Three architectures for agentic search: where to put the intelligence
hugobowne · x · 2026-08-29
Software Doug breaks down 'agentic search' into three distinct architectures based on where the intelligence sits:
- Skip search engines: Give the agent tools, Markdown files, and grep.
- Smart Retrieval Layer: Enhance the retrieval system to handle complex agent queries.
- Model Takeover: Fine-tune models to handle more of the search task.
Each choice changes the engineering workflow, metrics, and debugging requirements.
More from coding & agent
- Can agent Skills be repurposed to fragment repo context beyond 'things to do'? — amzwC137 · 2026-08-29
- llmog: Local tool for automatic dataset annotation and reclassification using LLMs — SavingsWeather1659 · 2026-08-29
- 4 Claude agents competed to build a music video editor, judged blindly by 3 more — BuffaloConscious7919 · 2026-08-29
- Firecrawl Developer Index launches on OpenAI plugin marketplace — devdigest · 2026-08-29
- User burns through 20x Claude Max plans weekly to maintain complex CRM automation — Fun_Ad7909 · 2026-08-29
- Self-driving sim in pure NumPy: 9 raycasts + genetic algorithm clears unseen track — rsasaki0109 · 2026-08-29