The best AI learning resources in the world are already free

 

Posting here is like having an archive where you store good information and it's also seem by other people... which they find interesting 💭 🤔 

These tool's are heavy duties, what I mean is they are good at what they are being programmed for.

 

10 platforms worth bookmarking:

 

1. Anthropic

👉anthropic.skilljar.com

 

2. Google

👉grow.google/ai

 

3. Meta

👉ai.meta.com/resources/

 

4. NVIDIA

👉developer.nvidia.com/cuda

 

5. Microsoft

👉learn.microsoft.com/en-us/training/

 

6. OpenAI

👉academy.openai.com

 

7. IBM

👉skillsbuild.org

 

8. AWS

👉skillbuilder.aws

 

9. DeepLearning

👉deeplearning.ai

 

10. Hugging Face

👉huggingface.co/learn

 

 

 

Anthropic just open-sourced a system that makes Claude dramatically better at specialized tasks.

 

It's called Agent Skills. And it's already powering Claude's built-in document features behind the scenes.

 

Skills are folders of instructions that Claude loads dynamically to improve performance on specific tasks.

 

One folder. One SKILL.md file. Claude reads it and follows it, every single time.

 

Here is why this matters right now:

 

Claude out of the box is general. But your work isn't general.

 

You have specific workflows, brand guidelines, document formats, and pipelines.

 

Skills let you encode all of that and Claude will follow it consistently, not just when it feels like it.

 

What's already inside the repo:

 

-> Word document creation (with formatting, tables, TOC)

-> PDF reading and generation

-> PowerPoint deck building

-> Excel spreadsheet workflows

-> Web app testing

-> MCP server generation

-> Creative, design, and enterprise comms skills

 

How to use it right now:

 

-> Claude.ai (paid plans) — skills are already available, upload custom ones

-> Claude Code — install via /plugin marketplace add anthropics/skills

-> Claude API — plug skills in via the Skills API

 

Creating your own skill takes 5 minutes:

 

Make a folder. Add a SKILL.md file. Write YAML frontmatter with a name and description. Add your instructions in Markdown below.

 

That's it. Claude will load it and follow your workflow exactly.

 

Real use cases:

 

-> Encode your company's doc formatting rules so Claude never gets them wrong

-> Build a client deliverable skill that always matches your brand tone

-> Create a data analysis skill that follows your exact reporting structure

-> Automate document pipelines with zero prompt re-explaining every session

 

From Anthropic. Actively maintained. Already 112k stars on GitHub.

 

100% Open Source. Apache 2.0 License.

 

 

 

🚨 BREAKING: Someone at Microsoft just open-sourced a tool that converts almost any file format into clean Markdown.

 

It's called MarkItDown. And it handles everything.

 

PDFs. Word docs. PowerPoints. Excel spreadsheets. Images. Audio files. HTML. ZIP archives.

 

One tool. One output format. Clean Markdown every time.

 

Here is why this matters right now:

 

Every AI coding agent, every LLM, every RAG pipeline works better with Markdown than with raw file formats. PDFs confuse them. DOCX files bloat the context. PowerPoints are basically unreadable.

 

MarkItDown fixes the input layer.

 

Drop any file in. Get clean, structured Markdown out. Feed it directly to your AI agent.

 

Full format support:

 

-> PDF documents

-> Word files (DOCX)

-> PowerPoint presentations (PPTX)

-> Excel spreadsheets (XLSX)

-> Images with EXIF metadata and OCR

-> Audio files with speech transcription

-> HTML pages

-> ZIP archives (processes contents recursively)

 

Use cases that are immediately obvious:

 

-> Feed a PDF research paper to Claude or GPT without fighting the format

-> Convert a client PowerPoint into structured notes in seconds

-> Preprocess entire document folders for RAG pipelines

-> Build document ingestion pipelines with zero formatting headaches

 

Install in one line:

 

pip install markitdown

 

That is it. Works as a Python library or a CLI tool.

 

From Microsoft. Actively maintained. Already one of the most starred Python tools on GitHub.

 

100% Open Source. MIT License.