Join /aichannel https://cura.network/aichannel/details?a

2 replies

1 recast

29 reactions

🪘free the weights🪘 https://huggingface.co/blog/nvidia/cosmos3edge

0 reply

0 recast

0 reaction

OpenAI's new Head of Strategic Futures (a.k.a. lobby guy) warns against AI communism after Kimi K3 😂 Gotta love the cry.

2 replies

0 recast

2 reactions

Got a Samsung SM-P610 tablet to replace the iPad for my 5yo daughter Claude installed LineageOS and now drives it entirely remotely, can debug, simulate taps, load content, log verbosely Basically making a fully custom OS What I did so far would have cost $50-100k pre-LLMs

5 replies

3 recasts

22 reactions

My logseq-based agent-connected knowledge graph is low key better than @garrytan's gstack https://github.com/mxjxn/hermes-logseq

1 reply

0 recast

2 reactions

https://www.dexerto.com/entertainment/china-bans-ai-boyfriends-and-girlfriends-over-addiction-and-birth-rate-concerns-3388737/

0 reply

2 recasts

5 reactions

As a dev, the one thing you can do right now, to improve your AI proficiency: learn about Harness Engineering. https://walkinglabs.github.io/learn-harness-engineering/en/

0 reply

0 recast

3 reactions

For those skeptical of further AI progress: here is why, even without another foundational breakthrough, and even if base-model reasoning improved only modestly from today, we should still expect substantial gains in economically valuable capability. Current models are often evaluated as though the model alone were the complete system. But on structured tasks, the same underlying model can often move from near-total failure to very high performance when placed inside a well-designed scaffolding and harness. https://schema-harness.github.io/ We're currently in the MS-DOS era of AI "operating systems". A large amount of present-day capability remains poorly elicited. In the future, with algorithmic improvements and vastly better AI processors, advanced AI systems will be able to run hundreds of Fable-level models in parallel at almost no cost. https://www.lesswrong.com/posts/75dnjiD8kv2khe9eQ/measuring-hardware-overhang

1 reply

0 recast

0 reaction

Who has tried Kimi K3?

1 reply

0 recast

0 reaction

Impossible Research created an agent harness called [schema] that plays games, writes code, reasons like a physicist, and saturates the ARC-AGI-3 benchmark. ARC-AGI-3 gives an agent a 64×64 grid plus legal actions, no rules, stated goal, or reward. The agent must discover both what the world is and how it works like a physicist: 1. State grounding -> identify objects, relations, and goals. 2. Mechanism discovery -> infer how these states change. [schema] handles the state and mechanism in one editable program, a symbolic world model. It designs experiments to verify hypotheses, backtests the program against history, and plans inside its world at zero action cost. Full blog: http://schema-harness.github.io Agent traces: http://huggingface.co/datasets/schema-harness/arc-agi-3-schema-traces

0 reply

0 recast

1 reaction

uptown-funk-your-bench just dropped https://www.tryai.dev/blog/ai-music-video-arena-claude-vs-gpt-5.6 Honestly a super interesting set of success and failure cases. Temporal and worldly cohesion are still clearly lacking but the visual at times are genuinely inventive. Curious if we are staring at the next Will Spaghetti.

0 reply

1 recast

6 reactions

if you're vibecoding something and you don't want to end up looking like all the other ai generated websites (you know the look), try aronia today. you get a distinct style/look for your website and a skill that enables your agent to sweat the details designers would usually agonise over so what you're building looks great. https://github.com/Chukwuka-Osakwe/aronia

4 replies

1 recast

14 reactions

why waste time say lot word when few word do trick? (eventually this will become post-human lexicon of glyphs)

0 reply

0 recast

0 reaction

From the former CTO of OpenAI, that's pretty cool they are getting into opensource. Let's make AI open again.

2 replies

0 recast

3 reactions

Grok Build is now open source https://github.com/xai-org/grok-build It joins Codex in being open source too https://github.com/openai/codex Who does not like free & open source software?

2 replies

1 recast

15 reactions