Posted by jttnr 4 hours ago
I spent so much time in the game playing the story, then playing skirmishes, building massive bases. "Hacking" the game to increase the max-unit-count to numbers my computer could barely manage.
Then I discovered HEAT.NET and started playing in a competitive clan. Then... life happened, and I have a lot of nostalgia for it now, but not enough time.
The first would be great, but probably very difficult technically. The second would probably cause a holy war in the community, as multiplayer RTS enjoyers glorify fighting against the genre's jank as an expression of skill (admittedly it would completely change all micro).
https://www.moddb.com/mods/loud-ai-supreme-commander-forged-...
With modern tooling, libraries and such, many of supcoms old technical issues can be address. (plus it now works on linux and mac!)
I think I'm close to a like for like version.
And of course the performance its leaps and bounds beyond what supcom could do, in the replies a few down, you can see a 64,000 unit air fight, and I'm pretty sure i can get that going at 60fps plus!
and end game vs the ai, is still 4x realtime, well beyond what supcom could do.
and of course higher player v player battles
https://forum.faforever.com/topic/10451/supcom-moho-engine-r...
"Punchy phrase. No X." "An open engine. Room to make it your own. "Same scene. Same instant. The same explosion. A different atmosphere." "A few things to know. The short answers."
Why is it always the same? What awful post-training, or synthetic data training pipeline, or RLHF gone wrong, produced this homogenised, low variety slop style? And it's totally possible for LLMs to write actually decent or at least varied prose and I would think Anthropic (or OpenAI) has the resources to fix the issue, they just choose not to.
Marketing copy wasn't this homogenous prior to LLMs, unless my memory fails me.
It was, or at least the most popular sites used it. LLMs just amplified it.
Like when you were a kid goofing around with some art supplies and you tried to mix every color of crayons, markers, watercolors, and/or colored pencils together to be both anything and everything all at the same time: It's just a sea of homogeneous brown yuck.
That homogenized yuck might work for code that is ultimately interpreted by a machine and that performs some kind of human-described function, but in writing (and in menu boards) it's just highly-emphatic sludge.
That _is_ the problem: ostensibly serious writing laced with marketing / clickbaity patterns is off-putting.
A) I was hit by a multiple _second_ load time (on a 1Gbps pipe btw)
B) The in game cursor was _visibly_ lagging
C) The gameplay itself was stuttery (and I am almost certain there is a hard one frame delay between most actions)
Then I thought, alright it could be Firefox, so I loaded it up in a fresh chromium image, and somehow it's worse. Then finally I looked at the source and it's very obvious why the performance is atrocious (the guy didn't really bother to tell his agent, "hey man, let's benchmark this and keep performance in check, maybe don't spam a billion draw calls everywhere y'know, maybe
Values below are medians across runs, including the median of each run's maximum tick time. Each pair reads *Go 1.25.0 → Go 1.27.1*; timings are ms.
| Mode | Process CPU/tick | Wall p50 | Wall p95 | Run maximum | Allocated bytes/tick |
|---|---:|---:|---:|---:|---:|
| Modern | 1.877 → 1.812 | 1.744 → 1.665 | 2.765 → 2.705 | 8.934 → 8.740 | 86,428 → 86,428 |
| Strict 3.1 | 1.210 → 1.101 | 1.144 → 1.041 | 1.577 → 1.498 | 4.555 → 4.594 | 19,871 → 19,871 |
| Community 3.9 | 1.591 → 1.447 | 1.433 → 1.290 | 2.407 → 2.299 | 9.323 → 9.310 | 16,721 → 16,721 |
fix this y'know ")
I would be surprised if that stood up in court but I doubt anyone cares for TA at this point anyway.