letting an agent stop and wait for a human answer instead of guessing is the real unlock. Email fits because the async part is already solved.
@gregisenberg Pressure to go viral is the one that hurts builders most. Writing for the 50 people who would actually use your product beats chasing reach every time.
@gregisenberg Same playbook as Google My Business, except now the model decides who gets recommended. Small businesses that fill their profile with clear, specific details will win this early.
@yongfook A progress bar is the bare minimum when one clip is a few GB. Switching Photos to Download Originals before you need the file saves a lot of guessing.
AI models keep getting smarter. Most of them still know almost nothing about your day.
The next big jump won't come from a bigger model. It'll come from better context: what you agreed, who you talked to, what you're actually working on.
#AI
@gregisenberg this is exactly it. solo in nz on a tiny watch app and half my time is just getting messy real-world context into the agents. whats your main source for that right now?
@levelsio same. been glued to the desk shipping a tiny watch app and the second i step outside my brain actually works again. you getting out more or still trapped