Developer asks if his orchestrated RAG chatbot is actually agentic AI or just a workflow
bhavyashah24 · reddit · 2026-10-03
A developer openly questions whether his self-described agentic project is truly agentic, laying out a batch of fundamentals questions plus a full case study.
Questions raised
- Are fixed-step RAG pipelines (chunking, embedding, vector store, generation) agentic at all?
- Does the same LLM perform differently with different peripherals (tools, memory, MCP, web search), and do agent benchmarks standardize peripherals across models?
- What's the real difference between a harness and an agent — continuous execution vs statelessness across turns?
- Why is DAG-based query decomposition rarely discussed, and how do you standardize inconsistent task breakdowns across runs?
Case study: vector DB with auto-updated stale chunks → intent classifier → orchestrator decomposing queries via DAGs → subagents retrieving from vector DB, conversation RAG, and web search → final synthesis. He wants to know how to make this fixed pipeline genuinely agentic.
More from coding & agent
- Replaying agent sessions with cached tool outputs to test fixes and model swaps safely — pauliusztin · 2026-10-03
- Stop building eval sets by hand: grouping agent failures into a regression suite — pauliusztin · 2026-10-03
- Dev argues models are now smart enough to skip GPUI and lay out pixels with raw math — zack_overflow · 2026-10-03
- antirez: AI Coding Tool's Constant Cyber Security Checks Are Blocking Real Work at Redis — antirez · 2026-10-03
- Cheap model's overnight bug-fix PRs: only 6 of 14 survived a stronger model's review — Aggressive-Narwhal-3 · 2026-10-03
- Deep Research MCP open-sources a pipeline to hook five research engines into Claude Code — PMinervini · 2026-10-03