Discovery
Best searchable video moments
Curated, transcript-backed clips from the public index. Each card opens a canonical moment page with timestamp, excerpt, and YouTube link — no rehosting.
559 moments · Updated when the materialized list is regenerated.
AI podcasts
Browse ai podcasts category →- "hate speech what is dangerous and all that kind of stuff I mean you've different value systems"Yann Lecun: Meta AI, Open Source, Limits of LLMs, AGI & the Future of AI | Lex Fridman Podcast #416 · 1:55:09
- Topic may be polarized; excerpt is one speaker’s framing.
- "life and the Deep question with AI is also what is life and what is"Andrej Karpathy: Tesla AI, Self-Driving, Optimus, Aliens, and AGI | Lex Fridman Podcast #333 · 2:51:08
- Short or messy excerpt — ranking is uncertain.
- "a convolutional embedding"An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale (Paper Explained) · 16:14
- Sounds like subjective take — verify claims independently.
- Short or messy excerpt — ranking is uncertain.
- "rmers aren't"An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale (Paper Explained) · 25:11
- Short or messy excerpt — ranking is uncertain.
- "l right with"An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale (Paper Explained) · 29:48
- Short or messy excerpt — ranking is uncertain.
- "training"Deep Dive into LLMs like ChatGPT · 1:04
- Short or messy excerpt — ranking is uncertain.
- "training"YouTube video 4b4MUYve_U8 · 3:04
- Short or messy excerpt — ranking is uncertain.
- "training"[GRPO Explained] DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models · 3:57
- Short or messy excerpt — ranking is uncertain.
- "training"[Classic] Generative Adversarial Networks (Paper Explained) · 2:26
- Short or messy excerpt — ranking is uncertain.
- "certainly"Michael Nielsen – Why aliens will have a different tech stack than us · 1:21
- Sounds like subjective take — verify claims independently.
- "inference"How GPT, Claude, and Gemini are actually trained and served – Reiner Pope · 0:25
- Sounds like subjective take — verify claims independently.
- "training"Mark Zuckerberg — Llama 3, $10B models, Caesar Augustus, & 1 GW datacenters · 2:50
- Sounds like subjective take — verify claims independently.
- Short or messy excerpt — ranking is uncertain.
- "inference"[Paper Analysis] The Free Transformer (and some Variational Autoencoder stuff) · 21:26
- Short or messy excerpt — ranking is uncertain.
- "inference"Energy-Based Transformers are Scalable Learners and Thinkers (Paper Review) · 1:55
- Short or messy excerpt — ranking is uncertain.
- "training"Variational Autoencoders · 1:44
- Short or messy excerpt — ranking is uncertain.
- "training"AI Scaling, Alignment, and the Path to Superintelligence — With Dwarkesh Patel · 2:14
- Short or messy excerpt — ranking is uncertain.
- "gemini-3"Ilya Sutskever – We're moving from the age of scaling to the age of research · 33:41
- Short or messy excerpt — ranking is uncertain.
- "basically"Mark Zuckerberg — Llama 3, $10B models, Caesar Augustus, & 1 GW datacenters · 1:48
- Sounds like subjective take — verify claims independently.
- Short or messy excerpt — ranking is uncertain.
- "actually"Mark Zuckerberg — Llama 3, $10B models, Caesar Augustus, & 1 GW datacenters · 2:00
- Sounds like subjective take — verify claims independently.
- Short or messy excerpt — ranking is uncertain.
- "probably"Andrej Karpathy — “We’re summoning ghosts, not building animals” · 13:55
- Sounds like subjective take — verify claims independently.
- Short or messy excerpt — ranking is uncertain.
- "learning"Adam Marblestone – AI is missing something fundamental about the brain · 0:45
- Sounds like subjective take — verify claims independently.
- Short or messy excerpt — ranking is uncertain.
ai-research
- "RAG failure modes cause hallucinations missing data chunking embeddings"Webinar: Fix Hallucinations in RAG Systems with Pinecone and Galileo · 19:48
- Sounds like subjective take — verify claims independently.
- Topic may be polarized; excerpt is one speaker’s framing.
- Snippet match: phrase_not_reflected_in_snippet
- Short or messy excerpt — ranking is uncertain.
chunking
hybrid-search
- "I spell a word wrong"What is Vector Search? | Vector Databases with Weaviate: Part 2 | Community Webinar · 46:41
- Topic may be polarized; excerpt is one speaker’s framing.
mcp-orchestration
Programming tutorials
Browse programming tutorials category →- "database"Learn Flask for Python - Full Tutorial · 15:26
- Sounds like subjective take — verify claims independently.
- "shortest"Algorithms Course - Graph Theory Tutorial from a Google Engineer · 6:42
- Sounds like subjective take — verify claims independently.
- "elements"Intermediate Python Programming Course · 0:51
- Sounds like subjective take — verify claims independently.
- "regression"Machine Learning for Everybody – Full Course · 18:53
- Sounds like subjective take — verify claims independently.
- Short or messy excerpt — ranking is uncertain.
- "actually"Machine Learning for Everybody – Full Course · 1:28
- Sounds like subjective take — verify claims independently.
- Short or messy excerpt — ranking is uncertain.
RAG pipeline
reranking
retrieval-evaluation
retrieval-infra
More indexed moments
- "our what what is our error with respect to this line of best fit"Machine Learning for Everybody – Full Course · 2:32:12
- Short or messy excerpt — ranking is uncertain.
- "ASI, Artificial Superintelligence, and what are the language models that we have today capable of doing"State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI | Lex Fridman Podcast #490 · 2:39:10
- Sounds like subjective take — verify claims independently.
- "our worldview"Michael Levin: Hidden Reality of Alien Intelligence & Biological Life | Lex Fridman Podcast #486 · 3:01:30
- Topic may be polarized; excerpt is one speaker’s framing.
- "our worldview specifically not able to see, right"Michael Levin: Hidden Reality of Alien Intelligence & Biological Life | Lex Fridman Podcast #486 · 3:01:34
- Topic may be polarized; excerpt is one speaker’s framing.
- "different ideas for the mechanism of how to add memory to these systems as we're increasingly seeing"State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI | Lex Fridman Podcast #490 · 2:42:42
- Sounds like subjective take — verify claims independently.
- "different ideas for the mechanism of how to add memory to these systems as we're"State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI | Lex Fridman Podcast #490 · 2:42:34
- Sounds like subjective take — verify claims independently.
- "fixes for all these issues"LangChain "RAG Evaluation" Webinar · 16:49
- Topic may be polarized; excerpt is one speaker’s framing.
- "Qdrant query API: prefetch dense+sparse with fusion=rrf; benchmark ranks reports Precision@10 and MRR@10 vs dense-only b"Qdrant hybrid search — RRF prefetch + Precision@10/MRR · 41:00
- Snippet match: snippet_under_88_chars
- Short or messy excerpt — ranking is uncertain.
- "Phoenix trace: retrieval step shows retrieved chunks; chunk relevance eval scores precision/ndcg; faithfulness gate on l"Arize Phoenix — retrieve span + chunk relevance eval · 42:00
- Snippet match: snippet_under_88_chars
- Short or messy excerpt — ranking is uncertain.
- "LangSmith @traceable retriever run: retrieve span trace shows document relevance scores; faithfulness=0.72 on labeled ro"LangSmith — document relevance eval on retrieve spans · 5:00
- Snippet match: snippet_under_88_chars
- Short or messy excerpt — ranking is uncertain.
- "Langfuse trace tree: retrieval span per query with inputs/outputs; production observability on retrieve+generation; debu"Langfuse — multi-step RAG trace + retrieval observability · 20:00
- Snippet match: snippet_under_88_chars
- Short or messy excerpt — ranking is uncertain.
Looking for more? Try the trending hub, full transcript index, research collections, or a seeded search landing.
Request a new source for indexing
Heuristic intake only — no account. Paste a public YouTube video, channel, playlist, or podcast URL you want searchable in a future crawl.