2026-06-16

Summary

• Added ai/prompt-engineering/ Day 7 few-shot auditor exercise; indexed in ai/README.md.
• Created math/ folder with Day 2 exponents & logarithms for the 50-day math track.
• Fixed ModuleNotFoundError (ollama) in prompt-engineering/ via a local pyproject.toml.

ai/prompt-engineering: Day 7 few-shot auditor exercise

• Added ai/prompt-engineering/ (script + README) demonstrating the four prompt techniques (system prompt, few-shot, CoT, XML tags) on a local-LLM Solidity screen; indexed in ai/README.md.
• Implements the Day 7 exercise: a safe deposit() exemplar primed as a user/assistant pair so the model returns severity=none in the same XML format instead of hallucinating a finding for clean code.

math: Day 2 exponents & logarithms

• Created new math/ folder (script + README) for the 50-day math track; Day 2 covers exponents/logs with three intuitions (compound interest, log scale, product→sum).
• Solves the ETH $1,666→$4,946 exercise: 2.9688× / 196.88%, and the inverse via logs (43.72%/yr discrete, 36.27%/yr continuous); notes the bridge to LLM log-likelihood and the LMSR cost function.

ai/prompt-engineering: fix ModuleNotFoundError (ollama)

• Cause: uv run from prompt-engineering/ walked up to the parent ai/ project (no ollama dep) and used ai/.venv, so import ollama failed.
• Fix: added a local pyproject.toml listing ollama to make the folder a self-contained uv project; uv sync now installs ollama 0.6.2 into a local .venv and the import resolves. Documented the cause + fix in the README Troubleshooting section.