Introduction to the A* Algorithm (2014)

Original Article ↗ Hacker News Discussion ↗

Reposts and “evergreen” content on HN

  • Some complain the article is repeatedly reposted; others argue many readers are new and haven’t seen it, so resurfacing is valuable.
  • Criticism of “hall monitor” behavior around reposts; pointing out duplicates is seen as low-value status-seeking unless it fights spam.
  • Suggestions for better platform features: “evergreen” items that get resurfaced periodically, personalized by what a user has already seen.
  • HN search is noted as useful for finding old discussions, which may contain extra “nuggets” beyond the original post.

Why A and this article remain popular*

  • A* is viewed as the obvious first pathfinding algorithm to teach: easy to visualize, broadly useful, simple extension of BFS/Dijkstra.
  • Several note that A* isn’t emphasized enough in CS curricula despite its power and conceptual elegance.
  • Many attribute the link’s repeated popularity to the quality of the tutorial: strong visualizations, examples, and clear explanations.

A vs “realistic” behavior in games*

  • One commenter dislikes A* as a “performance hack” that makes entities behave omnisciently and unrealistically.
  • Others respond that:
    • The “unfairness” is really about what information the heuristic uses, not A* itself.
    • Games prioritize fun and performance over realism; “reasonable” paths often look better than optimal ones.
    • Human-like limitations can be modeled by hiding information (fog of war, incomplete graphs) or using non-optimal heuristics.
  • Discussion of alternatives and refinements: navmeshes, formation following, cohesion costs, and special handling (e.g., water crossings).

Relationship to other algorithms and mental models

  • Several people frame BFS, DFS, Dijkstra, and A* as essentially the same framework with different data structures or priority functions.
  • A common teaching pattern:
    • BFS → queue; DFS → stack
    • Dijkstra → priority queue by cost-so-far
    • A* → priority queue by cost-so-far + heuristic estimate
  • Emphasis that admissible heuristics must underestimate to guarantee optimal paths, though inadmissible ones can be useful for style or speed.

A as “traditional AI” vs modern “AI”*

  • Some recall when algorithms like A*, logic, and planning were core “AI” topics; now “AI” is often used to mean deep learning/genAI.
  • Discussion of the “AI effect” and how once-understood techniques (like A*) stop being labeled AI.
  • In teaching, A* is placed into a “traditional AI” bucket distinct from machine learning and data science.

Resources, nostalgia, and side topics

  • Strong praise for Red Blob Games in general, especially the hex grid article and interactive visualizations.
  • Multiple commenters share nostalgia: this specific tutorial was their first encounter with A* or a formative learning experience.
  • Brief mentions of pathfinding in unknown environments (SLAM, D* Lite, ML-based exploration) and more advanced techniques like bidirectional search and pattern databases.
  • Minor threads cover pronunciation (“A-star”), jokes about Sagittarius A*, and gripes about poor real-world navigation software.