RecallRunHands-on guides to production RAG, AI agents, MCP and evals, with code that is actually run and numbers that are actually measured.https://recallrun.dev/JEPA vs LLMs: what LeCun's world models actually change for engineershttps://recallrun.dev/posts/jepa-vs-llm-world-models/https://recallrun.dev/posts/jepa-vs-llm-world-models/Analysis: JEPA predicts embeddings, not tokens. My text-chunk stand-in: token-space prediction collapsed 99.4% with 10x less data, embedding-space only 45.8%.Mon, 28 Sep 2026 00:00:00 GMTDo you need a vector database? Brute-force search benchmarked from 10k to 1M vectorshttps://recallrun.dev/posts/do-you-need-a-vector-database/https://recallrun.dev/posts/do-you-need-a-vector-database/Exact numpy search over 100k embeddings takes 3.5 ms on 2 vCPUs. Measured latency, memory and HNSW recall to show when an index actually pays off.Sun, 27 Sep 2026 00:00:00 GMTHybrid search in Python: BM25 + vectors with Reciprocal Rank Fusion, alpha swept 0 to 1https://recallrun.dev/posts/hybrid-search-bm25-rrf-python/https://recallrun.dev/posts/hybrid-search-bm25-rrf-python/Tested BM25+RRF on 2,713 FastAPI doc chunks: BM25 beat vectors by 49 MRR points on identifiers; vectors beat BM25 by 24 on paraphrases. alpha=0.8 balanced both.Sun, 27 Sep 2026 00:00:00 GMTJev vs LLMs: when a decision model beats a chat model (and when it doesn't)https://recallrun.dev/posts/jev-vs-llm-decision-models/https://recallrun.dev/posts/jev-vs-llm-decision-models/Analysis: Jev claims 70-500ms typed decisions (vendor-reported). My runnable stand-in, TF-IDF + logistic regression, routed text at 0.6ms and $0, 60% accuracy.Sun, 27 Sep 2026 00:00:00 GMTRAG chunking, measured: heading-aware chunks doubled hit@1 on the FastAPI docshttps://recallrun.dev/posts/rag-chunking-strategies-measured/https://recallrun.dev/posts/rag-chunking-strategies-measured/I tested 5 chunking strategies on 151 real docs and 773 queries. Fixed-size chunks crossed section boundaries 59% of the time. Heading-aware chunks never did.Sun, 27 Sep 2026 00:00:00 GMTThe RAG production checklist: 25 checks before real users see ithttps://recallrun.dev/posts/rag-production-checklist/https://recallrun.dev/posts/rag-production-checklist/25 checks across ingestion, retrieval, generation, evaluation and operations that catch the RAG failures demos hide. Free printable PDF included.Sun, 27 Sep 2026 00:00:00 GMTStop fake citations in RAG: validate them server-side in 25 lines of Pythonhttps://recallrun.dev/posts/validate-llm-citations-fastapi/https://recallrun.dev/posts/validate-llm-citations-fastapi/LLMs will cite passage [7] when you gave them 5. A tested validator that strips invented citations and returns a grounded flag your UI can trust.Sun, 27 Sep 2026 00:00:00 GMT