Lizzie Siegle is a developer advocate at Entire focused on making autonomous software accountable to the people who use it. Her work addresses agent memory and decision provenance, delegated authorization, and the infrastructure that preserves user context after an AI coding session ends.
From developer evangelism to agent infrastructure
Siegle studied computer science at Bryn Mawr College and interned at PubNub and Twilio before becoming a Twilio developer evangelist. She also directed Spectra, a Bay Area hackathon for people of underrepresented genders, and advocated for early-career opportunities in developer relations.
At Twilio, she translated messaging, machine learning, and conversational interfaces into practical projects, including ChatGPT over SMS, an AI personal-training application, and a text-message chatbot using LLaMA 2. Her Cloudflare work expanded into serverless AI infrastructure; one WNBA analytics dashboard combined Streamlit, LangChain, and Workers AI. She subsequently worked in AI developer advocacy at DigitalOcean before joining Entire.
- Agent memory and decision provenance: Siegle argues that commits and pull requests preserve generated code while losing the request, tradeoffs, tool activity, and verification behind it. Preserving that execution history gives subsequent developers and agents enough context to inspect decisions and resume unfinished work.
- Trajectory-aware code review: A plausible final diff does not demonstrate that an agent respected constraints or performed meaningful checks. Evaluating the sequence of actions that produced a change makes failures and unsupported decisions more visible.
- Stateful agent orchestration: Explicit planning, implementation, review, and error-handling stages can make coding agents easier to diagnose and constrain than open-ended tool-calling loops, especially when workflows encounter runaway execution, shifting goals, or exhausted context.
- Delegated authorization and per-user state: At AI Engineer World’s Fair 2025, Siegle and WorkOS engineer Nick Nisi jointly demonstrated authenticated agent access using an OAuth-enabled Model Context Protocol server. Her Cloudflare-focused contribution emphasized Workers infrastructure and per-user Durable Objects: agents need explicit permission to act and persistent context associated with the person they serve.
