I've indexed all Strange Loop conference talks so you can do semantic search (devblogs.sh)

Someone has indexed the complete Strange Loop conference archive into a semantic-searchable collection, letting you query talks by idea instead of exact titles or tags. That means developers and researchers can now find cross-cutting technical content—from ML safety and interpretability to language design, generative systems, and systems engineering—by meaning (via embeddings/vector search) rather than keyword lookup. For busy practitioners this speeds discovery of relevant talks, enables building retrieval-augmented tools (QA, summarizers, trend analysis), and turns a rich, multi-disciplinary conference into a machine-actionable knowledge base. The indexed talks include technically deep presentations valuable to the AI/ML community: Zac Hatfield‑Dodds on neural network “strange loops” (induction heads, feature superposition, RLHF feedback loops); Rabii’s Kolmogorov‑style theorem tying generator source length to artifact complexity; Cursorless (spoken programming) with tree‑sitter parse-tree scopes and LLM integration; navu’s fine‑grained CPU-histogram driven adaptive concurrency for JVM/Kubernetes; Wizard, a compact WebAssembly engine designed for observability and JIT experiments; and applied ML/engineering talks on scaling experiments, formal multi-language semantics, visualization-to-code compilers, and more. The index supports fine-grained retrieval for building RAG pipelines, dataset curation, citation mining, and cross-domain research synthesis—lowering the friction to reuse conference insights in code, papers, and products.