An ace software engineer is not an ace because of tooling.

It's not the plane, it's the pilot, or something like that.

Experiments already running. Write up:

"Autoresearch on DGX Spark: Error Bars, GPU Pools, and the Case for an Autonomous Research Swarm": https://github.com/matt-langston/autoresearch/discussions/1

My fork of Andrej's repo if you want to join in: https://github.com/matt-langston/autoresearch

The big takeaway isn't reverse engineering the ANE per se, but what Manjeet could do with his software engineering skills when accelerated by AI.

This is a good example of the present state of software engineering. Not future state - present state.

Comments URL: https://news.ycombinator.com/item?id=47010792

Points: 2

# Comments: 2

That's where the 4.5->4.6 jump hit me hardest - not routine tasks but problems where I need the model to reason about stuff it hasn't seen. It still fails, but it went from confidently wrong to productively wrong, if that makes sense. I can actually steer it now.

The cerebellum analogy resonates. I'd go further - it's becoming something I think out loud with, which is changing how I approach problems, not just how fast I solve them.

One data point for this thread: the jump from Opus 4.5 to 4.6 is not linear. The minor version number is misleading. In my daily work the capability difference is the largest single-model jump I've experienced, and I don't say that casually — I spent my career making precision measurements.

I keep telling myself I should systematically evaluate GPT-5.3 Codex and the other frontier models. But Opus is so productive now that I can't justify the time. That velocity of entrenchment is itself a signal, and I think it quietly supports the author's thesis.

I'm not a doomer — I'm an optimist about what prepared individuals and communities can do with this. But I shared this article with family and walked them through it in detail before I ever saw it on HN. That should tell you something about where I think we are.

The Standard Model predicts a specific value for the weak mixing angle, which determines the electron's vector coupling. My measurement at SLAC, along with other SLD measurements, consistently preferred a slightly different value than what LEP (the European competitor experiment) found using a different technique.

The key word there is "different technique." SLD used a polarized beam of electrons — a completely novel approach at the time — which gave us direct access to the left-right asymmetry without needing to untangle final-state effects. LEP extracted the same parameter from b-quark forward-backward asymmetry. Two fundamentally different methods probing the same physics, with different systematic exposures, giving different answers.

Both experiments had good resolution. We spent enormous effort characterizing the systematics, and they're small compared to the statistical uncertainty. But the two most precise determinations of this parameter disagreed at roughly the 3-sigma level — and that disagreement has never been explained. The world average splits the difference, and the Standard Model prediction is consistent with that average, so you could say "the SM is fine" if you squint. But nobody knows why the two experiments don't agree with each other.

It could be an unidentified systematic error in one experiment. It could be that something beyond the Standard Model is subtly shifting one measurement and not the other. That ambiguity is exactly what makes it a "dangling thread" rather than a resolved question.

Many of those models naturally include a dark matter candidate. I didn't mean to imply 'we found dark matter' — it's that the theories which could explain the discrepancy often come with one attached.

I measured the electron's vector coupling to the Z boson at SLAC in the late 1990s, and the answer from that measurement is: we don't know yet - and that's the point.

Thirty years later, the discrepancy between my experiment and LEP's hasn't been resolved.

It might be nothing. It might be the first whisper of dark matter or a new force. And the only way to find out is to build the next machine. That's not 'dead', that's science being hard.

My measurement is a thread that's been dangling for decades, waiting to be pulled.

"Differentiate y(x) w.r.t x, gamma and beta."

It not only produced the correct result, but it understood the context - I didn't tell it the context was layer norm, back-propagation and matrices.

This release is a step function for my use cases.

My screenshot came from here: https://docs.pytorch.org/docs/stable/generated/torch.nn.Laye...

Comments URL: https://news.ycombinator.com/item?id=44780352

Points: 2

# Comments: 0

1) the weak mixing angle

2) the three particle generations and the ratio of their masses

This is remarkable to me.

Apropos for current times.