PinnedInTowards AIbyMontasir Mahmud·Aug 28[Day 8/100] Agent Architectures Compared: Single, Multi-Agent, Hierarchical, SwarmThe question every agent project hits in week two: should we have one agent or many?A response icon4A response icon4
PinnedMontasir Mahmud·May 3100 Days of Learning Agentic AI: Full Topic RoadmapWelcome. If you are reading this, you want to become a serious agentic AI developer. Not someone who can wire up a demo in a weekend, but…A response icon1A response icon1
InTowards AIbyMontasir Mahmud·1d ago[Day 45/100] Hybrid Search for RAG: BM25 Finds What Your Vectors MissWhy embeddings blur error codes and endpoint names, how BM25 and Reciprocal Rank Fusion fix it in about 60 lines of Python, and when to let…A response icon2A response icon2
InTowards AIbyMontasir Mahmud·1d ago[Day 44/100] RAG Chunking Strategies: Why Half Your Answer Lands in the Wrong ChunkFixed, recursive, structure-aware, semantic, and hierarchical chunking, plus Anthropic’s contextual retrieval, tested on one PTO policy…A response icon1A response icon1
Montasir Mahmud·2d agoAI Could Pay Us All. History Says It Won’t Happen by Itself.In 1812, machines made Britain rich and left its workers waiting half a century for a share. AI is about to run the same experiment on the…
InTowards AIbyMontasir Mahmud·3d ago[Day 43/100] Vector Database Comparison for RAG: pgvector, Qdrant, Pinecone, and MoreSix databases compared on where they run, how they filter, hybrid search, and tenant isolation, then a tested move from LangChain’s…A response icon2A response icon2
InTowards AIbyMontasir Mahmud·6d ago[Day 42/100] How to Choose an Embedding Model for RAG (Without Trusting the Leaderboard)Dimensions, Matryoshka truncation, token limits, query prefixes, and cost, plus a small recall@k harness that tests any LangChain embedding…A response icon2A response icon2
Montasir Mahmud·Nov 8Weight-Space Programming: One LLM Can Directly Edit Another LLM’s WeightsGPT-6 Astra directly read and modified Qwen 0.8B’s weights, changing just 0.016% of them and improving its math scores by 509%.
InTowards AIbyMontasir Mahmud·Nov 7[Day 41/100] RAG Fundamentals: Why Your LLM Needs Retrieval, Not a Bigger Context WindowLong context and fine-tuning both look like shortcuts. Why retrieval wins for company documents, how to build a LangChain RAG baseline that…A response icon1A response icon1
InTowards AIbyMontasir Mahmud·Nov 4[Day 40/100] AI Agent Memory Privacy: Isolate Users, Redact PII, Delete for RealOne tool argument can leak every user’s memories. How to scope identity, enforce row-level security in Postgres, keep PII out of storage…