There's a huge difference between making your own streaming box that people have to buy in order to use, and buying the company that makes the streaming box built into basically every new TV.
"Built into" - are you talking about them shipping with the app? What's the difference between that and the Apple TV, Prime Video, Netflix apps built into those TVs?
Most panels are from China. Panels have a very long lifetime. Over their lifetime they generate way more than their price in oil. Europe is not a huge producer of oil and relies on imports to sustain its usage. Sourcing panels is effectively reducing the amount of money leaving Europe in the long term.
China. With that said, they have so much solar PV capacity that they’re barely breaking even, even when exporting tens of GW of PV panels a month. I argue it’s a net positive the solar PV printers in China are kept in business to maintain their annual output, the world needs as much solar PV as it can produce as fast as possible.
> That... basically kills the entire gaming industry.
> Am I missing something serious here
Only just that the video games industry as we've known it for the past few decades is basically already dead—at best, it's a hollowed-out husk of what it once was.
This isn't really much of an excuse given contemporary models though. My current game project has a GUI editor mode and it was not difficult at all whatsoever to set it up such that whenever I run a debug build of the game:
- It opens to the editor mode rather than the gameplay mode on launch
- It makes a .run/ directory next to the executable if one doesn't already exist
- It makes a timestamped directory within .run/ for this current debug run
- It automatically records stdout to stdout.txt, stderr to stderr.txt, and a crash.txt if the game crashes, in the directory for this run
- When the “take debug screenshot” function is invoked (which can be done by pressing F12), it saves a timestamped (based on time since executable launched) screenshot in the directory for this run
- Editor actions and 3D camera movements are recorded to playback.txt in the directory for this run
With all of this in place, I can do a debug build, run the game, do something in the editor, and take one or more screenshots where things went wrong. Then, Codex can see the log files and screenshots and try to diagnose the problem. When attempting to fix the problem, it can automatically recompile the debug build and rerun it with a launch option that plays back the latest recording file, which does the same sequence of editor actions/camera movements and takes screenshots at the same points in the process. Then it can compare this to the initial recorded run and see what needs to be fixed.
We could be having a GUI renaissance right now but for various primarily aesthetic reasons people are churning out TUIs, and personally I think it's a huge mistake.
Why is it only a matter of time? The AI-as-a-service companies are going to continue to improve their products by improving both the part that could be reproduced in a self-hosted setup, but also the “secret sauce” they put on top of that to make it a better product. There is no incentive for this “secret sauce” to be something that can be reproduced for self-hosting, is there?
What secret sauce? We already have open source tooling for tool use, web browsing, and code execution/computer use. Open weight models will win in the end.
AIaaS might keep an edge with multi-modal agentic workflows, but for 80% of general use cases, no "secret sauce" needed, the open weight models are already there, and tooling is constantly getting better.
The bottleneck is the cost of local hardware right now.
The "secret sauce" is vendor lock-in. A textbook case is the vmware broadcom situation. Vmware was cheap so corporations found little reason to use open source. Broadcom made vmware expensive but now those corporations are finding out that it is a lot of work (aka expensive) to switch infrastructure.
I think a major incentive could be to sell hardware. If Apple is able to get their hands on a local LLM capable of covering a significant % of what people use ChatGPT for, the pitch they can offer is:
"Free, private, offline ChatGPT so long as your laptop has X GB of RAM"
Beyond that, I wouldn't underestimate the incentive of "because I can". The "secret sauce" you refer to is effectively just a DB & a while loop that feeds text to a bunch of tensors. If an indie dev decides they want to release something that dismantles the OpenAI & Anthropic moats, there really isn't all that big of a technical barrier stopping them.
LLM inference decode is heavily dependent on memory speed, not just having lots of memory. You can't say "X amount of ram" because the memory bandwidth on an M1 is 68.3 GB/s versus the 614 GB/s of an M5 Max, or a 4090's 1.01 TB/s over GDDR6X.
This basically creates a bottleneck at the oldest/cheapest Apple Silicon machines, which are already crippled for context prefill.
But honestly, obsoleting a huge number of otherwise great Apple Silicon machines is something Apple would moment consider a major "pro" of building a compelling local AI stack.
With how much speculation around the difficult time Apple has had getting people to upgrade from M1, I'm sure they'd jump at such an opportunity.
It's also industry standard to do so. I don't think I've ever seen a team outsource something like pathfinding. Maybe in the Unity/Unreal space, which I'm very unfamiliar with. Dependencies are generally speaking not viewed as a good thing. They become vanishingly rare outside of certain things like physics engines, sound engines, vegetation, etc. And usually higher quality proprietary ones are chosen over OSS (though this is changing with physics engines in particular, mostly thanks to Bullet)
NIH is a cultural pillar. Even scripting layers are relatively split on if they're in-house or not. It's not uncommon to find both an in-house fork of Lua + a few other completely custom scripting engines all serving their own purpose.
Pathfinding middleware has traditionally been a thing though, e.g. apart from the mentioned ReCast (which is the popular free solution) the commercial counterpart was PathEngine (looks like they are even still around: https://pathengine.com/overview/).
If you need to do efficient path finding on random triangle geometry (as opposed to running A* on simple quad or hex grids) it quickly gets tricky.
What has undeniably declined is the traditional "10k US-$ commercial middleware". Today the options are either free and open source (which can be extremely high quality, like Jolt: https://github.com/jrouwe/JoltPhysics) or fully featured engines like Godot, Unity or UE - but little inbetween those "extremes".
I did mention this was changing in particular with physics engines. That being said, proprietary dependencies still reign supreme in places like audio engines. Something like OpenAL isn't really a replacement for FMOD or Wwise. I know those off-the-shelf engines roll their own replacements, but then they also roll their own pathfinding and navmesh generation as far as I know.
I've seen people genuinely wonder how one would go about making a 2D platformer without a generalized third-party physics engine—as though every single classic 2D platformer didn't have its own simple, bespoke physics simulation!
There are entire classes (even genres) of video games that don't require substantial third-party library support, and it's frustrating to see that this seems to be more and more of a minority view as time goes on.
It's not a terribly long list, it seems, even if it's non-exhaustive.
> I see the industry is of full double standards
Absolutely, and that's why it's so great. No two companies are going to look the same. The culture of the video games industry is a rather unique one that's kept its archipelago syndrome alive against all odds. It'll be a sad day if that ever changes.
> So, how many years until we'll see results, then?
-0.75 years.
Software development output (features, bugs, products) - especially at smaller companies like startups - has already accelerated significantly, while software development hiring has stayed flat or declined. So there has been a dramatic increase in human-efficiency. To me, that seems like a result, although it's cold comfort as a software engineer.
You probably won't see this reflected as a multiplication of new apps because the app consumer's attention is already completely tapped. There's very little attention surface area left to capture.
> You don't think capitalists are able to generate profit off of these LLMs currently?
Not sure where you are reading that. I said that they are able to be far more human-efficient because of LLMs, implying they are able to reduce costs relative to outputs/revenue, which means higher profits.
reply