Laszlo

Building a Kubernetes client with Cursor AI:
gimlet-io/capacitor

[Laszlo @laszlocph]

@rawkode @SaiyamPathak It’s a feature

[Laszlo @laszlocph]

@thorstenball That’s called “magyar kártya” 😀

[Laszlo @laszlocph]

@nocodelife Essentially you are working for an AI now?

[Mitchell Hashimoto @mitchellh]

The real future we need is better broad support for mapping each line in the diff back to a complete and public history including prompts. Thread sharing from Amp and Opencode help with this, but I need a `git blame` equivalent. The session exposes true expertise or slop.

[Mitchell Hashimoto @mitchellh]

@zeeg There's like a million things GitHub could be working on that would better position it in the AI world instead of whatever the fuck F-tier AI Copilot is. But the people who are driving whatever it happening clearly are too aloof or high on their own supply to figure this out.

[Laszlo @laszlocph]

@nocodelife It's English!

[Laszlo @laszlocph]

tldraw stops accepting PRs

[Laszlo @laszlocph]

If code has low value. Code from uncontrolled internet sources has negative value.

I'd rather have my own cursor say stupid things than a random internet entity.

[Ido Salomon @idosal1]

My entire childhood has led me to this moment...

I built AgentCraft - orchestrate your agents with your favorite RTS interface! ⚔️

Coming soon 👀

[Laszlo @laszlocph]

@thorstenball Ignore.
Many just want you to see their posts on your wall. Dishonest.

[Laszlo @laszlocph]

@badlogicgames Same on azure. Also for openai models

[Laszlo @laszlocph]

@mitsuhiko @steipete I mean yes.

But why can't we just pay the TV price for a TV?

I'd happily pay that extra 20% to not ruin the experience.

[Laszlo @laszlocph]

@mitsuhiko @steipete What's wrong with the old business model of TVs?

I buy a screen for it's fair price.
Then I plug it in.
End of story.

Why this does not work in 2026?

[Laszlo @laszlocph]

word of mouth === bragging

> impresses them so much
> that they wanna brag about it
> to their network
> their whole status spins around them discovering &
> consuming high quality cool products

[John Rush @johnrush]

If you read this til the end, the chance that you’ll give up on entrepreneurship is 81%

or you gonna learn the things that helped me keep going & get here

I’ll touch on
> product-vs-marketing debate
> if distribution is the king
> the truth I learned the hard way

Let’s go

If we ignore the enterprise sales, then it’s impossible to grow at a decent rate without word of mouth effect.

WOM happens if users love the product. Making smth users gonna love is the hardest job on earth, same as making a song that tops the charts.

People love if you solve the problem that they care enough about in a way that impresses them so much that they wanna brag about it to their network

their whole status spins around them discovering & consuming high quality cool products,

and if your product falls into this category then users share it to earn status.

Once the facts above are in place, the product ends up dominating its niche.

How to build such product in b2b:

> the general idea must be very relevant for the users and it must
>> make them more money
>> save them serious amount of time
>> remove serious amount of stress

> while you build the product, there are 1000 new little ideas you need to come up with for little flows and use cases
> just like in a great song, all of it must be great, not just a little part of it
> in most cases you need to reiterate on existing features
> basically remove your code and write a new one
> which is painful and you start adding new features instead of reducing their number and reiterating with a few of them)

The quality of your intuition is a dealbreaker here, because all those ideas have so many A/B/C decision to take

Your intuition can be hacked by you asking those questions directly from your users

Talking to users isn’t that easy, in most cases users give wrong answers

To get right answers you rather need to talk to them about their problems and then magically come up with solutions

Throw those solutions to your users and observe. If you see them complete the flow or come back to you saying “wow, great feature..” then you’re in the right path

All of this takes months or years, and while you iterate the market may just change or move forward and your whole product is obsolete

It means your innovation/iteration speed must be rapid, which is super difficult too

Ofc u need an initial push, to get some users to test your product on, and in almost all cases this is done just like Airbnb did(stupidly go around knowing the doors, and being okay with such incapable marketing effort)

So basically there is no “first I come up with an idea that I think is great, then I build the product and then I go market it”

All 3 happen simultaneously in one single process that’s impossible to even separate,

That’s why 99% of the startups are done by founders first, and they turn it into process-based system only after PMF,

Before PMF they are just swimming inside the ideation/building/marketing maze as it’s one inseparable flow

Long story short, building successful business is hard and you must learn to love the process, because the outcomes most likely gonna be meh for most of the time, if not ever.

And if your product isn’t doing well, most likely it’s everything to blame simultaneously: the idea, the product, the features, the UX, the niche, the timing, the marketing…

Improve everything. Don’t assume you’re good at any of this unless you’ve success in the past.

[Laszlo @laszlocph]

@stevekrouse 8 light minutes.

[Laszlo @laszlocph]

@steipete Cozy!

[Laszlo @laszlocph]

@ttunguz GPT wrappers FTW

[geoff @GeoffreyHuntley]

@laszlocph pr reviews are dead ;p

watch people waste 12 months
coming to this conclusion

[Laszlo @laszlocph]

@GeoffreyHuntley Agreed.

This is not the time to be conservative.

[Laszlo @laszlocph]

I guess we're gonna stop doing PR reviews in 2026.

[Daisy Hollman @The_Whole_Daisy]

So apparently @GeoffreyHuntley's Ralph Wiggum technique (and our Claude Code plugin for it) went a little viral over the holidays? I thought I'd share a typical prompt I use with it. All of the instructional videos I've seen don't give thorough enough completion promises, IMHO