References

Research and resources that inform the pedagogical choices in this program.

LLM-Assisted Development Education

Anthropic (2026). “How AI Assistance Impacts the Formation of Coding Skills.” RCT with 52 junior engineers. Students using AI for conceptual inquiry retained skills; those delegating code generation scored 17% lower on comprehension. https://www.anthropic.com/research/AI-assistance-coding-skills
CodeAid (CHI 2024, Microsoft Research). LLM assistant deployed to 700 students that deliberately avoids giving code solutions. Provides pseudocode, explanations, and fix annotations instead. Four design principles for educational AI. https://arxiv.org/abs/2401.11314
HypoCompass (IJCAI 2025). Students act as TAs helping an LLM debug code, forcing them to reason about code rather than consume generated solutions. https://www.ijcai.org/proceedings/2025/1217.pdf

Osmani, A. (2026). “My LLM Coding Workflow.” Practitioner guide for spec-driven, small-chunk LLM-assisted development. “You are the senior dev” principle. https://addyosmani.com/blog/ai-coding-workflow/
ASDLC.io. “Adversarial Code Review Pattern.” Builder/Critic separation with fresh sessions to avoid confirmation bias. Constitutional directive for the Critic role. https://asdlc.io/patterns/adversarial-code-review/
Fowler, M. “Architecture Decision Records.” Lightweight documentation of design decisions. MADR template. https://martinfowler.com/bliki/ArchitectureDecisionRecord.html https://adr.github.io/

Meta Engineering (2025). “LLMs Are the Key to Mutation Testing.” Using LLMs to generate meaningful mutants and tests to catch them. Validates test quality beyond coverage metrics. https://engineering.fb.com/2025/09/30/security/llms-are-the-key-to-mutation-testing-and-better-compliance/
LLM4TDD. “Best Practices for Test-Driven Development Using LLMs.” Write tests first as specifications, then have LLMs generate implementations that must pass them. https://arxiv.org/pdf/2312.04687

Koh, K. et al. (2019). MFAI Methodology. The Monthly Food Access Index methodology that FEAST implements. Defines how household food access scores are calculated based on store proximity, household resources, and trip frequency.
USDA Economic Research Service. “Food Access Research Atlas.” USDA’s definition of food deserts and low-access areas. Background for understanding what FEAST is modeling. https://www.ers.usda.gov/data-products/food-access-research-atlas/

Mesa: Agent-Based Modeling in Python. The ABM framework FEAST uses. Documentation for Model, Agent, GeoSpace, DataCollector. https://mesa.readthedocs.io/
Claude Code Best Practices. Anthropic’s guide for effective use of Claude Code, including context engineering. https://docs.anthropic.com/en/docs/claude-code/best-practices