aichannel

aichannel

practical ai news, tips, takes, wips, reviews, stories, and Q&A from the front lines of the ai unfolding.

nicholas 🧨 pfp

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

1 reply

4 recasts

18 reactions

Igor Yuzovitskiy pfp

Igor Yuzovitskiy

If you’re making an LLM .txt file for prompt engineers: 1. Keep it short—fits in one Cursor paste 2. One file only, no splits 3. Include basic sample prompts (e.g. user auth) MCPs don’t work well yet. Keep it simple.

1 reply

2 recasts

5 reactions

᠎ pfp

https://huggingface.co/collections/nvidia/llama-nemotron-67d92346030a2691293f200b https://arxiv.org/abs/2505.00949 reasoning models are "Wizards" to normal LLM chatbot "Villagers". we'll see two classes of behavior: -novel capability like Gemini finally beating Pokemon Blue -mature capability like DeepResearch with fewer calls of fewer parameters

1 reply

0 recast

0 reaction

Eun-oia pfp

Sketching is a conversation between hand and mind. AI’s more like a mirror, reflecting, not questioning⁉️ AI generated ::

10 replies

4 recasts

20 reactions

kazani.base.eth 🦂 pfp

kazani.base.eth 🦂

Interesting prompting strategy 👀 https://x.com/gregisenberg/status/1918661279641723158

2 replies

0 recast

2 reactions

kazani.base.eth 🦂 pfp

kazani.base.eth 🦂

How AI is changing our relationship to work | DW Documentary https://m.youtube.com/watch?v=xbE97Jra6io

0 reply

0 recast

0 reaction

David T Phung ⚛️⚡️🚀 pfp

David T Phung ⚛️⚡️🚀

https://youtu.be/nkhrEnuZi20?si=33DtP7Ysrf_agax-

0 reply

0 recast

1 reaction

simon pfp

https://www.forbes.com/sites/jackkelly/2025/05/04/its-time-to-get-concerned-klarna-ups-duolingo-cisco-and-many-other-companies-are-replacing-workers-with-ai/

1 reply

0 recast

2 reactions

Jacob pfp

i think it's more about outsourcing tasks + skills some tasks you could do but there isn't infinite time some skills you could learn but there isn't infinite time intelligence is maybe the most indispensable quality that augments the use of ai, which ai can't replace on its own

3 replies

1 recast

9 reactions

Justin Hunter pfp

@polluterofminds

This is so much fun. https://youtu.be/i21wN6CDsE0

0 reply

1 recast

6 reactions

kazani.base.eth 🦂 pfp

kazani.base.eth 🦂

You sent the message, but did you write it? https://davidduncan.substack.com/p/you-sent-the-message-but-did-you

0 reply

0 recast

0 reaction

shoni.eth pfp

Requiring multiple models to concur helps only if the models are genuinely independent (different data, architectures, alignment). Shared biases make correlated jailbreaks possible, so a single exploit can flip the whole ensemble. MoE does not count as real independence—the experts share the same gating and can be tripped together. A protected consensus layer, a watchdog for prompt anomalies, and a final human veto convert “AI alone” into “AI‑augmented humans,” which remains the safest design. Ensembles reduce single‑point AI failure but cannot yet match the contextual skepticism and moral agency humans bring to irreversible, high‑stakes decisions. WDYT? https://chatgpt.com/share/6816f6f6-b2fc-8010-a97e-e7b617d3ef3f

1 reply

0 recast

8 reactions

D-wayñe 🎩🕴 pfp

D-wayñe 🎩🕴

https://x.com/i/grok/share/c73aptKc93odFf06F7LlGfAO5

0 reply

0 recast

5 reactions

shoni.eth pfp

https://crowdfund.seedclub.com/c/cm9ujnboq0019s50t2ieiqcks

0 reply

1 recast

12 reactions

shoni.eth pfp

noobs always claim models are better based on benchmark snapshots, yet have little to no understanding of the benchmark or evaluation itself.

0 reply

1 recast

5 reactions