Does this update your priors?

On 31 May 2025 James bet a tungsten cube that quantum computing is obvious bullshit. Eric answered that AI will accelerate quantum computing by inventing workarounds to its physical limits; James called that the most credible objection, and said a single huge leap in metallurgy or other physical engineering, with AI as a necessary input, would be an early sign that Eric was right. This page lists the AI results since then, each with its source, its caveat and Eric's rating against that early sign. Eric wins the cube if James changes his mind.

Below: the bet as it was made, where each side stands, then 17 results since the bet and 6 from before it.

The bet

#tungsten-betx.com · times UTC

2025-05-31 16:54<jamesob>Would bet a lot of money that quantum computing is obvious bullshit.
2025-05-31 17:11<jamesob>Except AI, to its credit, has actually delivered interesting results. But it's virtual, and the limitations in QC are (at least in part) physical, and we've basically forgotten how to do things in the physical world.
2025-05-31 18:34<ercwl>They’re interlinked. Are you suggesting you would bet me copious amounts of money that AI cannot invent solutions/workarounds to the physical limitations within 5 years? […]
2025-05-31 20:28<jamesob>This is the most internally consistent and credible objection. But - there is an easy way IMO to early detect if you're actually right: if I see even a single huge leap in metallurgy or some other kind of physical engineering for which AI was a necessary input, I'll refresh my priors. I very much doubt we'll see that come out of autocorrect++. Then whatever discoveries are made must actually be buildable and operatable by human construction for some huge number of physical qbits. But I think we'll see early signs well ahead in more primitive forms of engineering
2025-05-31 20:30<ercwl>Let’s do a 3 year bet. Tungsten cube? I hold your morals to such a high standard, I don’t think we need to specify further.
2025-05-31 20:36<jamesob>Done 🤝
2025-05-31 20:41<jamesob>Eric's track record with tungsten does not bode well for me, but what is life without skin in the game?
2025-05-31 20:43<ercwl>Oh, only one caveat. If you’ve already conceded quantum computing is not bullshit before AI-driven contributions tilt the scale, you still lose, and have to submit the cube at that point. I just don’t want to get into an endless argument about whether human/AI was key.
2025-05-31 20:48<jamesob>Of course - bet is locked in as of now, so if tomorrow I change my mind, I can't renege.
-!- bet locked, with Eric's caveat that James still pays if he concedes for reasons other than AI
2025-05-31 20:58<jamesob>Yes you're probably right, and I'll defer to Eric on terms here because I don't mind losing, but writing the tweet I sort of had in mind a situation like "we built x new skyscraper that either wasn't possible to build or we did it for half the cost etc. given AI's discovery of a new material" In other words, something that surfaces into the real world in a big way.
2025-06-02 21:39<ercwl>Building a whole new skyscraper with AI-invented materials in 3 years, yeah, not betting my tungsten cube collection on that.
2025-06-02 22:47<jamesob>We could still try to put terms around "huge leap in physical engineering with AI as a necessary input" if you want! But I was definitely thinking of physical things that are along (or closely related to) the critical path of quantum, not e.g. some ACTG pattern hacking that results in some new pharma research paper.
2025-06-02 22:57<ercwl>Hm, yeah, sounds like a longer convo. AI-driven discoveries inferring huge leaps in metallurgy/other kinds of physical engineering, I’d bet a cube on. Whether those leaps will be along the critical path of quantum within 3 yrs, I don’t rule out, but wouldn’t bet a cube on.
-!- not agreed: Eric would not bet the cube on a skyscraper or on the quantum critical path within three years
2025-07-05 15:50<ercwl>Hey James, I’ve come up a way to simplify this bet. > Would bet a lot of money that quantum computing is obvious bullshit. I bet that within 3 years, you’ll change your mind, and you now believe that quantum computing is not obvious bullshit. Works?
2025-07-05 18:49<jamesob>Yes, let's do it
-!- terms simplified: Eric wins if James changes his mind about quantum computing within three years
2026-02-14 04:20<ercwl>”…if I see even a single huge leap in metallurgy or some other kind of physical engineering for which AI was a necessary input, I'll refresh my priors. I very much doubt we'll see that come out of autocorrect++…” [quoting a post on a GPT-5.2 result in quantum field theory, item P6 below]
2026-02-14 04:23<jamesob>> string theory > quantum field theory I rest my case!

status: James had not changed his mind as of 2 May 2026. On 24 April he wrote of quantum computing for the bullshit that it is [x.com]. The deadline is 2028-05-31, or 2028-07-05 if counted from the July message.

Where each side stands

What each of them said on 31 May 2025, with Eric's verdict so far and the results since the bet that it rests on. The names link to those results further down.

James's test, step by step

Each step includes the ones above it. Pick a step to keep only the results that reach it; the rest are dimmed further down. Counts are for results since the bet, and the ratings are Eric's.

TEST.EXEsince 31 May 2025

Physical-relevance index, by month

The index is 2.5 × P × H × W × C × (1 + log₁₀ n), where P is physicality 0–10, H the horizon (now 1.0, years 0.7, decades 0.4, someday 0.15), W the weight 1–4, C confidence 0–1, and n = 1 for every item here. When a step is picked, only the results that reach it are plotted, and months with none sit at zero. The y-axis keeps the same maximum whatever the filter.

Footnote: the Navier–Stokes claim

claimed
caveat