I'm blown away. This AI filmmaking workflow for precise camera control, multiple characters, and dialogue is insane: 1. Generate a start frame in Midjourney 2. Match the poses in Blender, animate the camera 3. Feed both to Seedance I didn't think this would work. Two consistent characters, solid performances, the move tracked perfectly through the entire beat. Even the soup looks great.

We talk about taste in design. Where does taste live in code?

Show the process, prove him wrong and you'll get a new sub x.com/Psshck26/status/206890…

The Rational Case For AAA on ThreeJS Or "why UE5 shipping a MCP was bad, real bad." The standard objection to ThreeJS as a serious game engine is technically accurate. It lacks the rendering fidelity, integrated tooling, physics depth, and asset pipeline maturity of Unreal Engine or Unity. A direct capability comparison favors the big engines on nearly every relevant metric. That objection mistakes the current state of a technology for its trajectory. The Cost Structure of Traditional AAA Production Traditional AAA production is built on high fixed costs: large headcount, expensive pipelines, long cycles, platform compliance, and heavy launch marketing. Iteration is costly. Every experiment carries real financial risk because the people and pipelines required are expensive and slow to mobilize. Studios respond predictably: fewer bets, earlier design locks, proven formats. AI-assisted development restructures this for smaller teams. A team running AI agents across production categories operates at a fundamentally different cost structure. Large studios can't adapt without risking public pressure and embarrassment. The Distribution Argument ThreeJS carries no distribution friction by default: no install, no launcher, no SDK, no licensing, no storefront submission. Link, play, embed, fork, update. No deployment pipeline required. Web frameworks exist, but they’re about code organization. Compile times run in seconds even at scale. Traditional platforms impose friction (deliberately!) at every stage: curation gates, platform fees, recertifications. A free SSL certificate is all you need in the browser. A browser game can reach real users within days of a working prototype. Behavioral data flows back in near real time. A feedback loop that scales directly with iteration speed. Traditional studios won't easily adopt this. Their infrastructure, milestone commitments, and platform relationships are built around the existing model. Attaching AI to a traditional pipeline improves efficiency at the margin but it won't restructure the feedback loop. Tooling WebGPU closes the theoretical gap (as browser limits evolve eventually). The performance ceiling of browser-based 3D is rising. ThreeJS is behind Unreal Engine 5 in rendering output at present. Open, modular, web-native tooling has a different scaling characteristic under AI assistance: it improves fastest where there are the most practitioners directing the most agents at the most specific problems. We can see the exponential growth of tooling around ThreeJS in realtime on X. Iteration and Market Entry Disruption in technology markets rarely begins at the top of an incumbent stack. It begins in segments the incumbents do not serve well and do not prioritize: experimental mechanics, web-native game formats, creator tools, procedural sandboxes, simulation toys, viral browser experiences, AI-generated worlds. These are also the segments where cost-per-experiment is the primary competitive variable and where ThreeJS combined with AI-assisted development is most competitive today. The correct strategic decision is to be active in the segments where the stack is already competitive, iterate on high volume and production experience and extend toward more demanding experiences as the tooling matures. Waiting for Three to match Unreal before committing to the platform means waiting until the window of structural advantage has already closed. So, invest in ThreeJS to profit from the next major cycle in gaming, or stay with current technologies to reduce your risk but cap your profits. Just stop acting like none of us know why we are taking this bet. We do. And if we're right, you all haters are cooked.

I made these models in under 5 minutes, the point of creating generative models is to help optimize the base then refine in blender!

2026: manual coding -> vibe coding 2027: manual 3D design -> prompting 3D design

11M downloads/week 🚀