This reminds me of Douglas Adams’ Hitchhiker’s Guide, where a civilization decides that phone sanitizers are useless, until removing them quietly collapses everything else. Declaring work “non-essential to progress” usually just means we don’t understand its role.

It’s something I’ve been noticing more and more over the last few years: everything gets turned into some kind of monetization scheme, or at least it feels like nothing has value unless it can be turned into money.

That said, good luck with your endeavor!

P.S.: RAMBnB? Do you eat RAM for breakfast in bed? Bold choice! :D

Reading through the (first few) comments and seeing people defending the use of pure AI tools is really disheartening. I mean, they’re not asking for much just that one reviews and understands what the AI produced for them.

The term “full stack” works fine within its usual context, but when viewed more broadly, it becomes misleading and, in my opinion, problematic.

Allocating a very large share of advanced memory production, especially HBM and high-end DRAM, which are critical for almost all modern technology (and even many non-tech products like household appliances) to a small number of U.S. centric AI players risks distorting the global market and limiting availability for other industries.

Even within Samsung itself, the Mobile eXperience (MX) Business (smartphones) is not guaranteed preferential access to memory from Samsung’s Device Solutions (DS) Division, which includes the Memory Business. If internal customers are forced to source DRAM elsewhere due to pricing or capacity constraints, this could eventually become economically problematic for a country that relies very heavily on semiconductor and technology exports.

I am pretty sure, in the next year we will see a wave of low end ram components coming out of china.

Yeah, I used ChatGPT to help me write this answer ;) (Unlike JPEGs, it works at the right abstraction level for text.)

I think the core issue isn’t push vs pull or frame scheduling, but why you’re sending frames at all. Your use case reads much more like replicating textual/stateful UI than streaming video.

The fact that JPEG “works” because the client pulls frames on demand is kind of the tell — you’ve built a demand-driven protocol, then used it to fetch pixels. That avoids queuing, sure, but it’s also sidestepping video semantics you don’t actually need.

Most of what users care about here is text, cursor position, scroll state, and low interaction latency. JPEG succeeds not because it’s old and robust, but because it accidentally approximates an event-driven model.

Totally fair points about UDP + Kubernetes + enterprise ingress. But those same constraints apply just as well to structured state updates or terminal-style protocols over HTTPS — without dragging a framebuffer along.

Pragmatic solution, real struggle — but it feels like a text/state problem being forced through a video abstraction, and JPEG is just the least bad escape hatch.

— a human (mostly)

This one looks more like a PowerShell automation and debloating script for power users than a classic one-click optimizer, but it still requires knowing exactly what each tweak does. Used without that understanding, tools like this can easily create confusing problems.

That said, this does follow a very old and well-documented pattern: build a consumer image generation tool, and a significant portion of users will try to push it toward sexual or nude imagery, especially involving women.

Even companies with massive resources struggle here. Try generating anything even mildly suggestive involving women with ChatGPT and see how many hoops you have to jump through and that’s after multiple layers of prompt and output filtering.

At that point, content moderation becomes an arms race. Keywords, rate limits, paid tiers, moderation APIs, users will route around all of it. Without huge ongoing investment, it’s a battle that’s very hard to win.

So your conclusion that the problem isn’t the product but the market resonates. A B2B pivot makes a lot of sense, because the incentives and user behavior are fundamentally different.

Edit: Even OpenAI seems to be acknowledging the limits here and has indicated plans for some form of adult mode next year. It will be interesting to see whether that also includes more relaxed image generation policies.