Learning plan / recalibrated

Build the middle.
Then go deeper.

Trinity has moved past beginner syntax drills. The next step is not a leap into advanced NOI topics; it is making familiar patterns deliberate, explainable, and repeatable. This page is the single plan: eight slow, gated blocks, each with its own Codebreaker problem list.

48AtCoder accepts
distinct problems
22Codebreaker solves
54 submissions
WK 03 / 22Block 2 by the calendar
blocks advance on gates, not dates

Done so far · Sat 10 Oct 2026

Where things stand.

Fix the two open attempts first, then finish the Block 2 list. Re-solves are short; fit one into every week.

Next up, in order

  1. 01 penguingatheringBlock 1 Fix it, don't rewrite it. Even the N = 1 subtask (answer 0) fails, so this is a bug, not the idea. Likely culprits: (K − X)² written with ^ (XOR in C++), or the minimum starting too small.
  2. 02 prefixsumsBlock 3 · pulled forward TLE: build a long long prefix array once, answer each query in O(1), and use fast I/O (sync_with_stdio(false), '\n' not endl).
  3. 03 sortBlock 2
  4. 04 swapsortBlock 2
  5. 05 hungrycats2Block 2 · mock reserve NOI 2026 Task 2. Skip it and do swords instead if the NOI 2026 paper is being kept unseen for the February mock.
  6. 06 trainorbusBlock 1 Last open problem from the Block 1 list.

Re-solve queue · blank file

  • Solved choosingcandies Choosing Candies · Classic O(N²) brute force, N ≤ 2000
  • Solved party_noi Party · NOI 2024 prelim 1-D DP: max sum, no two adjacent
  • Solved lunchbox Lunch Box · NOI 2016 sort, then greedy

Last session · Sat 10 Oct

First session after three weeks off: 3 submissions, none accepted. penguingathering scored 0 twice; prefixsums timed out.

The sequence · 21 Sep 2026 → NOI prelim, Feb 2027

Eight blocks. No rushing.

Each block earns the next one. Dates are default slots: if a gate is not secure, repeat the block with different problems and let Block 7 shrink instead.

Re-solve recent wins from a blank file and get used to reading constraints before coding. Two problems and all the re-solves are still owed, so this block stays open alongside Block 2.

Practise

Before coding, write the simplest correct approach and one sentence on why it fits the limits.

Good fits

  • short simulations
  • strings and grids
  • small bounded search
  • careful input/output

Gate

Close the block when the process is reliable, not when a calendar week ends.

  • penguingathering and trainorbus accepted
  • state the brute force and its complexity before coding
  • re-solve choosingcandies or party_noi without notes

Problems

  • Solved area_noi NOI 2023 prelim
  • Solved tourist_noi Tourist · NOI 2024 prelim
  • Solved area2 Area 2 · NOI 2026 prelim NOI 2026 Task 1, already seen
  • Solved typo Typo · AtCoder ABC221
  • Solved pairsums Button Pairs · UWCOI 2020 parity counting
  • Solved choosingcandies Choosing Candies · Classic
  • Solved party_noi Party · NOI 2024 prelim
  • Tried, not yet solved penguingathering AtCoder ABC156 C 0, 0 on 10 Oct
  • Not started trainorbus Train Or Bus · NOI 2025 prelim
  • Not started gotoschoolalt Go To School · AtCoder ABC142
  • Not started diamondalt Classic
RE-SOLVE
  • choosingcandies and party_noi from a blank file, 7–14 days after the first solve.

Frequency arrays, sorting before scanning, and greedy choices only where the proof is obvious. Stay with close variants of lunchbox rather than harder greedy.

Practise

Name the pattern (count, sort, or greedy) and its complexity before opening the editor.

Good fits

  • pair counts
  • duplicates and frequencies
  • small greedy after sorting
  • sort by a key

Gate

Name the pattern before implementing, and justify its complexity.

  • three related problems with at most a small hint
  • justify O(n log n) versus O(n)
  • re-solve lunchbox 7–14 days later

Problems

  • Not started sort Classic
  • Not started swapsort Swap Sort · HCI selection 2021
  • Mock reserve: keep unseen hungrycats2 Hungry Cats · NOI 2026 prelim NOI 2026 Task 2; swap for swords to keep the mock clean
  • Not started swords NOI 2023 prelim needs a sort
  • Solved lunchbox Lunch Box · NOI 2016 solved in April after 4 tries; re-solve
  • Not started lengthsortalt Dunjudge archive
  • Not started duckbuttonsstretch Duck And Buttons · NOI 2025 prelim
RE-SOLVE
  • lunchbox from a blank file: say why sorting first is safe before coding.

Answer many range questions without repeating work. prefixsums was attempted early, so the first lesson is already queued. No sliding windows until the prefix-sum model is reliable.

Practise

Draw the array with indices first. Build the prefix array by hand on a tiny example before coding.

Good fits

  • many range sums
  • running totals
  • range updates
  • recovering an array from differences

Gate

A correct prefix-sum solution makes its indexing convention obvious.

  • turn repeated range work into preprocessing
  • test the boundaries on paper
  • state the cost per query after preprocessing

Problems

  • Tried, not yet solved prefixsums Classic TLE (2.01 s) on 10 Oct
  • Not started differencearray Difference Array · Classic
  • Not started labels NOI 2020 final rebuild the array from its differences
RE-SOLVE
  • prefixsums from a blank file, including fast I/O, a week after it is accepted.

The operating system

Slow enough to stick.

The plan is deliberately lighter than four brand-new problems every week. Depth beats the appearance of speed.

02

New problems

Two or three from the current block's list, not a quota.

01

Guided review

Rebuild one solution with the key idea named first.

01

Timed mini-set

30–45 minutes, low pressure. Starts after Block 3.

01

Re-solve + explain

Repeat an old problem and explain the invariant or recurrence aloud.

Why this level

Read the signal,
not the fantasy.

The September plan moved too quickly; its verified problem lists are kept here, re-paced. These public snapshots show a learner with real traction, but not yet enough evidence for an advanced-topic sprint.

01

What is working

Direct implementation, simulation, counting, parity, sorting-based greedy, and a basic one-dimensional DP pattern are already appearing in accepted work.

02

What needs structure

Algorithm choice, complexity estimates, edge-case checks, and independent re-solving are the next layer — not remedial content.

03

What is deliberately later

Advanced DP and broad graph technique are conditional. They open after the middle blocks become reliable under light time pressure.

04

What the pace allows

About two new solves a week since late August. Kept up to February, that is roughly 35–40 more problems: enough for Blocks 2–5 done properly. Block 7's branches open only if that pace improves.