frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

SIMA 2: An Agent That Plays, Reasons, and Learns with You in Virtual 3D Worlds

https://deepmind.google/blog/sima-2-an-agent-that-plays-reasons-and-learns-with-you-in-virtual-3d-worlds/
41•meetpateltech•1h ago

Comments

Workaccount2•1h ago
>We’ve observed that, throughout the course of training, SIMA 2 agents can perform increasingly complex and new tasks, bootstrapped by trial-and-error and Gemini-based feedback.

>In subsequent training, SIMA 2’s own experience data can then be used to train the next, even more capable version of the agent. We were even able to leverage SIMA 2’s capacity for self-improvement in newly created Genie environments – a major milestone toward training general agents across diverse, generated worlds.

Pretty neat, I wonder how that works with Gemini, I suppose SIMA is a model (agent?) that runs on top of it?

FuckButtons•14m ago
That’s what it sounded like to me, a plain text interface between two distinct systems.
ukuina•1h ago
At 0:52 in their demo video, there is a grammatical inconsistency in the agent's text output. The annotations in the video are therefore suspected to be created by humans after the fact. Is Google up to their old marketing/hyping tricks again?

> SIMA 2 Reasoning:

> The user wants me to go to the ‘tomato house’. Based on the description ‘ripe tomato’, I identify the red house down the street.

m_w_•1h ago
I can't speak to the content of the actual game being played, but it wouldn't surprise me if there was an in-game text prompt:

> "The house that looks like a ripe tomato!"

that was transformed into a "user prompt" in a more instructional format

> "Go to the tomato house"

And both were used in the agent output. At least the Y-axes on the graphs look more reasonable than some other recent benchmarks.

golol•54m ago
The gap between high level and low level control of robots is closing. Right now thousands of hours of task specific training data is being collected and trained on to create models that can control robots to execute specific tasks in specific contexts. This essentially turns the operation of a robot into a kind of video game, where inputs are only needed a in low-dimensional abstract form, such as "empty the dishwasher" or "repeat what I do" or "put your finger in the loop and pull the string". This will be combined with high-level control agents like SIMA 2 to create useful real-world robots.
oersted•42m ago
I get why they do it, they are a business. I just wish Google would get off their ivory tower and build in the open more like they used to (did they? maybe I'm misremembering...).

They've acquired this bad habit of keeping all their scientific experiments closed by default and just publishing press releases. I wish it was open-source by default and closed just when there's a good reason.

Don't get me wrong, I suppose this is more of a compliment. I really like what they are doing and I wish we could all participate in these advances.

Zed Is Our Office

https://zed.dev/blog/zed-is-our-office
101•sagacity•1h ago•28 comments

Hemp Ban Hidden Inside Government Shutdown Bill

https://hightimes.com/news/politics/hemp-ban-hidden-inside-government-shutdown-bill/
75•bilsbie•1h ago•39 comments

GitHub Partial Outage

https://www.githubstatus.com/incidents/1jw8ltnr1qrj
81•danfritz•2h ago•42 comments

Launch HN: Tweeks (YC W25) – Browser extension to de-enshittify the web

https://www.tweeks.io/onboarding
30•jmadeano•1h ago•24 comments

The Monks in the Casino

https://www.derekthompson.org/p/the-monks-in-the-casino
26•pavel_lishin•50m ago•3 comments

BAML is hiring compilers/rust engineers (YC W23)

https://github.com/BoundaryML/baml/tree/canary/jobs
1•hellovai•10m ago

Checkout.com hacked, refuses ransom payment, donates to security labs

https://www.checkout.com/blog/protecting-our-merchants-standing-up-to-extortion
372•StrangeSound•7h ago•197 comments

Blender Lab

https://www.blender.org/news/introducing-blender-lab/
105•radeeyate•3h ago•33 comments

Android developer verification: Early access starts

https://android-developers.googleblog.com/2025/11/android-developer-verification-early.html
1222•erohead•16h ago•555 comments

SIMA 2: An Agent That Plays, Reasons, and Learns with You in Virtual 3D Worlds

https://deepmind.google/blog/sima-2-an-agent-that-plays-reasons-and-learns-with-you-in-virtual-3d...
41•meetpateltech•1h ago•6 comments

Kratos - Cloud native Auth0 open-source alternative (self-hosted)

https://github.com/ory/kratos
55•curtistyr•2h ago•37 comments

Denx (a.k.a. U-Boot) Retires

https://www.denx.de/
47•synergy20•3h ago•9 comments

We cut our Mongo DB costs by 90% by moving to Hetzner

https://prosopo.io/blog/we-cut-our-mongodb-costs-by-90-percent/
77•arbol•1h ago•57 comments

Tesla Is Recalling Cybertrucks Again. Yep, More Pieces Are Falling Off

https://www.popularmechanics.com/cars/hybrid-electric/a69384091/cybertruck-lightbar-recall/
125•2OEH8eoCRo0•1h ago•61 comments

Heartbeats in Distributed Systems

https://arpitbhayani.me/blogs/heartbeats-in-distributed-systems/
34•sebg•3h ago•8 comments

European Nations Decide Against Acquiring Boeing E-7 Awacs Aircraft

https://defensemirror.com/news/40527/European_Nations_Decide_Against_Acquiring_Boeing_E_7_AWACS_A...
79•saubeidl•1h ago•83 comments

Britain's railway privatization was an abject failure

https://www.rosalux.de/en/news/id/53917/britains-railway-privatization-was-an-abject-failure
362•robtherobber•3h ago•295 comments

Pebble: How to Build a Smartwatch: Software – Setting Expectations and Roadmap

https://ericmigi.com/blog/how-to-build-a-smartwatch-software-setting-expectations-and-roadmap/
14•teekert•2h ago•0 comments

Human Fovea Detector

https://www.shadertoy.com/view/4dsXzM
365•AbuAssar•16h ago•77 comments

Android 16 QPR1 is being pushed to the Android Open Source Project

https://grapheneos.social/@GrapheneOS/115533432439509433
205•uneven9434•13h ago•103 comments

Steam Machine

https://store.steampowered.com/sale/steammachine
2469•davikr•23h ago•1155 comments

A Challenge to Roboticists: My Humanoid Olympics

https://spectrum.ieee.org/humanoid-robot-olympics
23•quapster•1w ago•4 comments

Seed. LINE's Custom Typeface

https://seed.line.me/index_en.html
78•totetsu•7h ago•36 comments

Reverse Engineering Yaesu FT-70D Firmware Encryption

https://landaire.net/reversing-yaesu-firmware-encryption/
100•austinallegro•9h ago•14 comments

COBOL to Kotlin via Formal Models (IR and Alloy and Golden Master)

https://marcoeg.medium.com/from-cobol-to-kotlin-795920b1f371
17•marcoeg•5d ago•1 comments

Homebrew no longer allows bypassing Gatekeeper for unsigned/unnotarized software

https://github.com/Homebrew/brew/issues/20755
300•firexcy•19h ago•231 comments

Switching from GPG to Age

https://luke.hsiao.dev/blog/gpg-to-age/
68•speckx•1w ago•43 comments

GPT-5.1: A smarter, more conversational ChatGPT

https://openai.com/index/gpt-5-1/
487•tedsanders•22h ago•611 comments

Continuous Autoregressive Language Models

https://arxiv.org/abs/2510.27688
88•Anon84•1w ago•7 comments

Shader Glass

https://github.com/mausimus/ShaderGlass
58•erickhill•5d ago•10 comments