Guide

This page covers the whole workflow: pick a method, form a group (1–3, self-organized on Day 1), fork the central harness, build your solution, and open a pull request. The only hard timing constraints are: PR open by Day 4 morning (CI must be green), advisor buy-in decisions on Day 5 morning, and selected teams present on Day 5 afternoon. Paste each ▶ Prompt into Claude Code, Codex CLI, or OpenCode. Prerequisite: an AI coding agent installed — see Setup below.

// Methods

|n⟩

Exact Diagonalization

Full diagonalization of finite-size Hamiltonians.
Expert: Chen Cheng (程晨) · /method-ed

⊗

MPS / LTRG / DMRG / TEBD

Tensor-network methods for 1D and finite-temperature systems.
Expert: Wei Li (李伟) · /method-mps · /method-ltrg

⊞

PEPS / CTMRG

Projected entangled pair states and corner transfer matrix renormalization.
Expert: Hai-Jun Liao (廖海军) · /method-peps

∮

Quantum Monte Carlo

Stochastic sampling of partition functions and ground states.
Expert: Ming-Pu Qin (秦明普) · /method-qmc

↻

Monte Carlo Renormalization Group

Monte Carlo renormalization group methods.
Expert: Yan-Tao Wu (武琰涛) · /method-mcrg

𝒰

Quantum Circuit Simulation

Classical simulation of quantum circuits and gate-based algorithms.
Expert: Shi-Xin Zhang (张士欣) · /method-qcs

⪰

Semidefinite Programming

SDP relaxations for rigorous ground-state energy lower bounds.
Expert: Jie Wang (王杰) · /method-sdp

⌘

AI Agent and Knowledge Base

Harness skills and knowledge-base infrastructure.
Experts: Kun Chen (陈锟) · Jin-Guo Liu (刘金国) · tracks/agent-kb

Method slugs (used in folder paths and PR titles): ed, mps, peps, qmc, mcrg, qcs, sdp, agent-kb.

Source of truth: QuantumBFS/quantum.harness method table.

// Setup

Set up your AI agent tooling before the hackathon. The harness assumes you can fork a repo, run a make target, and converse with an agent in your terminal.

Tool	Type	Link
Claude Code	Terminal CLI	code.claude.com
Codex CLI	Terminal CLI	github.com/openai/codex
OpenCode	Terminal CLI	opencode.ai

Get one installed and authenticated before July 26.

Install superpowers and gh:

Install https://github.com/obra/superpowers and gh (GitHub CLI) if not already installed.
Authenticate me to GitHub via gh auth login if I'm not logged in.

Choose your model with /model — Claude Code: Opus 4.7 (high effort); Codex CLI: GPT-5.4 (xhigh effort); OpenCode: route to one of the above via your configured provider.

Fork, clone, and branch the harness:

Fork https://github.com/QuantumBFS/quantum.harness to my GitHub account using gh repo fork.
Clone the fork to ~/code/quantum.harness and add the upstream remote.
Create a working branch named group-<my-slug> off main.
Verify with git status, git remote -v, gh auth status, and the current branch name.

Pick <my-slug> as a short lowercase-hyphen team name (e.g. wolf-pack). Groups are free-form and self-organized — size 1–3, one PR per group.

// Onboard

Reproduce the reference result:

/track-starter

Run /track-starter with no argument. Follow it to reproduce the reference result, then report what passed, what failed, and the runtime environment used.

Brainstorm your attack on a challenge issue (pick one from the Challenges page):

Read challenge issue #<N> from https://github.com/QuantumBFS/quantum.harness/issues/<N>.
Brainstorm with me: surface 3–5 distinct angles, name the riskiest assumption in each, and recommend one to start with.

// Build & Submit

Work under solutions/<method-slug>/<my-slug>/ on your fork. When ready, open a PR — deadline: Day 4 morning.

Your PR should include:

Improvements to the harness system, such as a method skill upgrade.
Your challenge solution in solutions/<method-slug>/<my-slug>/.
One reproduction prompt in the PR comments, generated with the provided challenge-report skill.

Create a pull request to QuantumBFS/quantum.harness:main following the submission guideline at https://giggleliu.github.io/summer-school-2026/challenge-ideas#submission-guideline.

// Demo & advisor review

Use Day 4 afternoon to make your PR easy to review: run the demo and check the README. On Day 5 morning each track meets separately. Each advisor has one or more sponsor-backed Mac mini awards to allocate and can use them at their own discretion.

The judging question is simple. Would the advisor spend a Mac mini award on this result? If yes, the selected group has two obligations: spend about 2 hours polishing the skills/mini-harness so they can be merged into the repository, and provide the chat history (we will provide skill to extract it) for article writing or other research analysis.

// Resources

Learning:

MCP servers, CLI tools & skills for researchers: see the full Resources page.