We don't need more chatbots: an agent harness lead's field notes
threepointone · x · 2026-11-14
- Kylie, who spent the past year turning a Q&A bot into a platform-capable agent harness (prompt engineering, evals, monitoring, automated guardrails), argues that LLMs graduated from basic Q&A back in 2024.
- If a user can Google it, they don't need your chatbot. What keeps users engaged are action-oriented prompts: integrating product X into their existing stack Y, checking whether a vendor's recent releases matter to their setup, evaluating adoption paths for Z.
- Most agents on the market are just "information discovery" wrappers (car-configurator chats, founder net-worth lookups); she finds them useless. Even the dentist AI receptionist she likes probably works because the human team replies fast once the AI-routed request lands in their inbox.
More from coding & agent
- InferenceForge claims 43% higher LLM inference throughput than best AgentX baseline — kaiwei_chang · 2026-11-14
- Anthropic ships Self-Hosted Claude Managed Agents with first-party Evals product — threepointone · 2026-11-14
- David Khourshid: agents will force a rethink of libraries built to save developers typing — DavidKPiano · 2026-11-14
- Replit employee automated her sales job with AI, moved on to building GTM tools — amasad · 2026-11-14
- Compound V2 launches agent-native Office engines, tops PE Bench at 80.5% with 34% lower cost — peterjliu · 2026-11-14
- IT director says Hermes Agent built a full Windows domain flawlessly for 27 cents — NousResearch · 2026-11-14