Obsessively building YouGame (remix. upload. play) Made rollback brawl and P+ w/ ranked https://xitter.cf/t.co/eoLKGDf1W5 prev Google & GOAT Melee Coach
Joined November 2021
- Tweets505
- Following204
- Followers441
- Likes373
Pinned Tweet
I want to clear up some things about AI coding and software engineering in 2026 that a lot of the smash community seem to misunderstand from the perspective of an ex-Google AI engineer and startup founder because SOMEONE has to bite the bullet and tell you the truth even if you hate them for it.
Most of you guys seem to have this perception that if code is written by AI, it is slop, and if it is artisanal grass fed human code, it is high quality, when the EXACT OPPOSITE is the truth now. Frontier AI is better at coding than 99.9% of humans now, like Phillip is better than 99.9% of smashers, which is why every single software engineer from mag 7 to 1 person startups (hi👋) is using AI to code, and how you can get YouGame in less than a month and rollback brawl/P+ with ranked in less than a week.
The skill expression comes primarily from product design, UI/UX taste, system architecture design and how effectively you can coordinate swarms of AI agents to do useful things.
Rollback brawl for example, took about 200 agents total across 4 machines doing at peak 16 simultaneous rollback simulations across 46000 replays and cost 5 Claude max subscriptions across 3 days. Ranked and matchmaking took another 2 days. Polishing it up with everyone’s feedback will probably take a couple days to a week more depending on my bandwidth. It’s not theft, it’s not me being a god, it’s a swarm of AI agents working together to make whatever we want real.
Already, agents have picked up every hate tweet, every comment, every kind bug report and have triaged, prioritized, implemented and deployed much of the work. I’M the problem, because now that you guys are giving so much feedback, the agents can finish work faster than I can review, test and guide it. I can’t keep up with you all.
So please do all your favorite game developers and modders and founders a favor and stop forcing them to live in this don’t ask don’t tell world where they are all AFRAID to admit they use AI to build, like LITERALLY everyone else, for fear that their game or work is going to get labeled unskilled AI slop and cancelled, and instead join and collaborate with us in experimenting with these latest tools and early builds to collectively make the games we could only DREAM of as kids.
We’re entering a world where every existing game is going to become essentially open source. Decompiled and recompiled by agent swarms that can do the work that used to take humans years, in days. It’s going to mean that ANYONE is going to be able to make games, modify games, mashup games, and build on top of games in whatever way they want.
We’ve figured out how to scale AI agents, but we haven’t figured out a way to scale the human taste, judgement, intuition and collaboration needed to go from first iteration slop to polished gem and how to make sure everyone is rewarded in proportion to their contribution when something succeeds, from individuals contributors to original game publishers.
That’s the problem YouGame is trying to solve. How to create a collaborative open source system where everyone can remix, upload, collaborate on and monetize games, and where the site itself evolves in the same way.
The goal for me with this isn’t money. The purpose for money in Yougame is exclusively to reward great creators and to put as much AI compute in the hands of the people as possible so that we can collectively make the best games possible, because you guys already know, that the best innovations in gaming don’t come from greedy profit seeking, they come from passionate individuals working together who do it for the love of the game and the joy of creation.
I believe so deeply in the world of amazing games that we can create. One where every modder and builder can create and share and be fairly compensated for their amazing work without fear of DMCA or lawsuit or cancellation, but I can’t do it alone.
We have to find a way to do it together.
Yes, Linux and steam deck folks, dw, I haven’t forgotten about you.
(Integration testing underway, all open source)
Luis Rosias retweeted
Some observations about the "total" automation of software engineering:
1. There are more software engineers now than in 2025, and demand is growing.
2. Software engineers have never worked harder than they do now.
3. They largely don't write code anymore, agents write ~100% of the code.
4. The bottom line of the companies that employs these engineers and agents (i.e. the terminal purpose of all this code) is typically not budging as a result of AI adoption.
This is why you have to think in terms of systems. Automating one aspect of a system creates its own friction and moves the bottlenecks elsewhere.
Alright, what should we rename Orca to?
Post other cool name ideas if you like them better than these two.
31%YG Dolphin
69%Baiji
86 votes • Final resultsTop Dawg is hosting the FIRST EVER P+ rollback tournament tonight at 8PM ET!
start.gg/tournament/top-dawg…
Come enter and see how far you can go!
Luis Rosias retweeted
Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.
- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.
The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.
So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.
Judging by my tl there is a growing gap in understanding of AI capability.
The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code.
But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along.
So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions.
TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.
Rolling out a bunch of the big changes I’ve been hinting at little by little today. If there’s any issue at any point today, leave a comment and it’ll be rolled back to the previous stable version within 15 minutes.
I’ll stop deploying changes and leave things stable a few hours before the tourney tonight at 8. Thanks to all the early adopters for dealing with the messy update process so far.
Should start to become much more smooth, stable and community driven from now on.
Continuously improving every day!
Btw I know I haven't been posting as much the last few days. The main reason is that I'm working on a big refactor and new set of features, which will allow improving P+/Brawl rollback, and well any game on YouGame for that matter, to be completely open source, collaborative and community driven.
This was my goal from the beginning, but the way I released and the technology I released with communicated the exact opposite.
So stay tuned, there's some really big changes coming and I think it's going to be awesome.
We're back btw! Lmk if it feels TOO fast.
I think it might make sense to go up to delay 3 since this shaves off 12 ms. I need to do like a thorough analysis of controller input to pixel on regular P+ and here.
Delay 3 would definitely make connections better with higher ping by quite a bit.
As an aside, we're really getting to the tail end of bugs now. Almost have everything resolved. More feature request than bugs at this point.
Netplay will be temporarily down for an hour or 2.
I'm pushing/testing a v-sync netplay upgrade which will reduce delay by 14ms and a security update.
Should feel even snappier and work better across longer distances.
This is the beauty of open source.
You cannot "steal" open source. You don't have to ask for permission. Anyone can do with it whatever they want.
The only thing you have to do is give credit to the people who came before and make it open to everyone who comes after.
YouGame is not required to play rollback brawl or P+. I'm working hard every day to make it the best place to play rollback brawl or P+, but you can play wherever you want.
Rollback brawl is open source, and it belongs to everyone.
Alright dev update time!
34 fixes shipped on launch day, reported by about 129 people in total, the biggest ones were:
- your own in-game controls (tap jump off, Z jump) work online
- P+ Code Menu locked in Ranked and Casual (sorry to spoil the fun wario-man mains)
- Brawl Rev 1 discs work on Brawl and P+, and play vs Rev 2
- no CPUs bug
- Orca.exe is signed (looks less sus now)
Old Mac OS fix and getting rid of that sus google login prompt box on windows in ~2 hours.
It's insane how quickly we can fix things when everyone is reporting.
Not much else done today except for bug fixing, but next priority will be basically notifying you guys when bugs you reported are complete and making it easier to report/upvote downvote bugs. Keep them coming!
(Best place to report right now is bug button on top menu bar, but comments is also fine)
Hopefully can get back to some features and UX improvements tomorrow if bugs are mostly settled down.
Thanks for bearing with the rough patches today, the sailing should be much smoother now!
If I could go back in time:
Contact brawlback team as soon as I had a working initial build and collaborate on playtesting and integration.
Announce build as an early alpha build that shows a lot of promise but needs the help of community feedback to rapidly improve.
Be transparent about AI use and explain how agents would try to fix all the bugs people found.
Make cashbox conditional on us finding all the bugs to make the experience very high quality.
But alas, I’m a dumb human, so I can only promise to be less dumb in the future.
github.com/Lrosias/orca-netp…
Here's the Orca repo. It is a fully open source clean fork of Dolphin. The architecture.md explains the differences between this Brawback, and Slippi as well as what the methodology was in developing it. I tried to have the explainer be easy for anyone to understand.
The next step will be to make Orca more generic. The Orca approach should work not only for Brawl but for any Wii game with very little extra effort. I will also be extracting out the parts of the code from orca that are specific to yougame and putting it in separate modules.
The orca core should be easily portable to any other project after that.
I am genuinely sorry for how this played out and any hard feelings I caused.
I, like most of the people involved in smash projects, got extremely excited at the possibility of making a really great smash online experience for smash games, and charged straight ahead without thinking when I found I hit a breakthough. It's been a non stop 16+ hour a day obsessive building session for the past week, and almost as intense for the 3 weeks before that. I have never been more sleep deprived in my life.
In my obsession to get this to people as soon as possible, I didn't stop to think enough about how all of this coming out all of a sudden would affect the people who were just as passionate as me about it.
I would really like to know how I could make this right with everyone involved, and how we can work better together moving forward.