Skip to content
Paula Livingstone writing · projects · tools

Projects

The Frontier Engineer Project

Anthropic is investing $100 million to train 10,000 Frontier Deployed Engineers. This project asks whether an experienced industrial cybersecurity engineer can independently build the same underlying capabilities through targeted study, hands-on engineering and the delivery of a real agentic AI system.

ClaudeClaude CodePythonMCPRAGLLM AgentsLLMOpsCoursera

The Frontier Engineer Project › 01 — Agentic AI

Agentic AI — DeepLearning.AI

The first stage of the programme: moving from conversational LLM use to agentic systems built around reflection, tool use, planning, multi-step workflows and evaluation.

Why this is in the programme

The Frontier Engineer experiment begins with the transition from using a large language model as a conversational tool to engineering systems in which the model participates in a larger process.

This course provides the initial foundation in agentic design patterns, including reflection, tool use, planning and multi-agent workflows. It also introduces evaluation and error analysis, which become increasingly important as a model is given greater freedom to act.

Course

Agentic AI — DeepLearning.AI

The course covers agentic workflows, reflection, tool use, planning, multi-agent approaches, integration with external systems and the evaluation of agent behaviour.

Role in the experiment

This stage establishes the conceptual foundation for the rest of the programme. The objective is not simply to reproduce the examples in the course, but to understand which agentic patterns are useful, what new failure modes they introduce, and where conventional deterministic software remains preferable.

This page will be expanded as the course progresses to record the important concepts, experiments and changes in understanding that emerge.