Skip to content
Sulayman
Bowles

Algorithms / Constraint solving VI

Jane Street Puzzle Solver

One path.Every rule checked.

An exact-search Python solver for a three-dimensional knight puzzle, paired with an independent verifier and machine-readable audit tables.

Read the case study

Illustrative path motif, not the solved puzzle board.

My role
Solver author
Form
Python · Exact search · Independent verification
Period
July 2026
Status
Solver & verifier

Finding a plausible answer and proving that it follows every rule are different jobs.

01 / The problem

Turn the puzzle into constraints.

The July 2026 Jane Street puzzle combined schedule interpretation, three-dimensional knight moves, regional tower placement, and score arithmetic. A valid route had to satisfy all of them, including the no-repeat rule and the final stopping condition.

I treated it as an exact reconstruction problem. The search tests schedule interpretations against the arithmetic, then joins path segments between forced clue moves while enforcing the geometric and regional constraints.

Working principleThe answer is the last line
Recorded solver output33,609
Schedule constant
K = 7
Path length
54 moves
Check
Separate verifier

Geometry. Check three-dimensional knight legality, regional tower placement, and the no-repeat rule.

Arithmetic. Recompute the scoring schedule, clue scores, and unvisited-neighbor sums.

Verification. Run a separate check of the proposed path rather than treating search completion as proof.

Recorded outputs from the solver package. This display does not execute the original solver.

02 / The work

Do not let the search grade itself.

A solver can produce a convincing-looking result while repeating an error in its own assumptions. I wrote a separate plain-Python verifier to recompute legality and scoring from the board, tower placements, schedule, and path coordinates.

The repository also generates a machine-readable solution, a path table, a clue audit, and unvisited-neighbor sums. Those outputs make the answer inspectable without asking a reader to trust a search log.

03 / The judgment

Leave a reproducible record.

The retained solution records K = 7, a 54-move path, and a final answer of 33,609. The implementation uses the Python standard library; the search, verification, and artifact-writing responsibilities are separate.

The useful result is the complete chain: a constraint model, an exact search, a second check, and outputs that explain what was checked. The solver record is distinct from Jane Street’s own publication of correct submissions.

04 / Decisions

What I chose.
And why.

Exact reconstruction
Forced clues and scoring arithmetic constrain the search instead of relying on a sequence of guesses.
Independent verifier
Recheck the proposed path outside the search procedure.
Inspectable outputs
The path table and clue audit expose the work behind the final number.

05 / Result

What came out of it.

A retained solver package recording K = 7, a 54-move path, the answer 33,609, and separate verification artifacts.

Status & source notes

The numerical result is drawn from the retained solver repository. The public dossier does not embed the original code or establish official leaderboard confirmation. No claim of a fresh execution or independent organizer verification is made by this page.

Case study updated September 12, 2026.

Next case study / VIIInternship Aggregator Engine