Agentic Search
Synopsis
Agentic search is retrieval run as an active investigation rather than a single lookup. In the World's Fair material, it shows up as the move away from "vector search once and stuff it into context" toward systems that plan queries, combine retrieval modes, inspect source quality, follow missing evidence, and keep a decision trace of what was searched and why.
Origin And Context
The connected sessions frame agentic search as a convergence of RAG, hybrid retrieval, semantic code search, SQL, UI telemetry, long-context caching, knowledge graphs, web data, and agent evaluation. Turbopuffer's RAG talk argues that retrieval is broader than vector search; Ogilvy's hybrid RAG session adds SQL RRF and UI telemetry; Tesco and Turbopuffer connect the pattern to local code indexes and semantic code retrieval; Neo4j emphasizes decision traces over document dumps; Bright Data focuses on whether an agent actually searched the web in a verifiable way.
Why It Matters
The linked talks repeatedly point to the same failure mode: agents sound confident while retrieving the wrong evidence, ignoring user signals, overloading the context window, or claiming to have searched without a durable search trail. Agentic search matters here because it turns retrieval into an observable workflow: query choices, source triage, reranking, citations, gaps, conflicts, and stopping criteria can be evaluated instead of hidden inside one prompt.
How To Use It
Treat search as a loop with state. Start from the user question and available signals, retrieve across the right surfaces, rerank by task relevance, inspect primary or authoritative sources, record claims with citations, and branch when evidence conflicts. Use BM25 or exact search for identifiers and rare terms, vector search for semantic recall, SQL for structured facts, knowledge graphs for relationships, UI telemetry for product context, and cache or long-context strategies only when they improve evidence quality rather than merely adding bulk.
Where It Is Useful
Agentic search is useful anywhere an answer depends on evidence scattered across multiple systems: research agents comparing sources, enterprise assistants over internal knowledge bases, coding agents searching repositories, compliance workflows correlating documents, product agents turning user signals into pull requests, and support systems that need to explain which logs, tickets, docs, or web pages informed the response.
When To Use It
Use agentic search when the question is underspecified, evidence is distributed, freshness matters, exact citations are required, or the agent must reconcile conflicting sources. A direct database query, deterministic API call, or simple keyword lookup is still better when the answer lives in one known structured source and the task does not require exploration.
Active Use Cases
- Research agents that cite and compare sources.
- Hybrid RAG over documents, SQL, UI telemetry, and web data.
- Semantic code retrieval for coding agents.
- Enterprise knowledge agents with source-grounded answers.
Related Slide Decks
- youtube aHhB3sjGjkI slides — Agents Building Agents - Alfonso Graziano, Nearform (24 extracted slide frames)
- youtube jVjt 2g8NMY slides — A Genius With Amnesia - Victor Savkin, Nx (19 extracted slide frames)
Related Scheduled Sessions
- 2026 07 01 session vector isn t enough hybrid search and retrieval for ai engineers — Vector Isn't Enough: Hybrid Search & Retrieval for AI Engineers; Jeff Vestal (Day 1 — Workshop Day · 2:20pm-4:20pm · Track 7; official schedule)
- 2026 06 29 jo kristian bergum the unreasonable effectiveness of bm25 for agentic search — The unreasonable effectiveness of BM25 for agentic search; Jo Kristian Bergum (Day 2 — Session Day 1 · 11:10am-11:30am · Search & Retrieval; official schedule)
- 2026 06 29 will bryk the search engine for the agentic web — The Search Engine for the Agentic Web; Will Bryk (Day 2 — Session Day 1 · 11:40am-12:00pm · Search & Retrieval; official schedule)
- 2026 06 29 maximilian david rumpf where rl will take search — Where RL Will Take Search; Maximilian-David Rumpf, Lotte Seifert (Day 2 — Session Day 1 · 2:50pm-3:10pm · Search & Retrieval; official schedule)
- 2026 06 30 han xiao autoresearch for dense retrieval test time compute with frozen embedding models — Autoresearch for Dense Retrieval: Test-Time Compute with Frozen Embedding Models; Han Xiao (Day 3 — Session Day 2 · 11:10am-11:30am · Autoresearch; official schedule)
- 2026 06 30 elie bakouch the era of auto research — « the era of (auto) research »; Elie Bakouch (Day 3 — Session Day 2 · 12:05pm-12:25pm · Autoresearch; official schedule)
- 2026 06 30 tim sweeney closing the loop an autonomous ai research agent — Closing the Loop: An Autonomous AI Research Agent; Tim Sweeney (Day 3 — Session Day 2 · 1:30pm-1:50pm · Autoresearch; official schedule)
- 2026 06 29 dhruv nathawani teaching agents to search building synthetic training pipelines with nvidia data designer — Teaching Agents to Search: Building Synthetic Training Pipelines with NVIDIA Data Designer; Dhruv Nathawani (Day 1 — Workshop Day · 11:05am-12:05pm · Workshops Day 1; official schedule)
- 2026 06 30 erina karati autoresearch in a multi agent ai village — Autoresearch in a Multi-Agent AI Village; Erina Karati, Arunachalam Manikandan (Day 3 — Session Day 2 · 3:45pm-4:05pm · Autoresearch; official schedule)
- 2026 07 01 stephen chin crabrag why automated assistants need graph memory not more tokens — CrabRAG: Why Automated Assistants Need Graph Memory, Not More Tokens; Stephen Chin (Day 4 — Session Day 3 · 10:45am-11:05am · Graphs; official schedule)
- 2026 06 29 zhengyao jiang hands on autoresearch cracking openai s parameter golf — Hands-on AutoResearch: Cracking OpenAI's Parameter Golf; Zhengyao Jiang, Dixing Xu, Vayum Arora, Dhruv Srikanth (Day 1 — Workshop Day · 2:20pm-4:20pm · Workshops Day 1; official schedule)
- 2026 06 30 benoit schillings research to reality with google deepmind — Research to Reality with Google DeepMind; Benoit Schillings (Day 3 — Session Day 2 · 10:05am-10:25am · Autoresearch; official schedule)
- 2026 06 30 richard socher first steps toward automated ai research — First Steps Toward Automated AI Research; Richard Socher (Day 3 — Session Day 2 · 10:45am-11:05am · Autoresearch; official schedule)
- 2026 06 30 tejas bhakta autoresearch for kernels — Autoresearch for Kernels; Tejas Bhakta (Day 3 — Session Day 2 · 2:50pm-3:10pm · Autoresearch; official schedule)
- 2026 06 30 roland gavrilescu autoresearch in the wild — Autoresearch in the wild; Roland Gavrilescu, Julian Bright (Day 3 — Session Day 2 · 3:20pm-3:40pm · Autoresearch; official schedule)
- 2026 07 01 brendan rappazzo alphalab autonomous multi agent research across optimization domains with frontier llms — ALPHALAB: Autonomous Multi-Agent Research Across Optimization Domains with Frontier LLMs; Brendan Rappazzo (Day 4 — Session Day 3 · 10:45am-11:05am · AI in Finance; official schedule)
- 2026 06 29 nyah macklin rag needs a map using graphrag to retrieve connected context — RAG Needs a Map: Using GraphRAG to Retrieve Connected Context; Nyah Macklin (Day 1 — Workshop Day · 11:05am-12:05pm · Track 2; official schedule)
- 2026 06 29 jess wang agentic vs vector search an eval driven approach to coding agent performance — Agentic vs. Vector Search: An Eval-Driven Approach to Coding Agent Performance; Jess Wang (Day 2 — Session Day 1 · 11:40am-12:00pm · Expo Stage 2 NW; official schedule)
- 2026 06 30 nixon dinh the death of keyword search and the rise of agent readable catalogs — The Death of Keyword Search and the Rise of Agent-Readable Catalogs; Nixon Dinh (Day 3 — Session Day 2 · 11:10am-11:30am · Expo Stage 3; official schedule)
- 2026 07 01 george he everyone talks about document search but what about results — Everyone talks about document search, but what about results?; George He (Day 4 — Session Day 3 · 1:55pm-2:15pm · Expo Stage 4 SE; official schedule)
- 2026 06 30 stefania druga memory harnesses for long running research agents — Memory Harnesses for Long-Running Research Agents; Stefania Druga (Day 3 — Session Day 2 · 11:40am-12:00pm · Memory & Continual Learning; official schedule)
- 2026 07 01 zubin aysola aria how we built autoresearch with autoresearch — ARIA, how we built autoresearch with autoresearch; Zubin Aysola (Day 4 — Session Day 3 · 11:10am-11:30am · Expo Stage 2 NW; official schedule)
- 2026 06 29 peter werry beyond rag build a relational context engine from scratch — Beyond RAG: Build a Relational Context Engine from Scratch; Peter Werry (Day 1 — Workshop Day · 12:10pm-1:10pm · Workshops Day 1; official schedule)
- 2026 06 29 valeria wu fon speech to speech model research at google deepmind — Speech-to-Speech Model Research at Google DeepMind; Valeria Wu Fon, Tom Ouyang (Day 2 — Session Day 1 · 11:10am-11:30am · Voice & Realtime AI; official schedule)
Related People
- Zhengyao Jiang
- Charlie Guo
- Brandon Waselnuk
- Christopher Manning
- Kent C. Dodds
- Abhishek Bhardwaj
- Jeff Vestal
- Jo Kristian Bergum
- Will Bryk
- Maximilian-David Rumpf
- Lotte Seifert
- Han Xiao
- Elie Bakouch
- Tim Sweeney
- Dhruv Nathawani
- Erina Karati
- Arunachalam Manikandan
- Stephen Chin
- Dixing Xu
- Vayum Arora
- Dhruv Srikanth
- Benoit Schillings
- Richard Socher
- Tejas Bhakta
Related Companies
- NVIDIA
- Weco AI
- Google DeepMind
- OpenAI
- Neo4j
- Bright Data
- Unblocked
- Oracle
- Amazon AGI Lab
- Elastic
- Weights & Biases by CoreWeave
- Introspection
- Exa
- turbopuffer
- LlamaIndex
- Artificial Analysis
- Prime Intellect
- DatologyAI
Transcript And Resource Support
Transcript-backed resources
- youtube UM6sFg_jdlE — RAG is dead, right?? — Kuba Rogut, Turbopuffer
- youtube Akm1sqvWG4A — Bypassing the Multimodal Tax: Hybrid RAG, SQL RRF & UI Telemetry - Abed Matini, Ogilvy
- youtube zKk7sDMGDEQ — Benchmarking semantic code retrieval on Claude Code — Kuba Rogut, Turbopuffer
- youtube T0HhO4YtTfE — AI System Design: From Idea to Production - Apoorva Joshi, MongoDB
- youtube OXMMN XbxwA — Research to Reality: Bringing Frontier ML Research to Production - Vaidas Razgaitis, Higharc
- youtube htM02KMNZnk — WF2026: Software Factories & Keynotes ft. Microsoft, OpenAI, OpenClaw, Z.ai (GLM), MiniMax, HF
- youtube wFTVEDYVJT0 — Building Agents with Amazon Nova Act and MCP - Du'An Lightfoot, Amazon (Full Workshop)
- youtube x5GEVnkuRw — Structuring the Unstructured - Cedric Clyburn, Red Hat
- youtube vh2VGuQ3zhY — The 100-Tool Agent Is a Trap - Sohail Shaikh & Ankush Rastogi, Prosodica
- youtube Jx4ZFEAq6bY — User Signal Dies at the Retrieval Boundary - Sonam Pankaj, StarlightSearch
- youtube dRmWYHuIJxM — We Cut 94% of AI Coding Tokens With a Local Code Index - Rajkumar Sakthivel, Tesco
- youtube XovaGv4f39A — When All Context Matters: Extended Cache Augmented Generation - Luis Romero-Sevilla, Orbis
- youtube btxGmN8RvNU — Your Agent's Biggest Lie: "I Searched the Web" — Rafael Levi, Bright Data
- youtube iNkFlCiij0U — The Art & Science of Benchmarking Agents — Vincent Chen, Snorkel AI
- youtube EcqMYoIV57A — Why More Context Makes Your Agent Dumber and What to Do About It — Nupur Sharma, Qodo
- youtube B9h9ovW5H9U — Why your agents need decision traces, not just documents — Zach Blumenfeld, Neo4j
- youtube QuuIywMG4s8 — Evals Are Broken, Use Them Anyway — Ara Khan, Cline
- youtube zMiSRliEzv4 — Self Driving Products: Product Signals to Pull Requests — Joshua Snyder, PostHog
Quote signals
- “And what I Turbo puffer what we think this actually means, you know if you break down rag into retrieval augmented generation, you know retrieval is not just vector search.” — youtube UM6sFg_jdlE
- “Um So what we're finding now is that a lot of people are no longer doing the simple rag you know the the Twitter quote unquote rag of just doing vector search once and throwing it into the context windows.” — youtube UM6sFg_jdlE
- “So you know not not a public benchmark but you can trust the numbers they give us.” — youtube UM6sFg_jdlE