frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: KradleBench – Multi-agent AI evals in Minecraft

https://www.kradle.ai?1
1•jamest•11m ago
Hi HN -

This is James and Kemal. In past lives we led Firebase and Tensorflow/JAX respectively. HN has been a particularly big part of my (James’) career and the trajectory I’ve taken would be meaningfully different without it. Thank you!

Our team shipped a benchmark: KradleBench. We test AI in Minecraft Worlds.

1. The worlds are multi-agent. Humans and other AIs are in there too, often pursuing their own goals. Collaboration and conflict skills fall out of this property.

2. The world are dynamic. They constantly change, which means that agents need to react. “Capture the Flag” has a single winning condition, and many different ways to achieve it. This leads to many emergent and creative strategies.

3. The worlds are real-time. They are not turn-based, they do not wait for AIs to finish thinking. Nature balanced speed and intelligence in humans, we can understand this in AIs too.

Underneath all this is Kradle, a platform to test and understand intelligence. It’s a petri dish where any intelligence type (LLM, World Model, Human) can be compared and generate useful information to improve current systems, whether it’s on capabilities or safety.

Some findings so far: - A skilled human was #1, until yesterday when Astra 6.2 surpassed them. We lose badly on solo tasks and out-compete on multi-agent ones - Astra 6.2 iterates on its strategy 4x faster than Fable 5.2. The Claude family tends to think hard and stick with strategies. The OpenAI models tend to be more reactive and iterative, which serves them well in multi-agent, dynamic, real-time evals.

We’ve got videos on all the top model’s page, and we’ll be producing them when a new model launches. Sign up on the site to get them and for when the platform opens.

Thanks for reading,

James & Kemal

VB Harness – Make local coding agents verify their own work

https://github.com/Viktotovich/vb-harness-for-agentic-building
1•vova2k•59s ago•0 comments

Local search for coding agents that knows which decision is still current

https://github.com/jacobemanuel/leks/blob/main/docs/comparison.md
1•Iniraug•1m ago•0 comments

Show HN: I made my own Voyager Golden Record

https://yashmouje.com/record
2•yashmouje•5m ago•0 comments

Pity the Teacher Because Student Traits More Significant Than Teachers/Schools

https://pubmed.ncbi.nlm.nih.gov/27919311/
1•CGMthrowaway•7m ago•0 comments

GTA VI

https://www.rockstargames.com/VI
1•TechTechTech•7m ago•0 comments

Show HN: Poptato – App launcher, command runner, and fuzzy menu for Linux

https://github.com/skorotkiewicz/poptato
1•modinfo•7m ago•0 comments

Eval for Email Classification

https://housecat.com/blog/email-classification
2•algent2•8m ago•0 comments

Show HN: KradleBench – Multi-agent AI evals in Minecraft

https://www.kradle.ai?1
2•jamest•11m ago•0 comments

GPT-6.2 Astra Formalizes Smoktunowicz's Gap Theorem

https://www.vals.ai/blogs/smoktunowicz-gap-theorem
1•theanonymousone•11m ago•0 comments

The AI Repo Man: Exploiting Agentic AI Debt Collection Platforms

https://eaton-works.com/2026/11/18/credit-nirvana-hack/
3•notfried•11m ago•0 comments

What to do now that Last.fm is going to charge for API access

https://www.coryd.dev/posts/2026/what-to-do-now-that-lastfm-is-going-to-charge-for-api-access
1•cdrnsf•14m ago•0 comments

Field Notes from China

https://parthchopra.substack.com/p/field-notes-from-china
1•probe•15m ago•0 comments

Ask HN: New Accounts

2•illidan1000•16m ago•3 comments

Hearthly, the Smart Home That Cares

https://jikol1906.github.io/Hearthly/
1•apetresc•16m ago•0 comments

FrontierBreach: Measuring frontier model attack susceptibility

https://www.promptarmor.com/frontier-breach
1•cmitsakis•16m ago•0 comments

Can a Phone Last 50 Years?

http://crywallcommunity.blogspot.com/2026/11/can-phone-last-50-years.html
1•mike5gao•18m ago•0 comments

California Sues OpenAI

https://apnews.com/article/chatgpt-openai-hugging-face-ai-risk-1dd7be56134b4f09fb2fb974803edc69
4•leafingensemble•18m ago•0 comments

Show HN: Fun with bicycle trainers and web APIs

https://joelriley.com/fun-with-trainers
2•okayestjoel•19m ago•0 comments

PlayStation network experiencing a service disruptions

https://status.playstation.com
4•ljahier•20m ago•0 comments

Cost per token explains 92% of a model's cost per task

https://micjm.com/blog/cost-per-task-vs-token/
2•seizethecheese•20m ago•0 comments

It Came from the Desert Enhanced (Recomp)

https://tim-rachor.itch.io/it-came-from-the-desert-enhanced
1•doener•21m ago•0 comments

Democrats probe whether Lutnick helped Cantor avoid Nvidia-linked losses

https://www.reuters.com/world/us/democrats-probe-whether-howard-lutnick-helped-cantor-avoid-nvidi...
2•JumpCrisscross•23m ago•0 comments

New proposed EU law will gut the protections for birds, fast track data centres

https://birdwatchireland.ie/
4•austinallegro•23m ago•1 comments

OpenAI expands Ganassi IndyCar partnership to full-season sponsorship of Palou

https://chipganassiracing.com/news/openai-expands-ganassi-partnership-to-full-season-sponsorship-...
1•tekdude•23m ago•1 comments

Rethinking "RL Generalizes, SFT Memorizes": The Role of SFT Data

https://arxiv.org/abs/2611.14912
1•Anon84•26m ago•0 comments

Recursive self-improvement has started [video]

https://www.youtube.com/watch?v=1NHPDzpfR6Y
1•doener•28m ago•0 comments

LeBron James sues makers of AI sports-betting app, saying they used his likeness

https://abcnews.com/Technology/wireStory/lebron-james-sues-makers-ai-sports-betting-app-138229791
1•randycupertino•28m ago•1 comments

Nvidia Kumo Tabular: Pushing the Accuracy-Efficiency Frontier

https://arxiv.org/abs/2611.23001
1•Anon84•29m ago•0 comments

16M Segmentation Masks

https://huggingface.co/datasets/ibm-esa-geospatial/TerraMesh-Masks
1•marklit•30m ago•0 comments

Show HN: SourcePost – a hand-reviewed directory for AI agents (API and MCP)

https://sourcepost.dev
1•User1988_m•30m ago•0 comments