HNHacker News
TopNewBestAskShowJobs

William_BB

166 karma · joined September 28, 2024

submissionscomments
William_BB··on Queueing to publish in AI and CS
NeurIPS scoring system is inherently subjective. People will have wildly different interpretations of, say, 3 vs 4, or 4 vs 5. You can get lucky and draw only reviewers that, on average, "overrate" papers in their batch. The opposite can happen too, obviously. 4444 vs 3444 is just noise.
William_BB··on Queueing to publish in AI and CS
There's two main problems. First, the field is too saturated -- too many people publishing papers. Second, the reviewers are the same people that publish papers -- it's a zero sum game (regardless of acceptance criteria).

Submission numbers this year have been absolutely crazy. I honestly don't think it can be solved.

William_BB··on The Theatre of Pull Requests and Code Review
Respectfully, I disagree. I understand people have vastly different experiences and preferences. In my ideal world, a PR is a unit of work that has an in-state and an out-state. I don't have to see an initial commit within a PR with a full fledged spec and then wonder if any future commits within the same PR overrode those changes. A PR will rarely be clean from the start.

The way it appears to me, if there's multiple commits submitted as separate PRs, then maybe the PR wasn't so atomic to begin with.

William_BB··on The Theatre of Pull Requests and Code Review
PR is the atomic level of work. I'd argue PR-level history (i.e. squash) is often enough and is way cleaner. Why would I care about "commit A", "change parts of A because I misunderstood a requirement", "improve A based on code review" etc.?

If I want that granularity, I'd go read the original PR and the discussion that took place.

William_BB··on Processing Strings 109x Faster Than Nvidia on H100
Could you elaborate on SVE and SVE2? Is that because it's only 128 bits? I think my Macbook (Apple silicon) is one of the two
William_BB··on Oxford loses top 3 university ranking in the UK
Local UK rankings have always been BS. The only rankings that really matter are international ones. Even then, I'm a strong believer that the best you can do when ranking universities is ranking them in tiers. The ranking within the tier is useless.

Oxbridge will always be tier 1. Undergraduate teaching quality + a rigorous selection process cannot be beat.

William_BB··on Tracking trust with Rust in the kernel
Out of curiosity, how would one go about implementing this in C++?
William_BB··on Microsoft is officially sending employees back to the office
I'm sorry, but this is a meaningless stat when evaluating remote work productivity.

Microsoft is a huge company with a lot of momentum and so many products that you can't reasonably conclude anything about remote work from this stat alone. Worse, you don't even account for inflation (!).

What about revenue per employee before covid? What if I chose any arbitrary time period in the last 20 years? Lastly, maybe revenue per employee could've grown significantly more during 2020-2024 if everyone was in the office?

William_BB··on You don't want to hire "the best engineers"
This is so interesting to hear. From what I've seen, probably half of recent yc startups have founders below 30. I wonder how senior talent views being interviewed by people who are essentially junior/mid developers.

I'm in my 20s with good credentials and have quite a few friends in the startup world. I would never feel comfortable interviewing someone with 10+ years of industry experience.

William_BB··on Why tail-recursive functions are loops
I studied FP in college. I currently work on a large low latency C++ codebase. I honestly have no idea how I'd use pure functional concepts everywhere in this context. I'm also probably doing it wrong.
William_BB··on NautilusTrader: Open-source algorithmic trading platform
Interesting stuff. I'm a bit confused on several points though.

If your strategies care so much about performance, can you really achieve the same or comparable performance with Cython? At that point, wouldn't those strategies be better suited for a custom "re-implementation"?

I suppose I'm not exactly sure how "high performance" and "low latency" this project is. Are there any latency stats on this project?

William_BB··on C++26 Reflections adventures and compile-time UML
You sound like you subscribe to "Orthodox C++".

Speaking seriously, I agree there's definitely a lot of bloat in the new C++ standards. E.g. I'm not a fan of the C++26 linalg stuff. But most performance-focused trading firms still use the latest standard with the latest compiler. Just a small example of new C++ features that are used every day in those firms:

Smart pointers (C++11), Constexpr and consteval (all improvements since C++11), Concepts (C++20), Spans (C++20), Optional (C++17), String views (C++17)

William_BB··on Does the Bitter Lesson Have Limits?
Interesting. Are these models the SOTA in the options trading industry (e.g. MM) nowadays?
William_BB··on Enough AI copilots, we need AI HUDs
The original example was (paraphrasing) "rerunning 10-100 tests that take 1ms after each keystroke".

Even with incremental builds, that surely does not sound plausible? I only mentioned C++ because that's my main working language, but this wouldn't sound reasonable for Rust either, no?

William_BB··on Enough AI copilots, we need AI HUDs
There's no way this would work for any serious C++ codebase. Compile times alone make this impossible

I'm also not sure how LLM could guess what the tests should be without having written all of the code, e.g. imagine writing code for a new data structure

William_BB··on Use Your Type System
What do you think about this but in C++ (e.g. with explicit constructors)? Has anyone had any experience with it? Did it succeed or fail?
William_BB··on Mathematics for Computer Science (2024)
In my experience, coursera/khan academy courses have never been able to compete with a rigorous university course. They're great resources when you need alternative explanations, but never stood up on their own.

I think long lecture playlist is a feature, not a bug. It's much harder to commit to such material when you're not full timing education.

William_BB··on My experience with Claude Code after two weeks of adventures
The reason is probably complexity and the task at hand.

In my experience, LLMs are great at small tasks (bash or python scripts); good at simple CRUD stuff (js, ts, html, css, python); good at prototyping; good at documentation; okay at writing unit tests; okay at adding simple features in more complex databases;

Anything more complex and I find it pretty much unusable, even with Claude 4. More complex C++ codebases; more niche libraries; ML, CV, more mathsy domains that require reasoning.

William_BB··on The messy reality of SIMD (vector) functions
Thank you so much, this explains it well. I was initially afraid that the dispatch would be costly, but from what I understand it's (almost) zero cost after the first call.

I only code for x86 with vectorclass library, so I never had to worry about portability. In practice, is it really possible to write generic SIMD code like the example using Highway? Or could you often find optimization opportunities if you targeted a particular architecture?

William_BB··on The messy reality of SIMD (vector) functions
Could you elaborate on the dynamic dispatching features a bit more? Is that for portability?
William_BB··on Demonstration of Algorithmic Quantum Speedup for an Abelian Hidden Subgroup
It's mathematically sound and the quantum primitives it uses are well understood.

The limiting factor in practice, as with everything quantum, is noise. You are right -- we don't know for certain until it's implemented. I suppose it's part of a bigger question: whether quantum computing will work at all. My knowledge of quantum hardware is limited, so I can't really comment more on this.

William_BB··on Demonstration of Algorithmic Quantum Speedup for an Abelian Hidden Subgroup
As the poster above mentioned, it's widely accepted that Shor works. We simply don't have hardware to run the full version.

The quantum papers on "factorization as optimization" are borderline scams though. I wouldn't put those papers in the same sentence as Shor.

William_BB··on Demonstration of Algorithmic Quantum Speedup for an Abelian Hidden Subgroup
A standard undergraduate quantum computing course should suffice. Most of them would follow Nielsen and Chuang's book.
William_BB··on What Problems to Solve (1966)
It depends on what "working on those problems" means to you. If you want to work on those problems as a software engineer, that sounds like an achievable goal.

To me, the interesting, fulfilling bits of building the next Google/ChatGPT/AI/climate change lie in the theory. Arguably with the exception of Kubernetes, this theory does not come from software engineering. As much as I enjoy software engineering, it's a trade. It's a tool to get the job done. And recently, I realized I like building things just as much as I like "the theory".

To me, that was a bitter pill to swallow. I'm not an ML engineer, but I suspect this is also the reason why you can find so many posts about ML engineers trying to pivot to ML scientist roles.

William_BB··on Andrej Karpathy: Software in the era of AI [video]
[flagged]
William_BB··on Quantum Computation Lecture Notes (2022)
I meant practically usable NISQ
William_BB··on Quantum Computation Lecture Notes (2022)
"When quantum happens in a few years" -- do you mean NISQ (i.e. VQA, QAOA) or actual fault tolerant quantum computing that can run Shor?
William_BB··on Modern C++ – RAII
I take the following approach: - Stack by default - Unique ptr if needed on the heap - Shared ptr if needed to share ownership

Although unique ptr is zero cost after make_unique(), I avoid polluting my heap unnecessarily. I've never benchmarked this though (keeping various objects on stack vs heap as unique ptr and how that impacts memory accesses)

I'm quite junior. Appreciate anyone pointing out if anything I said doesn't make sense.

William_BB··on Push Ifs Up and Fors Down
My first thought was conditional branches inside the for loop based on the element as well. By any chance, do you know how hard it is for compilers to auto-vectorize something like this? I am generally not sure where the boundary is.
William_BB··on Initialization in C++ is bonkers (2017)
The standard says that "Copy elision is <...> one of the two allowed forms of optimization, alongside allocation elision and extension,(since C++14) that can change observable side-effects"

I agree you have a valid point though. I'd be interested to know the committee's reasoning.

← PreviousPage 3 of 4Next →