Generality, Inc.

Humanity is advancing full tilt towards general intelligence. Generality, Inc. is building the measures to take us beyond pointwise progress. | YC W25

[saner @sanerp110]

We are launching @outshipai (YC w25) a hiring platform that records and analyzes how engineering candidates use coding agents.

Seeing a candidate work with AI can give you a lot of signal about their thought process and whether they’re the engineer you’re looking for.

[saner @sanerp110]

New LLM game tournament 🎲🕹️!
Four in a row from @basvanopheusden and @Ikuperwajs's "Expertise increases planning depth in
human gameplay"

... and our reasoning depth analysis. 🧵

[Ahmed @ah20im]

Impressive!
In my experience, GPT-5 is highly steerable, allowing it to demonstrate its intelligence across a wide range of use cases, including games.

[saner @sanerp110]

GPT-5 and o3-high are incredibly good at chess!

How good? They have a secret superpower 🧵(1/9)

[saner @sanerp110]

@UnslothAI This is really exciting! We just launched https://t.co/bTPTEgSDzr: hundreds of strategy game RL environments to improve LLM reasoning.

https://t.co/QZCJsU9Ko6

[Generality, Inc. @generalityinc]

btw. check out this Racing Kings analysis!

https://t.co/0dV6KIkdm4

[saner @sanerp110]

Racing Kings seems to be one of the hardest chess variants for LLMs on https://t.co/bTPTEgSDzr

And it reveals a lot about instruction following. 🧵

[Generality, Inc. @generalityinc]

Gold checkmark 🚀

[saner @sanerp110]

Games are actually great at measuring and teaching many skills we associate with intelligence.

Long horizon planning: Some games take 300+ moves. Many of them require players to consider multiple moves options, how an opponent would react, and reason about the best moves to make.

Instruction Following: They have long, complex rules and put LLMs in scenarios where they have to first reduce to a set of legal moves, and then pick the best move from this set. Game arena shows that models have an illegal move rate of +70% on some games.

and many more meta skills like creative problem solving, long context behavior.

We believe that these are generalizable skills that models can learn through games but apply in other areas like coding, task oriented dialogue, and more.

If you get a chance to read the reasoning traces on https://t.co/cn8xAsY1xz I think this will make a lot of sense.

[saner @sanerp110]

Racing Kings seems to be one of the hardest chess variants for LLMs on https://t.co/bTPTEgSDzr

And it reveals a lot about instruction following. 🧵

[Generality, Inc. @generalityinc]

@mvrckhckr @ycombinator @sama Spot on 🎯

[Generality, Inc. @generalityinc]

@dimeji_san @ycombinator I can explain!

[Generality, Inc. @generalityinc]

@ycombinator GPT-5 is goated!
https://t.co/uSFsjW2ObA

[saner @sanerp110]

Openai is dominating https://t.co/bTPTEgSDzr. (and deepseek v3.1 is incredibly good)

[Generality, Inc. @generalityinc]

@mvrckhckr @ycombinator Yes.
It was a joke around @sama saying "we will ask AGI how to make money once we have it"

[Generality, Inc. @generalityinc]

@mvrckhckr @ycombinator Making LLMs extremely smart and then asking them

[Y Combinator @ycombinator]

Game Arena from @generalityinc is the largest LLM strategy game tournament to date.

Games are great for measuring LLMs on instruction following, long-horizon planning, and problem-solving. In fact, models that are Olympiad-level at math and coding often struggle to make accurate moves in Game Arena (median illegal-move rate: 11.4%).

Game environments also resist contamination and saturation: every run is unique, so there’s no risk of training-data leakage, and the bar keeps rising as models improve.

Game Arena has digitized hundreds of board games into fresh new environments, and today it’s debuting its first tournament, with models like GPT-5 High, Claude Opus 4.1, and DeepSeek V3.1 from OpenAI, Anthropic, Qwen, DeepSeek, Google, and more going head-to-head.

You can watch game replays and full model reasonings on https://t.co/PkQysj7u3r.

Congrats on the launch, @sanerc110 and @kaylalee278!

[Generality, Inc. @generalityinc]

@ycombinator Which model should we add next?