A quantumful of throughts
The bull case for peer review is that, when done right, it is an extremely powerful tool for vetting and improving scientific results. The bear case is that peer review is rarely done right; instead, it is a low-throughput, mixed-quality process that critically relies on highly skilled, unpaid labor. Wherever you sit on this, I think we agree that some aspects of peer review are badly broken.…
The joke is that every time that a new AI model comes out, someone throws it at the Collatz conjecture just in case. It’s actually probably not the worst idea, but “nails high-school physics but cannot yet create the theory of everything” is not the most useful description of the level of physics proficiency of AI models. What I need to know is: as of mid-2026, how good is frontier AI at frontier…
By now, we have all seen how AI software engineers, AI chatbots, and AI copywriters change the game. In this blog series, I wanted to go broader, and extrapolate from my personal and professional experiences in quantum computing on how AI will impact deep tech R&D. This post is part 1, which is all about how crazy good AI can be at doing everyday physics .
Writing this blog over the last 2.5 years has been a highly rewarding experience. This is both because I genuinely love writing, and because many of you seem to enjoy it. Time for a sneak peek behind the scenes, a preview of what’s coming up, and an ask for your help!
One of the main metrics used to describe a quantum computer is qubit coherence time, typically denoted as $T_2$. There is just one problem: $T_2$ is pretty useless at predicting the effect of qubit decoherence on computational errors! In this post, I explain why that is, and argue for adopting a new coherence measure more fit for the task: $T_{3}$.
When I left academia in 2021 to join Oxford Ionics, not much changed day-to-day to start with. Despite significantly different goals and structure, life in a small-scale QC startup was just not that different from life in an academic lab. Sure, there were business goals to meet, and more engineering resources to utilise - but my brain was working in very much the same way it did at the university.…
Late in 2024, the team at Google Quantum AI released a free online course on quantum error correction taught by Austin Fowler. I checked it out expecting “more of the same”, but it blew my mind and I literally couldn’t put it down until it was done. Here is a belated review, and a discussion of what made it work.
My favourite application of new technologies is when they reduce the BS and let us spend more time doing what we love. In the recent weeks, I’ve been feeling exactly this about the recent wave of AI tools. I want to share the story of two side-projects of mine, and how LLMs made them successful.
The fundamental proposition of physics is that there are some fundamental laws from which all else can be derived. You would therefore think that once you know the fundamental laws of quantum mechanics - the Schroedinger equation and the Born rule - you are ready to answer any question about any quantum system. But is that really the case?
While this blog is primarily about quantum, I am keenly interested in any work pushing the boundaries of computation. Thus, when the folks at Normal Computing came up with [2308.05660] Thermodynamic Linear Algebra last year, it immediately grabbed my attention. The paper presents what seemingly looks like a rather straightforward recipe for a better analog computer to solve linear algebra…