Math Academy #4, Part 2 – Knowledge Graph Engineering: Mental Models & War Stories
8 snips
Dec 3, 2025 Building a knowledge graph resembles city planning, where too many prerequisites can create a cognitive traffic jam. The hosts dive into the evolution of their tooling, emphasizing the importance of efficient command-line tools over complex UIs for internal tasks. Justin shares his journey from research coding to handling real-time systems, highlighting the challenges of production deployment. Alex proposes a 'papercuts team' to tackle minor content issues proactively. They also tease an upcoming differential equations course and discuss its curriculum relevance.
AI Snips
Chapters
Transcript
Episode notes
Prefer CLI Tools For Internal Workflows
- Build command‑line tools for internal, expert workflows instead of full UIs to move faster and reduce maintenance.
- Reserve UI work for customers; internal tools can be powerful CLI utilities.
From Notebooks To Production Nightmares
- Justin learned production engineering the hard way: Jupyter research code became a live service that he had to deploy and maintain.
- Deploying and fixing catastrophic failures taught robust logging, testing, and error containment.
Design For Failure And Validate Changes
- Accept that failures will happen and design code to limit scope and cascade effects.
- Validate graph changes by running the exact selection code to guarantee safety before applying them.