I discovered how assisted generation speeds up LLMs using smaller models, explored AI-powered Advanced Paste in PowerToys, and investigated Turing complete systems like Magic: The Gathering. I also checked Rev.ai for cost-effective audio diarization.
I discovered LanceDB as a scalable ChromaDB alternative and explored Meta's image models. I also tested Gemini's code execution, analyzed audio diarization limitations, and compared AI development tools like Cursor, Bolt, and Replit for different skill levels.
I learned about JavaScript's Temporal object, appending hidden data to PDFs, and using embeddings for ML classification. I also explored AI-driven business models, including zero-employee companies, and how compute dominance shapes global AI power.
I investigated Amazon Nova model costs, surveyed why employees avoid internal LLMs, and evaluated JavaScript text splitters. I also learned about Unicode characters in ChatGPT citations and how to install Docker on Windows without admin rights.
I challenged several LLMs to generate funny Tamil puns ending in ".ai". While DeepSeek and Claude struggled with accuracy, Gemini captured the cultural nuance perfectly with clever wordplay like Tholl.ai (annoyance) and Kaval.ai (worry).
I explored using Pyodide for browser DOM access and PyMuPDF4LLM for converting PDFs to Markdown. I also compared AVIF to GIF compression, experimented with Opus audio encoding, and researched Anthropic’s contextual retrieval methods for improved RAG performance.