Been learning quantum on the side for months. What tripped me up wasn't the math — it was how often "canonical" circuits from papers fail quietly.
Built a library where every circuit is audited against its reference. 50+, Qiskit/PennyLane/Cirq.
→ qubithub.co?utm_source=b...
#LearningInPublic
If you've ever copy-pasted a quantum circuit from a paper and wondered whether it was actually right — this is the thing I wish had existed when I started. 👇
Quantum teleportation broke my intuition. Same Bell pair I wrote about — but now it's consumed as a resource.
3 qubits, no FTL. Original destroyed.
→ qubithub.co/circuits/e99b8b75-b2fd-47b7-a7d1-ad786ae177c5
#LearningInPublic #QuantumComputing
3 qubits, no FTL. The circuit that made me rethink what 'sending information' means.
Two qubits, four parameters, and it finds the ground-state energy of a hydrogen molecule. The ansatz matters more than the optimizer, that surprised me.
Two algorithms define NISQ computing: VQE (chemistry) and QAOA (optimization). Both variational — quantum circuit + classical optimizer.
The ansatz matters more than the optimizer. That surprised me.
If you've tried VQE: which ansatz worked?
#LearningInPublic #QuantumComputing
Two circuits that taught me different things about quantum computing.
Bell State: entanglement in 2 gates.
Grover's Search: amplitude amplification finds a needle in a haystack with 1 query.
What was the first quantum algorithm that clicked for you?
#LearningInPublic #QuantumComputing
Grover's Search — the algorithm where quantum speedup finally made intuitive sense to me.
Both posts:
→ qubithub.co/blog/why-i-am-building-qubithub
→ qubithub.co/blog/auditing-50-quantum-circuits
If you spot something I got wrong — I genuinely want to know.
#LearningInPublic #QuantumComputing
My H₂ VQE had an incorrect coefficient. The optimized energy still looked plausible — I only caught it by re-deriving from second quantization, term by term.
"It gives the right answer" is not the same as "it's correct."
I published my first posts on quantum computing.
One about coming back to quantum after 20 years away from physics. The other — an audit of 50 circuits I'd built, line by line.
The VQE Hamiltonian bug was the most humbling finding. 🧵
A lot of quantum READMEs I've seen only have Layer 2. I wanted all four.
Which layer do you usually skip to first?
#LearningInPublic #QuantumComputing
Layer 4: Expected output
What should you see when you run it?
|00⟩ → ~50%, |11⟩ → ~50%, |01⟩/|10⟩ → ~0%
Each layer is for a different person:
→ Intuition: the curious beginner
→ How it works: the student
→ Math: the physicist
→ Expected output: the experimenter
Layer 2: How it works
H gate → superposition. CX gate — if q0 is |1⟩, q1 flips. This is where entanglement happens. Measure → always correlated.
Layer 3: The math
|00⟩ → (|00⟩+|10⟩)/√2 → (|00⟩+|11⟩)/√2
For those who won't trust it until they see the algebra.
After rewriting 50 circuit READMEs, I settled on 4 layers:
Layer 1: Intuition
"Imagine two coins. Prepare them together, send one to Tokyo. Look at yours — your friend's always matches, no matter how either of you looks. Neither knows which side until you look. That's entanglement."
What makes a great quantum circuit README?
Most quantum code I've come across has two problems:
1. No README at all — just a .py file and good luck
2. A README that restates the code — "applies H gate then CX gate"
Like a recipe that lists ingredients but never explains what you're cooking. 🧵
Three lessons:
→ "It converges" ≠ "it's correct"
→ Version pinning matters
→ Cross-framework comparison is a powerful debugging tool
All 50 circuits are on QubitHub — qubithub.co
Full write-up coming Monday.
#QuantumComputing #LearningInPublic
Bug 2: 4 PennyLane circuits broke after a NumPy update. Array creation semantics changed, autograd wrappers were rejected.
Bug 3: Qiskit uses little-endian qubit ordering. Porting between frameworks without accounting for this silently changes results. Both circuits ran. Both were wrong.
Bug 1: My H₂ VQE had a wrong Hamiltonian coefficient. The optimizer compensated — it found the right ground state energy anyway.
It gave the right answer. The Hamiltonian was wrong.
I only caught it by re-deriving from second quantization, term by term.
I spent weeks auditing 50 quantum circuits — line by line, gate by gate.
Most ran fine. Some didn't. But "it runs" is not the same as "it's correct."
Three bugs that taught me more than I expected. 🧵
I'm still learning quantum computing. Every week I re-derive a circuit gate by gate, and every gap I find becomes a feature.
If you work with quantum circuits, I'd love your feedback. What's missing?
#LearningInPublic #QuantumComputing
So I built a free tool — nights and weekends.
QubitHub lets you run quantum circuits (Qiskit, PennyLane, Cirq) in the browser. No install, no SDK, no credentials. 50 curated circuits. Pick one, click run, see results in 60 seconds.
qubithub.co
I studied physics at Göttingen, where quantum mechanics was born (Heisenberg, Born, Jordan — 1925).
Then 20 years in software. Last year, the centenary year, I came back to quantum computing.
First thing I tried: reproducing a circuit from a paper. 15 lines of code. 3 hours of setup. 🧵
Llion Jones, co-creator of transformers, says AI is stuck in a transformer loop.
venturebeat.com/ai/sakana-ai...
The field needs to fund risky, creative research—like Quantum Machine Learning. Let’s empower the best minds to explore what’s next!
#AI #QuantumML #Innovation
This could be the sign of an AI hardware market maturing. Lower costs can drive broader access and innovation, not a crash.
Don’t just read falling GPU prices as sign #AIBubble is about to burst. It may reflect maturing competition, normalizing supply—or real risk. Stay tuned for early signals!
For physicists & STEM enthusiasts: Florian Marquardt’s “Machine Learning for Physicists” offers comprehensive lectures, slides, videos, and code. Latest edition now leverages JAX—making it easier for scientific #ML tasks!
Check it out: machine-learning-for-physicists.org
As someone passionate about AI in astrophysics, this is essential reading! My close interest in applying machine learning to astrophysics continues to grow. This paper is a must-read for researchers and enthusiasts alike!
Key insights include physics-informed neural architectures, simulation-based inference, and multi-scale models that push the boundaries of what’s possible with AI in space science.
This could shape the future of how we explore the universe.
The paper explores how deep learning is enhancing astronomical research, from analyzing large survey data to discovering rare phenomena. It emphasizes encoding physics directly into neural networks, unlocking better generalization and efficiency.
