Advent of Code 2025: Number of puzzles reduce from 25 to 12 for the first time

Original Article ↗ Hacker News Discussion ↗

Reduced number of puzzles (25 → 12)

  • Many welcome the change because they regularly dropped off around days 7–18 due to holidays, work, or family.
  • Several say they only ever finished ~half the problems anyway; 12 feels more realistic and less stressful.
  • Some note that puzzle difficulty spikes late in the month just as personal time shrinks, making the old format hard to enjoy.
  • A minority are disappointed, having just managed to complete all 25 last year or liking the full “calendar to Christmas” feel.

Time, stress, and personal life

  • Common theme: AoC increasingly competed with family time, holiday prep, and jobs; a daily hard puzzle felt like a “stressful Christmas calendar.”
  • People describe all‑nighters, frustration, and burnout trying to keep a perfect streak, then relief at feeling “permission” not to.
  • Others already treated AoC as a January or slow‑period activity and say the format change won’t affect them.

Suggestions for alternative formats

  • Popular ideas (mostly rejected or seen as trade‑offs):
    • Release a puzzle every two days to still span the month.
    • Make each part (1 and 2) separate days to get back to ~24 items.
    • Insert rest days or easier late‑month puzzles.
  • Counterarguments: simple part‑2s would make some days feel trivial; daily releases preserve flexibility (participants can self‑pace).

Removal of the global leaderboard

  • Widely expected due to increasing AI use and “LLM farms” auto‑submitting answers in seconds.
  • Some will miss the race and the thrill of waking at release time to speed‑solve; others felt the leaderboard always skewed the event toward unhealthy competition and timezone unfairness.
  • Private leaderboards and informal competitions remain popular as a compromise.

Debate over “Advent” and event identity

  • Extended argument over whether “Advent of Code” implies 24/25 problems (advent calendar) or just a pre‑Christmas coding season.
  • People cite differing cultural traditions (24 vs 25 doors, liturgical Advent length, 12 days of Christmas) to argue that 12 puzzles is still thematically fine.

AI, cheating, and puzzle design

  • Some try to design puzzles humans can solve but LLMs struggle with (e.g., altered rules that trigger AI priors, ARC‑like pattern tasks).
  • Others argue that if you enjoy solving, AI “cheaters” on a leaderboard shouldn’t matter, but acknowledge policing any AI policy is impossible.
  • A side note shows an LLM clearly leaning on AoC‑specific terminology, suggesting training exposure.

Accessibility and site UX

  • One participant criticizes the default color scheme as low‑contrast and basically unreadable, noting that alternate stylesheets or browser reader modes aren’t a practical fix for most users.

Overall sentiment

  • Dominant mood: mild sadness at the end of an era (25 days + global leaderboard) mixed with strong appreciation for a decade of free, high‑effort puzzles and support for reducing the creator’s burden so AoC can continue.