Comments URL: https://news.ycombinator.com/item?id=48749928

Points: 2

# Comments: 0

I hadn't considered HTML and I'm definitely going to try this.

gnu and emacs already have a long history of cli and text friendly solutions that LLM dev agents can easily use and are trained on.

Or for structured data, just use a database. Dev agents can work with SQL just fine.

I didn't know about Kiro specs. I've been playing around with my own org-mode based approach with mixed success in keeping dev agent work tracked:

https://github.com/farra/dev-agent-work

My take: The "not a game" positioning feels like surrendering before the real design challenge begins. With our first experiment (Retail Mage [1]), we deliberately created a hybrid between traditional game structure and AI improvisation. We released it precisely because we were getting similar feedback from folks that an early implementation was too open ended and didn't "feel like a game." While I think we proved our point, we also learned a lot (including some mistakes we made).

That said, I _also_ think we can build _more than games_ with this technology. I think we’re edging into a new genre or medium: more improvisational and participatory than either traditional games or film. Structure still matters in this space, though. Without it you're really in the fanfic engine or writing assistant space. That’s fine! But it's a different goal.

Honestly, I'm very interested in this space. I was just speaking on a panel about this Monday at GamesBeat in LA and a few of us have a new podcast about this topic "Playing With Inference" [2] with folks like Nick from AI Dungeon as guests.

[1] https://www.jamandtea.studio/news/making-retail-mage-a-new-a... [2] https://playingwithinference.com/

Some other highlights, specific to the Hacker News folks:

- We have our own AI backend, running in AWS. - Costs were a major issue and we did a lot to lower it - We use a hybrid approach of mixing traditional game AI with machine learning - We have not just AI NPCs, but an AI world interaction model

Anyway, happy to answer questions!

Comments URL: https://news.ycombinator.com/item?id=43366815

Points: 3

# Comments: 1

"So we have won. That is decisive. The world will be never ever like before. Globalists have lost their final combat. The future is finally open. I am really happy."

https://x.com/AGDugin/status/1854136340184490282

Dugin wasn’t exaggerating when he said, “We have won.”

Today’s White House display is exactly what he meant. A decade ago, the GOP would’ve never echoed Russian talking points. Now, they’re openly aligning with them.

This isn’t just isolationism—it’s the dismantling of U.S. hegemony from within.

We use LLMs for real-time gameplay. We released Retail Mage last year on Steam as a tech demo of what we can do. We've found FP8 to be the current sweet spot for accuracy and performance for our use case. It's one of the many techniques we applied last year to bring our real-time inference costs down by 3 orders of magnitude to make Retail Mage releasable.

Anyway, we're just sharing this with anyone who finds it useful.

You can read a few more note by our ML engineer, Yudi, on LinkedIn here:

https://www.linkedin.com/feed/update/urn:li:activity:7290455...

Comments URL: https://news.ycombinator.com/item?id=42870523

Points: 1

# Comments: 1

https://distill.pub/2020/growing-ca/

"Differentiable Model of Morphogenesis"

I like to joke that Retail Mage is "late stage capitalism come for Diagon Alley." You and your friends are mages working the end of your shift at Mage Mart, seeking 5 star reviews from the wacky customers and trying not to burn the shop down, which sometimes happens.

We've built what we think is one of the more sophisticated implementations of GenAI in games so far. NPCs are powered by embodied AI agents that can observe their environment, plan actions, remember interactions, and engage in natural conversation. Going beyond AI agents, we also introduce an AI-based interaction system: Players and NPCs can just type or say what they want to do, and we translate that into the 3D simulation. This approach lets us bring "rulings, not rules" style gameplay to video games for the first time.

The technical journey has been... interesting! Some highlights:

- All inference runs in the cloud. We're part of Amazon's Generative AI Accelerator (the only gaming company)

- We manage inference ourselves, using some of the latest techniques in batching, forking and KV-cache reuse

- Over the last year, we've reduced our inference cost by 3 orders of magnitude

- After testing over two dozen models, we're currently using Google's Gemma

- We have our own in-house AI tools for Unreal Engine, including in-engine/runtime inference testing

- We developed our own natural language unit testing suite for our models and features

And we launch today! Try it out: https://store.steampowered.com/app/3224380/Retail_Mage/

Retail Mage is just the start. We have so many other prototypes and experiments. As a game developer, I'm legitimately excited to bring a new genre of games to players. Happy to answer questions!

Comments URL: https://news.ycombinator.com/item?id=42129310

Points: 4

# Comments: 0

We're running this on GPUs hosted in AWS (we're part of the AWS GenAI Accelerator) where we manage the inference ourselves (for feature, cost and performance reasons).

Getting realtime inference to function, at even a modest scale, for a realtime 3D video game has been... fun. And definitely interesting. This game includes just a few of the AI features we've prototyped, so we're excited to build more games with more advanced features soon.

Steam page: https://store.steampowered.com/app/3224380/Retail_Mage/

Happy to answer any questions!

Comments URL: https://news.ycombinator.com/item?id=41881976

Points: 1

# Comments: 2

If you don't actively stay up with the tech, your understanding will atrophy. Consequently, because you intend to remain a PM, your tech will never be as strong as someone who is dedicated to it and you need to be okay with that.

Another classic example for this are engineering managers: people management (like project/product management) is its own discipline with a deep set of skills, research, training, etc. To become a good EM means you _will_ sacrifice some of your technical edge. That "deep technical understanding" that good EM's maintain is built on a foundation of years as a practitioner and, even then, requires continual investment.

I don't mean to discourage you, I just want to ensure you have realistic expectations.

Given all that and given your goals, I think you should just ask yourself a question only you can answer: how do you best learn?

If that's in the classroom, then go get that masters (online or in person). If that's with hands on projects, then set aside time for tinkering and get going. Likewise with books, videos, meet-ups, etc. You know what motivates you and what techniques keep you focused (because you'll need that discipline).

Regardless, I wish you luck and enjoy the journey!

In my case:

- I had 10+ years of experience in linux/unix development

- Career change to AAA game development which is predominantly Windows based

- Day to day do Unreal C++ developing in Windows with Windows toolchains

- Still use WSL (and thus bash) for everything else, including DevOps related work, cloud/container development work, and personal notetaking/productivity via emacs+org-mode.

So, some of us have reasons to co-exist in multiple OS's. I tend to write up scripts in Python if I can so that there's at least a chance that I can run them in both Windows + Unix environments.

I know it's not as popular or sexy as it used to be, but the whole point of a foundation like Apache was to avoid these situations, even more than the way the Linux Foundation is setup. Apache _explicitly_ manages projects to avoid these downsides.

- Single corporation ownership. Projects cannot get out of the Incubator unless they demonstrate a diverse and healthy community. That doesn't mean popular, it doesn't necessarily mean best-in-class, but it means that there shouldn't be just one entity backing a project.

- Membership in Apache is _personal_ not a seat for a given company. If you're a committer on an Apache project and you move jobs, you're _still_ a committer on that project

- The Foundation owns the trademarks. There have been fights about this in the past, but the whole idea is that the _community_ owns the name, so some corporation can't claim to be the sole or official owner by naming their company or product after the open source product.

The core premise of the Apache Software Foundation is community over code, that healthy, diverse communities have a better chance of standing the test of time than open source projects backed by a single individual or company. That's the thesis at least.

The is starkly different from several other foundations, notably the Linux Foundation or Eclipse Foundation which are modeled more around industry consortiums.

Both models have their place, but I believe Apache better models the core values many of us feel strongly about when it comes to free and open source software.

That said, this is still good work, especially considering your experience and age (I checked out your 2D physics library on github and I found the trailer on youtube).

One suggestion: As someone in the game industry, I personally wouldn't label it an MMO as it doesn't include many of the features, particularly the "massive" part of MMO, that is usually associated with that term. The work is already impressive enough, you don't need to give it a label that could open it to criticism.

Otherwise, this is solid work. If you're aiming for a career in the game industry eventually, it's a good foundation to build on.