Nari

Freelance Full-Stack Web Engineer & Tech Advisor

[Unsloth AI @UnslothAI]

You can now run Kimi K2 locally with our Dynamic 1.8-bit GGUFs!

We shrank the full 1.1TB model to just 245GB (-80% size reduction).

The 2-bit XL GGUF performs exceptionally well on coding & passes all our code tests

Guide: https://t.co/kRig50dJHR
GGUFs: https://t.co/GFO4kgxQ2q

[Google AI Developers @googleaidevs]

Introducing Gemini CLI, a light and powerful open-source AI agent that brings Gemini directly into your terminal. >_

Write code, debug, and automate tasks with Gemini 2.5 Pro with industry-leading high usage limits at no cost.

[Adam Silverman (Hiring!) 🖇️ @adamsilverman]

You do not want to miss the MCP updates & announcements from Windows, Gradio, Caucus, Anthropic, Firecrawl, Postman, Perplexity, Cline, and more this week! 

I summarized everything you need to know below. 

(save for later)

[Google Gemini @GeminiApp]

We’re also launching Veo 3, our state-of-the-art video generation model.

Veo 3 lets you generate videos with sound effects, background noises and even dialogue.

#GoogleIO

[Adam Silverman (Hiring!) 🖇️ @adamsilverman]

A week in AI Agents is like a year in traditional software. 

Here is everything that happened this week in AI Agents from OpenAI, AgentOps, LangChain, You, Flowwise, Convergence, CopilotKit, Aomni, Box, Stagehand, Cohere, & more. 🧵 

(save for later)

[Nari @0xnari]

I tried Qwen3:4b. By default, it generates responses with reasoning enclosed in <think> tags.
If you want to disable that, you’re supposed to add /no_think at the end of the prompt,
but in many cases, it didn’t seem to work properly.

[Cognition @cognition]

Project DeepWiki

Up-to-date documentation you can talk to, for every repo in the world.

Think Deep Research for GitHub – powered by Devin.

It’s free for open-source, no sign-up!
Visit deepwiki com or just swap github → deepwiki on any repo URL:

[Nari @0xnari]

DeepWiki is incredibly useful.
Not only does it turn a target GitHub repository into easy-to-understand documentation, but it also offers an AI chat feature.
This AI chat is smart and has been extremely helpful.

[Marcin Ignac @marcinignac]

Every 20 seconds, a new Spirit Plant blooms—composed of 4 traditional Polish herbs and an ever-shifting palette of colours.
In just the first 3 days of Expo 2025, visitors have co-created thousands of unique digital plants at the Poland Pavilion. The result? An endlessly evolving garden of collective imagination.

#Expo2025 #GenerativeDesign #PolandAtExpo

[Anthropic @AnthropicAI]

A few researchers at Anthropic have, over the past year, had a part-time obsession with a peculiar problem.

Can Claude play Pokémon?

A thread:

[DeepSeek @deepseek_ai]

🚀 Introducing NSA: A Hardware-Aligned and Natively Trainable Sparse Attention mechanism for ultra-fast long-context training & inference!

Core components of NSA:
• Dynamic hierarchical sparse strategy
• Coarse-grained token compression
• Fine-grained token selection

💡 With optimized design for modern hardware, NSA speeds up inference while reducing pre-training costs—without compromising performance. It matches or outperforms Full Attention models on general benchmarks, long-context tasks, and instruction-based reasoning.

📖 For more details, check out our paper here: https://t.co/HJiqzwnUV7

[ollama @ollama]

DeepSeek's first-generation reasoning models are achieving performance comparable to OpenAI's o1 across math, code, and reasoning tasks!

Give it a try! 👇

7B distilled:
ollama run deepseek-r1:7b

More distilled sizes are available. 🧵

[griffain @griffaindotcom]

https://t.co/Q3YVYhI5yx

[el.cine @EHuanglu]

what??

NVIDIA just dropped Project DIGITS, a $3,000 personal AI supercomputer that’s small enough to look like a Mac Mini but packs 1,000x the power of your average laptop.

Handles AI models with up to 200 BILLION parameters.

This is incredible..

[Nari @0xnari]

I think the combination of AI agents and memecoins would be a good match for experimenting with some ambitious projects.

[Nari @0xnari]

Thanks!
$PENGU airdrop

[Elon Musk @elonmusk]

侘び寂び

[Alex Cheema @alexocheema]

M4 Mac Mini AI Cluster

Uses @exolabs with Thunderbolt 5 interconnect (80Gbps) to run LLMs distributed across 4 M4 Pro Mac Minis.

The cluster is small (iPhone for reference). It’s running Nemotron 70B at 8 tok/sec and scales to Llama 405B (benchmarks soon).

[Nari @0xnari]

@stand_english @metaplex おめでとうございます🎉

[Awni Hannun @awnihannun]

MLX Swift example can also QLoRA fine-tune Llama 3.2.

Here's the 1B fine-tuning on my iPhone 15 Pro at > 150 toks/sec. A this rate only takes a few minutes to learn some decent adapters fully on-device.