Loading video...
Video Failed to Load
My vtuber website progress 😊 Added/refine some interpolation and imported VRMA pose from VRMPosingDesktop. Also added sway + IK bones! Texts are slop placeholders though lol #VRM #b3d #ThreeJS #DaisyUI #Tailwind
21 Comments

I've wanted to make something like this for a very long time

My god i need to do something this. Thats so cool.

HOLY.. looks so sick! I like it 0.o

so cool!!!! it kinda reminds me of the persona menu screens

ill subscribe to Cursor next week, any tips for me to stretch usage?

I use plan mode a lot. Never do one shots. Just follow regular dev patterns and write plan first so you can review implementations prior to building. I never use anything else other than Composer/Grok and it's been working fine for me. You get extra limits for these models.

@neko23423 This is the way, basically AI-SDLC 😎

@neko23423 𝒮𝓅𝑒𝒸 𝒟𝓇𝒾𝓋𝑒𝓃 𝒟𝑒𝓋𝑒𝓁𝑜𝓅𝓂𝑒𝓃𝓉 😎 Or whatever buzz word people uses nowadays. Personally I don't like the term vibecoding because it sounds like I'm just rolling the dice to get the outcome I want without streering lol

Heheh no worries, its the wild west stage of things at this point and everyone's still figuring things out what to call things and what works for them 😅 I think the term will stick tho, as Anthropic released an AI-SDLC playbook, my guess it will pick up in the vernacular soon Where I work we kinda built and adapted our own playbook, beginning with AI-Product DLC and up to the SDLC.. we also named it as such 😁 Anyways I digress, all I mean to say is that you're doing well, and that you'll see more consistency and stability in your codebase 👍

At my irl work we have Cursor rules (CLAUDE.MD? I don't use Claude lol) defining code patterns and architecture, basically SWE standards. AI is good with following patterns so that keeps our codebase consistent. We still need to review code/architecture though, I don't trust AI enough to do loop engineering 😂

Yah true, loop engineering is still something not easy to do... If you're token-rich you can just give Claude Opus 4.8 a rubric to critique PRs and you're set For enterprise where you're paying API rates it's atrociously tough to balance quality vs cost 🫠 Tbf, it took us almost an entire year of experimenting to get it to a useable degree, and building an "inner harness" inside of Cursor made it easier to scale PR reviews But yep, final human review an approval still is needed, we made it a point that the person still is accountable for their agent's output 🤷♂️

We used to look at Cursor analytic page to see who used the most token. Now being on top is like a symbol of shame. 😂 I don't work in a SAAS company but our work is systems critical that we have to go back and forth to get so many feedback/validation prior to release and there are a lot of gotchas with how our legacy systems are setup. Loop engineering is good with Fast Waterfall/product releases but I don't think it will ever work for me. 🫠

impressive for not opening a single file of code

slop

Love this!

That looks sick!

reminds me of @TensorAlice waifu

Did you model the character yourself?

What application interface is this?

It's Cursor IDE!

😍😍
