Stef

Move fast and make things. 🦋

[Stef @stef]

@mattpocockuk One thing I added just this week is Vercel edge config as a way to control the nextjs flags I want rather than having to redeploy to turn flags on and off (rather than relying on Env).

[Stef @stef]

Yes this is kind of my workflow. Experiments per agent task with linear tracking the work as we go. Ensure the work doesn't disrupt the current process/feature. Front end preferences toggles for each flag for the users that are testing it so they can see current reality versus new reality. Once good feedback promote the flag as default with a Doppler config change. And repeat!

[Stef @stef]

@threepointone Yes this is basically what I'm trying to build

[Stef @stef]

@dok2001 @nextjs @Vite I think Vercel messed up with their pricing for a Nextjs deployment in this new world. $20+ min for non-hobby, per collaborator seat pricing (feels like pure margin for them), just doesn't make sense. Good to see you offering another way

[Stef @stef]

@sassy_009 Thanks for posting! I'm not on here much either but this popped up and I love your music - listening now 🎉

[Stef @stef]

@ericzakariasson @Rasmic While I am chatting to an agent, currently I can't select some text, press delete and rewrite it. I have to hold delete to get the caret back to the start of the text and start typing. Could you add a text editing pass to the bug bash?

[Stef @stef]

@fractaledmind Sounds great - perhaps you could post the caniuse scores for the various browser features this relies on?

[Stef @stef]

@flybayer @hiitsricardo Sorry what... is that the deposit or the total?

[Stef @stef]

I may be missing some context, but I see a company that has an existing trademark for several years and a new entrant that has chosen to use the same mark for an overlapping service in the same sector. The existing company MUST defend.

That's how trademarks work, otherwise the original company loses/weakens their position.

I don't have any skin in the game here but surely we would all benefit from this conversation if it's clear that this is how the world works?

[Stef @stef]

Tou don't nerd to speak check sf. Type correctly to be able to work with an LLM because they can deal wight the typos and still urndrfjsbg ehst IDB gong on. This is an amazing feature

[Stef @stef]

Oh wow. What happened over at Authy? Did they get rid of their product design team and go with vibe coding instead? 🤯

[Stef @stef]

@d4m1n I want to be able to get Cursor to run somewhere cloud-based that's not my laptop and make a feature I just came up with, and then make a PR that I can review. Ist that what this is? Are you saying you *don't* want that?

[Stef @stef]

Microservices actually seem quite well-suited to the new vibecoding era. Small scope, easy to reason about in a context window.

[Stef @stef]

@Azikara @PrajwalTomar_ It's not in production. It's just a command that runs in local development to generate code

[Stef @stef]

@eriktorenberg Really pleased for you!

[Stef @stef]

Maybe I'm using the wrong phrase then. Perhaps there are degrees of vibe code? I'm finding I'm able to make very promising things happen in a creative / product director mode (having done both those jobs) but the minute you need to flip to CTO / senior engineer, it's like the facade shifts.

[Stef @stef]

The main problem with vibe coding is that it turns your familiar codebase into the dreaded "somebody else's code". Here lies an opportunity, I think.

[Stef @stef]

"Any sufficiently advanced technology is indistinguishable from magic." (Arthur C. Clarke's third law)

Which must explain why every user interface now has a little ✨ button which does something inexplicable.

[Stef @stef]

@akothari @bleikamp Navigation. It seems wild to me that you have this semi structured data plus an AI feature but have no navigation layer on top other than a slow search and a sidebar that doesn't have what I want to see in it or can use to find what I want to see

[Stef @stef]

@jlongster @charliermarsh This pattern of rendering content as responses stream in from the LLM seems to not be that compatible with the React pattern of re-rendering the entire element for every token. Seems like a gap for something new