Welcome back to Daily AI Skills.

Here’s what we are covering today:
1. The AI Biology of Claude
2. Gemini 2.5 Experimental - Best Coding Model Free for everyone?
3. Alibaba Qwen 2.5 - Omni 7B

+ App Flow Document Template for Cursor & Windsurf

The Biology of a Large Language Model

Anthropic has published two research papers shedding light on how its AI assistant, Claude, processes information. These studies offer insights into the internal mechanisms behind capabilities such as multilingual reasoning and advanced planning.

Key findings:

• Researchers developed an "AI microscope" that maps Claude’s internal "circuits," revealing how it transforms input into output for critical tasks.

• Claude operates using a universal "language of thought" across multiple languages, processing concepts similarly in English, French, and Chinese.

• When composing poetry, Claude anticipates several words ahead, selecting rhyming options before constructing lines to align with those choices.

• The team also identified a built-in tendency to avoid speculation unless backed by strong confidence, offering insights into how the model mitigates hallucinations.

Read the research papers by Anthropic here

Gemini 2.5 - The Best Model for Coding is now free for everyone

Google has made Gemini 2.5 available to all free Gemini users, a move likely to increase competitive pressure in the AI space. Initially launched a few days ago, Gemini 2.5 Pro Experimental is the company’s most advanced model yet, featuring built-in reasoning capabilities.

Key Highlights:

• #1 on LMArena Leaderboard – Demonstrates state-of-the-art reasoning across math, science, and coding.

• Strong coding performance – Scores 63.8% on SWE-Bench Verified and 68.6% on Aider Polyglot, excelling in web apps and agentic code applications.

• 1M token context window – Soon expanding to 2M tokens for processing entire code repositories and massive datasets.

• New Canvas feature – Allows users to run code live directly within the interface.

• Now free for all users – Expected to intensify competition, with API pricing coming soon.

Try out the model here: https://gemini.google.com/app

Alibaba Unveils Qwen2.5-Omni-7B

Alibaba has introduced Qwen2.5-Omni-7B, a new multimodal AI model that can process text, images, audio, and video simultaneously while being efficient enough to run on consumer devices like smartphones and laptops.

Key Highlights:

• "Thinker-Talker" system – Enables real-time processing across multiple modalities with both text and speech outputs.

• Superior speech capabilities – Outperforms specialized audio models in speech understanding and generation benchmarks.

• Optimized for consumer devices – Runs efficiently on phones and laptops, making it ideal for real-world applications like real-time audio descriptions for visually impaired users.

• Available now – Released on Hugging Face and GitHub, with Alibaba positioning it as a foundation for practical AI agents.

Read the full announcement by Alibaba here

App Flow Document Template for Cursor and Windsurf

📩Forward it to people you know who are keeping pace with the changing AI world, and stay tuned for the next edition to stay ahead of the curve!

Maybe you missed out on these:

1. Copilot vs Claude vs Gemini

2. How I Built an AI Influencer in 2 Minutes—and Landed Sponsorships

3. Microsoft Muse, Helix and Google's AI Co-Scientist

4. Bolt vs Lovable

5. Your AI Stack Just Got Superpowers: 4 New Inference Providers Live!