🎯 The Challenge

Join BlueQubit's quantum hackathon, run together with DuQuantum 2026 at Duke University! Most participants will compete online, and some will join in person at Duke. The competition is open to all experience levels, from quantum enthusiasts to seasoned researchers.

In 2025, Google's Quantum Echoes experiment (Nature, 2025) reported a verifiable quantum advantage. Its Willow chip measured out-of-time-order correlators (OTOCs), which track how quantum information scrambles. Google estimates it did this about 13,000× faster than the best known classical method on one of the world's fastest supercomputers. You take the classical side.

• The task: four hidden circuits, G1 to G4. Estimate one OTOC per qubit, 140 graded numbers (plus 50 for the samples), by running the circuits on BlueQubit by ID. Three sample circuits (S1 to S3) have open QASM you can download and simulate any way you like.
• Scoring: each circuit is its own problem, worth corr² × its points (the squared correlation with the exact values, from 0 to 1). Samples are worth 100 points each, and G1 to G4 are worth 300, 600, 800 and 1,000.
• AI is encouraged: the challenge is built for AI agents. You steer, and your agent runs the simulations.

New to OTOCs? Start with our OTOC tutorial.

📅 Schedule

All times are Eastern Time (ET).

• Hacking starts: Saturday, October 24, 11:30 am ET (circuits and full challenge details are released at this time) 🚀
• Hackathon duration: 24 hours
• Hacking ends: Sunday, October 25, 11:30 am ET (final submissions due)

DuQuantum in-person participants: check-in opens at 9:00 am ET on Saturday at the Wilkinson Building, and the closing ceremony ends at 4:30 pm ET on Sunday.

⚡ Quick start

1. Before the event: create a BlueQubit account and join this hackathon. Online, you compete individually. In person at Duke, a team of up to 4 shares one account.
2. At kickoff: redeem your $30 credit coupon, code UOECJBRJDA (Account → Billing → Redeem Coupon).
3. Work in Hybrid mode on a SMALL machine. It can stay on for the full 24 hours, and the SDK and your API token are already set up there. Claude Code is installed and ready to go, and we encourage you to use it: give it the getting-started notes below and let it run BlueQubit's exact, MPS and Pauli-path simulators. You can also work on your own computer.
4. Need more power? Occasionally switch to a LARGE machine for notebooks, or ask your AI agent to bring one up and run simulations on it. LARGE machines have a limit of 6+ hours (check the usage page). There are no GPU notebooks, but bq.run can run GPU jobs (device="gpu"), also with limited availability (check the usage page). Machine time and larger jobs use your credits.
5. Submit one CSV per circuit on that circuit's problem, with a short note on your method. Up to 20 submissions per problem, and your best one counts.
6. Deadline: 11:30 am ET, Sunday, October 25.

🤖 Getting started with your AI agent

Paste these into your agent (Claude Code or another). In a Hybrid machine, the SDK and your token are ready. On your own computer, use Python 3.11+, copy your token from See API token (top bar), and run export BLUEQUBIT_API_TOKEN=<your token>.

1. The task

BlueQubit OTOC challenge. Seven circuits: S1, S2, S3 (open samples) and
G1, G2, G3, G4 (hidden). Each circuit is one problem.
Every circuit has one ID per qubit: Duke_<circuit>_<q>, q = 0 ... n-1
(e.g. Duke_S2_5 = circuit S2, qubit 5). Each ID runs its own QASM program.
The value to estimate for an ID: run its circuit from |0...0> and take <Z>,
the expectation value of Pauli Z on qubit q. A number in [-1, 1].
Qubit q is q[q] in the QASM register. Keep that order: swapping qubits
ruins the score. All answers are hidden.
You don't need the physics terms (OTOC, echo, butterfly) to submit.

2. Samples: download the circuits

Problems 1-3 each have a zip: one OpenQASM 2.0 file per ID, named
<ID>.qasm (e.g. Duke_S1_0.qasm). Gates: u3, cz, x, z. No measurements:
compute <Z> on q[q] from the final state. Each file is a different
circuit, so simulate each one. Use any tool, or run them by ID (below).

3. Run by ID (the only way to run the graded circuits)

pip install -U "bluequbit>=0.19.0b1"

import csv
import bluequbit

bq = bluequbit.init()                      # uses BLUEQUBIT_API_TOKEN
circuit = "S1"                             # S1 ... S3, G1 ... G4
n = bq.get_circuit_info(f"Duke_{circuit}_0").num_qubits   # free
rows = []
for q in range(n):
    cid = f"Duke_{circuit}_{q}"
    z = "I" * (n - 1 - q) + "Z" + "I" * q  # Z on qubit q, q0 = rightmost
    job = bq.run(cid, device="cpu", pauli_sum=[(z, 1.0)])   # waits
    rows.append((cid, job.expectation_value))
with open(f"{circuit}.csv", "w", newline="") as f:
    csv.writer(f).writerows(rows)

Devices: cpu, gpu, mps.cpu, mps.gpu, pauli-path (settings: see the docs).
cpu is exact and only fits small circuits. Small jobs are free, and larger
ones use your credits.

📘 Devices, settings, size limits and costs: the BlueQubit Python SDK docs (every bq.run option), the platform docs and the tutorial Choosing a quantum simulator.

4. Submitting and scoring

One CSV per circuit, uploaded on that circuit's problem (Submissions tab),
with a short note on your method. Rows: circuit_id,value, one per ID,
any order. Optional header line: circuit_id,value. IDs exactly as listed
(case-sensitive). Values: any finite number, plain or scientific notation.
Any file name.
The upload checks the IDs: a file with a missing or wrong ID is rejected
before you submit, and doesn't use an attempt.
Up to 20 submissions per problem. Your score appears right away, and the
leaderboard counts your best submission on each problem.

Score per problem: corr² x the problem's points, where corr is the Pearson
correlation between your values and the exact values over that problem's
IDs. corr² runs from 0 to 1. Scale and offset don't matter, so no
normalization is needed. The sign does: a negative correlation scores 0,
and so do all-equal values.

Start with the samples: get a CSV for S1 submitted end to end, then build from there. More: tutorials.

🆘 Questions and help

Email rhik@bluequbit.io, or ask in the #duke-2026 channel on the BlueQubit Discord. Asking the organizers is always allowed, including for online participants.

🎟️ Credits

Each participant (or in-person team) gets $30 in BlueQubit credits at kickoff. Coupon code: UOECJBRJDA. Redeem it at Account → Billing → Redeem Coupon. Track spending on the usage page. Small jobs are free.

💬 Frequently Asked Questions

1. Who can participate? Students, researchers, hobbyists, and quantum enthusiasts (18 or older) from anywhere can join online. In-person participation at Duke is limited to accepted DuQuantum 2026 participants.

2. Can I participate with a team? Only in person. Online participants compete individually, and only the account holder is recognized. In-person DuQuantum participants at Duke can form teams of up to 4, under Duke's guidelines, sharing one account.

3. How is this related to DuQuantum at Duke? Most participants in the BlueQubit hackathon will compete online, and some will take part in person at Duke as part of the DuQuantum 2026 hackathon. Everyone appears on the Leaderboard tab under the "World" view, and DuQuantum participants also appear under the "Duke" view.

4. Do I need prior experience in quantum computing? Not really. Knowing what a quantum circuit and an expectation value are will get you started. Experience with quantum simulators and approximation methods will give you an edge.

5. How long does the challenge last? 24 hours, from 11:30 am ET on October 24 to 11:30 am ET on October 25.

6. What kind of problems can I expect? Seven, one per circuit: three open sample circuits and four hidden graded ones. See "The Challenge" above.

7. How do I prepare?

• Get comfortable with quantum circuits and expectation values. BlueQubit's tutorials are a good place to start.
• Work through our OTOC tutorial: what OTOCs measure, why they're hard to simulate, and how to run exact, MPS and Pauli-path simulations on BlueQubit.
• For background, read Google's Quantum Echoes paper, Observation of constructive interference at the edge of quantum ergodicity (Nature, 2025).
• Read up on approximate methods such as matrix product states and Pauli propagation.

8. Which tools can I use? The graded circuits can only be run by ID, on the BlueQubit platform or with the BlueQubit Python SDK. For the sample circuits, you can also download the QASM files from Problems 1–3 and use any tools you like. AI assistants are welcome on every problem.

9. How are winners selected? By total points across all seven problems (3,000 in all), with each problem counting your best submission. Each submission also records a time penalty: the time since the start. Your leaderboard time penalty adds these up over your best submission on each problem, and the lower total breaks ties.

10. How do I keep track of my compute? On the usage page.

11. Is there a participation fee? No. Registration is free!