Clear advancement of the method is its low level of representation can be used to construct high-level abstractions, if necessary.

2 notes:

1. The game has nice "dynamics" which makes the agents' behaviours interpretable through the lense of self-organisation.
2. Unfortunately, this project has one of the worst implementations that I've ever seen, but looks still viable as a code game for OpenSpiel, idk.

Wdyt? (if any, ofc).

Sorry for a late one, but I noticed an interesting application of alphaevolve. It's something in between environment and algorithm generation — Code Wars game based on LLMs.

pub.sakana.ai/drq/

sometimes it's getting tiring of being alone all time in professional sense.

Resource-Efficient Model-Free Reinforcement Learning for Board Games Board games have long served as complex decision-making benchmarks in artificial intelligence. In this field, search-based reinforcement learning methods such as AlphaZero have achieved remarkable suc...

This approach looks weirdly similar to the Magnetic Mirror Descent:
arxiv.org/abs/2602.10894

you have "computer" vision

Attention and Augmented Recurrent Neural Networks A visual overview of neural attention, and the powerful extensions of neural networks being built on top of it.

Also, a very subtle but a reminder that you *can* represent a transformer model as a NTM.
distill.pub/2016/augment...

Technically, we can build programmes of varying complexites, true.

I wish this startup all the best

what does a quokka say if they want to kill themselves?

However, I think one of the authors has a little bit unconventional definition of "operational system." OLTP? Or operating system?

Can LLMs Be Computers? | Percepta We build a computer inside a transformer — executing arbitrary C programs for millions of steps with exponentially faster inference via 2D attention heads.

Percepta does nice things. Blog

1. www.percepta.ai/blog/can-llm...
2. www.percepta.ai/blog/beyond-...

Why are google's own models so incompetent about their own software?

This seems illogical.

A thing from cooperative AI:

www.cooperativeai.com/curriculum/1...

based

can you please explain what singularity means in that screenshot and how it differers from a matrix being singular or its singular values?

wait what

fantastic

It's this time of year again: why?

sigh...

What can a 70-year-old idea about digital organisms teach us about the future of AI and artificial life? 🧬💡

📄 Read our report: arxiv.org/abs/2603.08463

#ArtificialLife #ComplexSystems #OpenEndedEvolution #ArtificialIntelligence #ALife

Magic: The Gathering is Turing Complete $\textit{Magic: The Gathering}$ is a popular and famously complicated trading card game about magical combat. In this paper we show that optimal play in real-world $\textit{Magic}$ is at least as hard...

a more understandable and sweet result:
arxiv.org/abs/1904.09828

you can do programme synthesis with Magic The Gathering, yo

#woke

However, the hiring web sites state about equal opportunities. It doesn't work in my head like that. Also, no, students from third world countries do not differ much from ones from EU or NA universities because the ed system is pretty much the same everywhere. But the former lack those opportunites.

A thought. I read a lot of internship report type papers and everywhere see that the students are from very top universities. But what the point for top companies do that. Those students work with professors that have (in)direct affilications with the said companies and can provide the resources.

Deep Incentive Design with Differentiable Equilibrium Blocks Automated design of multi-agent interactions with desirable equilibrium outcomes is inherently difficult due to the computational hardness, non-uniqueness, and instability of the resulting equilibria....

A based find from scholar-inbox:
arxiv.org/abs/2603.07705

Optimal play in Yu-Gi-Oh! TCG is hard Motivated by the results for Magic: The Gathering presented in [CBH20] and [Bid20], we study a computability problem related to optimal play in Yu-Gi-Oh! Trading Card Game, a popular card game develop...

I am not that deep into the complexity but this paper results are interesting: arxiv.org/abs/2603.02863

can't compute the optimal strategy =(

I think I've put it together

how can one be excited about the fact of the publication rather than about importance of its result?

reproducibility should be the process: it may be a product or may be a guidace. For example, there are some people how know how AlphaZero really works because they'd implemented it. Because of the business side they can't provide an implementationm, however, can consult on expected results/issues.