@bcherny 1) claude generates the code
2) your or claude then open a PR
3) it looks for bugs
seems a bit redundant when surely it could’ve done 3) at 1)
I'm Boris and I created Claude Code. I wanted to quickly share a few tips for using Claude Code, sourced directly from the Claude Code team. The way the team uses Claude is different than how I use it. Remember: there is no one right way to use Claude Code -- everyones' setup is different. You should experiment to see what works for you!
Model can write all code that’s possible today. But -
Can it innovate?
Can it build the next k8s level tool?
What is the prompt to manual code ratio for building tools like A2UI?
@Google
junior pm: i have to run my first global alignment meeting and i'm freaking out
senior pm: what's your plan?
junior pm: schedule the call, walk everyone through the proposal, get buy-in
senior pm: how many people?
junior pm: like 8 stakeholders across 4 time zones
senior pm: and you're meeting them for the first time... in the alignment meeting?
junior pm: well yeah. that's the point of the meeting
senior pm: there's your problem
junior pm: what do you mean?
senior pm: alignment meetings don't create alignment
junior pm: then what's the point?
senior pm: they confirm alignment you already have
junior pm: i don't follow
senior pm: have you talked to any of these 8 people yet?
junior pm: i sent them the pre-read
senior pm: that's not talking
junior pm: so i should... call them first?
senior pm: every single one. 1:1. before the big meeting
junior pm: that's 8 separate calls just to prep for 1 meeting
senior pm: now you're getting it
junior pm: seems like a lot
senior pm: it's the whole job
junior pm: what would i even say in these calls?
senior pm: you're not saying. you're listening
junior pm: listening to what?
senior pm: their fears. their motivations. their hidden objections
junior pm: why wouldn't they just say that in the meeting?
senior pm: politics. pride. audience
junior pm: so they'll tell me privately what they won't say publicly?
senior pm: if you ask right
junior pm: how do i ask right?
senior pm: frame it as soliciting their expertise
junior pm: "i'm new to this and want your input"?
senior pm: exactly. people love feeling like the wise advisor
junior pm: and then i just... shut up?
senior pm: shut up and write everything down
junior pm: what am i looking for?
senior pm: conflicting goals between stakeholders. who wants what and why
junior pm: ok say i do all this. what changes?
senior pm: two things
junior pm: go on
senior pm: first, you adjust your proposal based on what you learned
junior pm: makes sense
senior pm: second, you sprinkle their exact phrases into your presentation
junior pm: why?
senior pm: people support what they helped create
junior pm: even if they didn't technically create it
senior pm: especially then
junior pm: what about the difficult ones? the ones who'll disagree no matter what?
senior pm: you pre-negotiate
junior pm: how?
senior pm: in your 1:1, you say "i know you have concerns about X. i can't fully address them. can you disagree and commit in the meeting?"
junior pm: you can just ask that?
senior pm: you'd be surprised how often the answer is yes
junior pm: because you asked privately
senior pm: because you respected them enough to ask
junior pm: so the alignment meeting itself is basically...
senior pm: a ceremony
junior pm: confirming decisions already made
senior pm: in hallways. in 1:1s. in coffee chats
junior pm: the meeting before the meeting
senior pm: now you're ready to run alignment
In the middle of an advanced voice chat with ChatGPT, I asked, "Why is Sam Altman so dismissive of @elonmusk?" - it keeps making weird sounds, while working perfectly well for other statements. Tried multiple times. Weird.
IBM couldn’t build an OS so they were forced to partner with Microsoft.
Apple can’t build an AI so they’ll be forced to partner with the next Microsoft.
The only way out is to collaboratively fund a competitive open source model.
Stumbled upon this top-notch video on YouTube - https://t.co/syPECTWutp
Makes me wonder if there's a way to improve my YouTube feed, showing more of this good stuff instead of the attention-grabbing, useless content that's become the norm over the last few years.
That awkward moment when you hit the limit on the smart model and have to use your brain.
The difference between the models as a collaborative worker is very apparent.