A capstone, in three conference rooms

In January none of us had seen a qubit. In April we were the ones explaining them.

I'm writing this the afternoon after our final presentation at Wood Mackenzie. The deck is uploaded, the happy hour is booked, the last bit of adrenaline is just starting to leave. So before I lose the shape of it, I want to put this down.

January

The project arrived like a door you didn't know was there. SIPA calls them "capstones"; Wood Mackenzie had commissioned one on quantum computing in the energy sector. Six of us were assigned — Akshay, Coco, Ruby, Terry, Vincent, me. None of us had touched quantum before. Our honest combined starting knowledge was: a half-remembered Feynman anecdote, two PBS episodes, and the general sense that IBM had one and China had one and they were probably a big deal.

The first meeting was on the top floor of IAB. Jeanne Fox — our advisor — was measured about it. Ten to twelve hours a week. Rotating meeting secretary. Find a two-hour block, in person, every week. Ben would meet us every other Friday. Midterm at Wood Mackenzie the Friday before spring break. Final at Wood Mackenzie in late April. That was the whole architecture.

What she didn't say, because she didn't need to, was that the first month would be hard in a particular way. Not the technical hardness. The open-endedness. When you don't know what you don't know, you can't plan — you can only read, and then read something that the first reading made you think to read.

February, mostly

Some weeks in February I felt like I was reading in a language with the nouns removed. Variational Quantum Eigensolver. Quadratic Unconstrained Binary Optimization. Noisy Intermediate-Scale Quantum. Each term was fine in isolation; it was the recursion that killed you. Every definition used three other terms that themselves needed unpacking. You could spend a day on one paragraph.

The team was also tired. Six people on a Wednesday afternoon, going through readings none of us were quite confident we understood, trying to decide which applications to assign to whom. Ruby took nuclear and the policy chapter. Terry took grid optimization and the case-studies survey. Vincent took solar, and built the technoeconomic matrix the rest of us would later score every application against. Coco took quantum sensing and the QC 101 chapter. Akshay — project lead — took batteries, the hardware-modalities chapter, and the AI-data-centers piece. I took finance and the Monte Carlo explainer. The handoffs were not clean. Nothing was clean in February.

I remember a meeting where someone — I think it was Terry — just said I don't know what we're supposed to do this week and we all sat there for a minute. Then someone else made a list on the whiteboard. That is mostly how we got through it.

March

Things turned in March. I can almost date it.

March 13, we took the train into Wood Mackenzie's office and presented the midterm to Ben. The deck was rough — it still had "Sample text" in three places and a lorem ipsum paragraph under battery chemistry — but the bones were there. Six applications. A three-axis matrix. A rubric. Ben liked the QC 101 structure. He gave us a list of real corrections: Energy saving varies across use cases, make that explicit. Nuclear needs computational framing, not just nuclear framing. Spell out acronyms at least once.

We walked out into Hudson Yards afterward and it was one of those days where the wind is cold but the light is already long. I remember feeling, for the first time in the project, that we were building something rather than just not-failing. It is a small distinction and it changes everything.

After that the work got faster. We started an Excel scoring harness. The matrix tightened. Interviews — Pete Shadbolt from PsiQuantum, Zach Rainey from Fujitsu, a few others — came through in March and early April, and each one was calibration. You learn quickly that experts disagree more than the papers suggest, and that this is information, not noise.

April

April was the month of tightening. The deck moved from Draft 1 to Draft 3 to Draft 4. Every meeting Ben would circle something and say make it more concrete and we would make it more concrete and by the next meeting he would circle something else and say make it more concrete. Late in April he said accuracy drives timing and we wrote it down and it became the first of our four overarching findings. That's how most good ideas entered the deck — Ben said a thing offhand and we wrote it down.

And then today. We presented to Ben in person and Trent on the phone, and the deck landed cleanly, and there were questions we could answer, and then it was done. Four months.

Trent

There is one funny thing I want to put on record before I forget, and it is that I have never actually met Trent Yang.

Trent is the other client lead on our project. He works in a sustainability-focused venture fund. He was introduced to us in November. He has been on almost every Zoom call we've had, and he asks good questions — particularly about whether the efficiencies we're attributing to quantum could also be obtained classically, which is the right question to ask and the one that keeps you honest. His voice I would recognize in a crowded room. His face, on a laptop screen, yes.

But he was not at the midterm at Wood Mackenzie. He was not at the final today. He has been reliably somewhere else for every in-person moment of this project, often an airport. On April 10, he joined the client review from a terminal because Google Meet had broken on his phone; we ended up calling out slide numbers so he could follow along.

I'm not complaining. Trent was an excellent client, by every measure that matters. He just exists for me the way a character in a podcast exists — a voice, an opinion, a recognizable cadence, and no body. I will probably never see him in person, and I find this strangely fitting for a capstone on quantum computing. There is a joke here about superposition that I am going to resist making.

What I'm left with

A few things, in no order.

That a team of six people who don't know something can, given four months and one good advisor, know it well enough to teach it back. Not at the PhD level — we are not physicists — but at the level where a VP of research at Wood Mackenzie nods and says yes, that's the framing. That is a lot of what a good brief is.

That the second-hardest thing in a capstone is the team. Six people, six schedules, six standards for what done means. Most of our friction was about that, not about quantum. We got through it by doing the boring coordination work — agendas, rotating minutes, 24-hour turnaround on notes, master reference list. The boring work is what made the interesting work possible. I wish I had known that earlier.

That the hardest thing is sitting with not-knowing long enough to let the shape emerge. In January I kept wanting to commit to a thesis before we had earned one. Jeanne kept pulling us back to the reading. She was right and I was wrong, and I will remember that the next time I want to commit early.

And that I'm grateful — to the team, to Jeanne, to Ben, and (by Zoom) to Trent.

The conference room is empty now. The deck is somewhere on a Wood Mackenzie server. Everyone's going to a bar on 10th Avenue. I'm closing the laptop.