Nick 🚧

i tried running OpenClaw locally. macbook → insecure permissions hell + errors. docker → config + dependency hell. it boots → my macbook turns into a space heater. vps → ports, perms, secrets, security paranoia. so i built Clawster.io paste keys → deploy. live sandboxed OpenClaw agent in 28 seconds.

@shadcn glassmorphism drip hits so hard. thanks a lot!

You can now build translucent dropdown menus, selects, and comboboxes on shadcn/create. Mix in your custom styles, icons, fonts, and radii and build something different. Your own preset. All it takes is a few clicks.

@leodev what's so bad about supabase? i've seen large startups using it in production with no issues. it's not perfect, but it gets you up and running in no time and scales easy.

Why do vibe coders always choose supabase as their backend? There are so many better options that will give higher quality results.

@jackbutcher seeing is cheap. seeing valuable things is what matters more

You are not what you can make. You are what you can see.

@SimonHoiberg opus 4.5? that was eons ago

Ok, I take it back. Now after using GPT-5.4, Opus 4.5 feels retarded 😂 I haven't written a line of code in months. I started letting AI review it's own PRs, and I now trust its judgement better than my own. I have a team of OpenClaw agents run major parts of my business, and a personal AI assistant that handles everything for me. I'm outside most of the time now, talking to my team on Telegram through voice. I don't even need a laptop anymore for large parts of my work. I can't believe how fast this is accelerating.

@lukatofocus talking to every trial user directly, that's converting around 15-20% so far. it's a small sample but the ones who try it tend to stick. what's your benchmark for a tool like this?

@nickvas_dev cutting costs 50% before you scale is the right order. most founders optimize for cost after they're bleeding money. 34 trial users, what's the conversion to paid looking like so far?

Today's 6h deep work session: → Cut model generation and cut costs by 50% → Made the video pipeline 2x faster → +34 users on trial Building an AI UGC video tool as a solo founder. No investors. No team. Just shipping. What did you ship today?

The 50% cost cut took me 20 minutes. Sometimes you overthink it.

40/730

39/730

38/730

37/730

36/730

@jeffreyhuber i think the real tax is that you can rebuild your entire data model when you hire someone who actually knows databases. serverless postgres kinda kills the "operational burden" excuse

more of you should read this dx.tips/oops-database

35/730

34/730

33/730

32/730

31/730

@JakeManger great work jake! how many iterations did it take until it started to be profitable? did you speak about the distribution and marketing challenges you hit?

1 year ago I launched a file converter app Started as a broke PhD student just trying to make rent Today it hit 2,280 users and supports me financially Supports 3,985 file conversions (documents, videos, images, audio, emails) All local. No uploads. No subscriptions. Just drag, drop, convert Best part? I use it almost daily for my own photography and work (I even used it to crop and trim the videos in this post) Building something you actually need hits different howtoconvert.co

yeah claude code seems to get all the heat now, until the next shiny thing pops out lovable seems better structured around the process for ui centric work, it's a pity that the code it renders is not really production ready have you tried forge.new ? had some really amazing results so far with it

@nickvas_dev Just watching everyone use Claude rn . and honestly for the simply small stuff I was doing like wallet trackers and vol trackers and stuff. lovable has been a lot better with on chain stuff . Would like to dig deeper into automating some basic taste for now but haven’t yet

been using replit and lovable to make random shit for the trenches hit my ref so i can get free credits lmaoo lovable.dev/invite/ZL4RRAW