Advent of Code 2024

Original Article ↗ Hacker News Discussion ↗

Advent of Code 2024 is prompting both enthusiasm and soul‑searching among programmers who use it as a yearly coding ritual, a way to learn new languages, or a shared puzzle tradition with friends and colleagues. Many are worried that AI coding assistants and automated LLM pipelines undermine the public leaderboard and the event’s spirit, while others argue that using modern tools is no different from relying on libraries or Stack Overflow and that private boards or personal goals matter more than race times. Alongside debates over cheating, fairness, and burnout, participants trade tips on tooling, parsing strategies, accessibility tweaks, and unconventional approaches—from solving puzzles in SQL, F#, or Ada to running them on old hardware or inside homegrown languages.

AI, cheating, and the global leaderboard

  • A 9‑second double‑star solve on Day 1 was traced to an AI‑generated solution and later-removed apology, triggering debate about LLM “cheating.”
  • Many argue LLMs make the public leaderboard meaningless: models read faster than humans and can be automated to fetch puzzles, generate code, run, and submit answers.
  • Others say AI is now a normal tool (like Stack Overflow or autocomplete) and should either be allowed explicitly or moved to a separate AI leaderboard.
  • Several liken AI use on the public board to aimbots in games or Stockfish in chess tournaments; others counter that programming isn’t inherently a sport and tools shouldn’t be forbidden.
  • Some are impressed by the automation challenge itself (pipelines, benchmarking o1‑style repeated runs), but still see it as incompatible with the event’s spirit.

Competition vs. personal enjoyment

  • Many participants say they ignore the global leaderboard due to time zones, cheaters, and extreme competition; they prefer private boards with friends or colleagues.
  • A recurring pattern: people enjoy the first ~7–12 days, then puzzles become time‑consuming and stressful, leading to burnout or abandonment.
  • Strategies include: setting per‑puzzle time limits, skipping hard days, finishing after December, or doing only first stars.
  • Some view AoC as a fun tradition and a way to practice problem solving, not a career or productivity exercise; others advocate doing side projects instead for longer‑term benefit.

Learning, languages, and tooling

  • Large contingent uses AoC to learn or practice languages: F#, Gleam, Rust, Go, Swift, Ada, SQL/SQLite, K/APL, Elixir, Lisp variants, Prolog, bash, Excel, Whitespace, custom languages, even NES/STM32 targets.
  • Many build personal frameworks/CLIs, input parsers, grid/graph utilities, or benchmarking rigs; some note they over‑invest in frameworks instead of solving puzzles.
  • AoC is contrasted with LeetCode: AoC is seen as more playful, story‑driven, and community‑oriented, with less emphasis on textbook algorithms and more on parsing and ad‑hoc problem solving.

Difficulty, algorithms, and accessibility

  • Disagreement over how “beginner‑friendly” AoC really is: some say you can get far with loops and brute force; others note recurring need for more advanced ideas (graphs, DP, CRT, linear algebra).
  • Several stress that optimal algorithms are often not required for personal success; brute force plus patience works for many inputs.
  • Site UX is widely criticized: tiny thin font, dark theme, and poor mobile support; people recommend browser reader modes, user CSS (Stylus), userscripts, or CLI tools to fetch and re‑render puzzles.