HNHacker News
TopNewBestAskShowJobs

mwcampbell

11,185 karma · joined February 6, 2013

Note to PR folks: I'm not the InfoQ.com editor named Matt Campbell. Please don't email me about your upcoming press release. If anyone knows a better way to clear up this confusion, please contact me.

I'm a software developer specializing in accessibility for blind and visually impaired people. From mid-2017 to late 2020, I worked for Microsoft on the Windows accessibility team, though my comments were always my own opinion posted on my own initiative.

I'm legally blind myself; I have enough sight to read a desktop computer screen up close with slightly enlarged fonts, but I often use a screen reader when browsing the Web, doing email, and other tasks that don't involve code. (Don't get me wrong; totally blind people can program, but I never got used to doing it that way myself.)

Of course, there's more to me than visual impairment. When it comes to programming, I'm also particularly interested in the challenges of developing cross-platform applications that integrate smoothly with the host platform. Outside of programming, I love music and love to sing. Pre-COVID, I often did karaoke (of course, I memorize the words). And like many nerds, I read science fiction.

Email: mattcampbell@pobox.com

Mastodon: https://toot.cafe/@matt

Blog: http://mwcampbell.us/blog/

Location: Wichita, Kansas, US

[ my public key: https://keybase.io/mwcampbell; my proof: https://keybase.io/mwcampbell/sigs/9sbxkfOoTyW6C8IiDB50bP_ZWa-V73AfbdmvUlfOKe8 ]

submissionscomments
mwcampbell··on My AI skeptic friends are all nuts
To anyone else who was going to protest that the name _is_ at the top, apparently it depends on one's browser window width and zoom level. I was a bit surprised to find that increasing the zoom level even affects the order when reading with a screen reader.
mwcampbell··on My AI skeptic friends are all nuts
I think where tptacek is right, though, is that if we're going to hold this position without hypocrisy, then we need to respect copyright as long as it exists. He's right that many of us have not done that; it's been very common to violate copyright for mere entertainment. If we want the licenses of our own work to be respected, then we need to extend that respect to others as well, regardless of the size of the copyright holder.
mwcampbell··on Peer Programming with LLMs, for Senior+ Engineers
I don't deny that, under the right circumstances, these tools can produce results that feel indistinguishable from magic, or like science fiction as you put it. But I don't think it's worth the costs. To me, the two most concerning costs are the unreliability, and the massive amounts of stolen training data and underpaid labor (the RLHF process) required for these models. I'm not comfortable relying on a tool built on such foundations.

My bet, and I realize this might just be wishful thinking, is that the high order bit for being an effective software developer in the near future will be skill at using more reliable and non-exploitative automation tools, such as programming languages with powerful macro systems and other high-level abstractions, to stay competitive with developers who sling LLM-generated code. So I'd better get started developing that skill myself.

mwcampbell··on Peer Programming with LLMs, for Senior+ Engineers
How did we end up here, just accepting that some coding tasks require repetitive work, and turning to a probabilistic text synthesizer that requires massive training data to automate that? We're having this discussion on a site whose founder famously wrote, over 20 years ago, that succinctness is power, and even wrote this site in a programming language that he designed to take that principle as far as he could. Now, two decades later, why have we so completely retreated from that dream?

I bear some responsibility for this, since I was one of the people who basically said, in the 2010s, that we should just give up and use popular languages like JavaScript because they're popular. I regret that now.

mwcampbell··on The copilot delusion
> Your boss, stakeholders, and especially non-technical people literally give 0 fucks about "quality code" as long as it does what they want it to do.

We've all already heard this message a thousand times. There's no need to keep repeating it. Don't discourage people, especially young people, who want to do great work.

mwcampbell··on Litestream: Revamped
Do you have a citation on Turso moving away from Fly? I'd be particularly curious about anything Turso has written on _why_ they're moving away.
mwcampbell··on Edit is now open source
I wonder if you used GitHub Copilot or some other LLM-based code generation tool to write any of the code. If not, that's a lot of code to write from scratch while presumably under pressure to ship, and I'm impressed.
mwcampbell··on Stability by Design
As a contributor to the Rust GUI ecosystem [1], I've been affected (somewhat) by that churn and have also contributed to it. The best defense I can give for doing what might be considered refuckering is that when my library eventually hits 1.0, I want to meet the same standard of stability as Rust itself, and when that point comes, as @munificent said [2], I want my library to represent my best understanding of how to design an API for what I'm doing, with no legacy baggage. Hopefully I don't piss off current users between now and 1.0.

[1]: https://accesskit.dev/

[2]: https://news.ycombinator.com/item?id=43933068

mwcampbell··on Stability by Design
It's no wonder that newcomers to Win32 find it intimidating.
mwcampbell··on Mississippi Can't Possibly Have Good Schools
> We really need an alien invasion. Any Trisolarans listening?

Except we shouldn't even hate _them_, because they're sentient beings too. I guess we should only hate the worst aspects of our own nature.

mwcampbell··on As an experienced LLM user, I don't use generative LLMs often
I think it's natural for programmers to hold LLMs to a higher standard, because we're used to software being deterministic, and we aim to make it reliable.
mwcampbell··on All four major web browsers are about to lose 80% of their funding
> The codebase is deliberately under-architected (a legacy of the scars of working on Gecko), so many things you'd expect to be well defined re-usable modules that could be worked on independently of Google are actually leaky hacks

I'm guessing you're specifically referring to Gecko's early over-use of XPCOM, which the Gecko team itself had to clean up in a long process of deCOMtamination [1].

I'm hopeful that if Servo ever gets enough funding to become a serious contender among browser engines (hey, KHTML was once an underdog too), that it might walk a middle path between overuse of COM-ish ABIs and what you described about Chromium. Servo is already decomposed into many smaller Rust crates; this provides a pretty strong compile-time boundary between modules. Yet those modules are all statically linked, and in a release build, that combined program gets the full benefit of LTO. Of course, where trait objects are used, there's still dynamic dispatch via vtables, but the point is that one can get strong modularity without using something COM-ish everywhere.

Incidentally, the first time I built Chromium (or more specifically, CEF) from source in late 2012, I was impressed as I watched hundreds of static libraries being linked into one big binary. Then as I studied the code (though not deeply enough to learn the things you described), I saw that Chromium didn't use anything COM-ish internally (though CEF provided a COM-ish ABI on top). That striking contrast with Gecko's architecture (which I had previously worked with) stuck with me. I wonder how much the heavily reliance on static linking and LTO (meaning whole-program optimization), combined with the complete lack of COMtamination, contributed to Chrome's speed, which was one of its early selling points.

[1]: Mozilla used to have a dedicated wiki page about deCOMtamination, but I can no longer find it.

mwcampbell··on Migrating away from Rust
I'm not making an argument from grievance about my own code being plagiarized. I actually don't care if my own code is used without even the attribution required by the permissive licenses it's released under; I just want it to be used. I do also write proprietary code, but that's not in the training datasets, as far as I know. But the training datasets do include code under a variety of open-source licenses, both permissive and copyleft, and some of those developers do care how their code is used. We should respect that.

As for our tendency to disrespect the copyrights of art, clearly we've always been in the wrong about this, and we should respect the rights of artists. The fact that we've been in the wrong about this doesn't mean we should redouble the offense by also plagiarizing from other programmers.

And there is evidence that LLMs do plagiarize when generating code. I'll just list the most relevant citations from Baldur Bjarnason's book _The Intelligence Illusion_ (https://illusion.baldurbjarnason.com/), without quoting from that copyrighted work.

https://arxiv.org/abs/2202.07646

https://dl.acm.org/doi/10.1145/3447548.3467198

https://papers.nips.cc/paper/2020/hash/1e14bfe2714193e7af5ab...

mwcampbell··on Migrating away from Rust
> Why would I valorize discarding this kind of automation? Is this just a craft vs. production thing?

The strongest reason I can think of to discard this kind of automation, and do so proudly, is that it's effectively plagiarizing from all of the experts whose code was used in the training data set without their permission.

mwcampbell··on Prepper Disk
> These things are great lookup tools, much better than web search

Here's an article that argues otherwise, convincingly IMO:

https://buttondown.com/maiht3k/archive/information-literacy-...

mwcampbell··on Accessible open textbooks in math-heavy disciplines
A lecturer who is fluently multilingual might indeed smoothly switch accents when pronouncing foreign words. But it's still the same voice, and (if they're well practiced at it) they don't have to pause in mid-sentence to switch languages, as text-to-speech systems usually do. And eSpeak can switch languages while still being the same voice, since it's a rule-based, parametric synthesizer. But, at least with NVDA, a mid-sentence HTML span with a different lang attribute still causes a (short) break in the intonation on either side. That's too bad, because a multilingual parametric synthesizer like eSpeak could be like the ultimate polyglot speaker, impressing us all with how smoothly it switches languages.
mwcampbell··on How to distrust a CA without any certificate errors
> (As an erstwhile pentester, btw, fuck the OS certificate store; makes testing sites a colossal pain)

Can you please explain? I'm just curious, not arguing.

mwcampbell··on There isn't much point to HTTP/2 past the load balancer
This is why I configured my company's AWS application load balancer to disable HTTP2 when I first saw the linked post, and haven't changed that configuration since then. Unless we have definitive confirmation that all major load balancers have fixed these vulnerabilities, I'll keep HTTP2 disabled, unless I can figure out how to do HTTP2 between the LB and the backend.
mwcampbell··on File Pilot: A file explorer built for speed with a modern, robust interface
I'm curious, how did you conclude that AccessKit is a huge dependency? The binary size does vary by platform, with the Linux backend being especially large due to its use of a pure-Rust implementation of D-Bus. But the Windows x64 DLL is 268 KB, and the macOS x86-64 dylib is 437 KB stripped. The impact of the static library on your binary size will probably be about the same, or maybe a little less because most applications don't require all of the little setter and (especially) getter functions. I'm not happy with the size on Linux, but I think it's acceptable on Windows and macOS, given that accessibility is, in many contexts, a must-have.
mwcampbell··on File Pilot: A file explorer built for speed with a modern, robust interface
Nice work. I noticed, though, that there's no accessibility for blind people or others who need assistive technology that uses the platform accessibility API. If you can accept a Rust dependency, here's my project for solving this problem: https://accesskit.dev/ We do provide pre-compiled static libraries of the C binding, so you don't need to directly introduce Rust into your build process. I specifically kept immediate-mode toolkits in mind when designing AccessKit. Or if you'd rather just look at our code and implement the accessibility APIs yourself, that's cool too.
mwcampbell··on We were wrong about GPUs
> Apple's chips are dramatically faster than any other kind.

Any idea why? Is it because of some patent they hold?

mwcampbell··on We were wrong about GPUs
> we'll open a ticket with Apple and block the ticket in our system.

Wouldn't it be annoying to be blocked on Apple rather than shipping on your schedule?

mwcampbell··on Show HN: I made an open-source laptop from scratch
Not to take anything away from this amazing achievement, but if you want a similar open-source laptop, also based on an RK3588, that will actually be manufactured and sold, check out the MNT Reform Next: https://www.crowdsupply.com/mnt/mnt-reform-next
mwcampbell··on WASM GC isn't ready for realtime graphics
> This is good in rust

Is it actually going to be feasible to use WasmGC in Rust? I haven't yet found anything on this.

mwcampbell··on The Origins of Wokeness
> What gives Paul Graham away is what he doesn't mention and may be what bothers him more: the old heresies that the surprisingly innovative and even rebellious "conventional-minded" abolish.

Can you give an example of what you mean here?

mwcampbell··on Performance of LLMs on Advent of Code 2024
Steve Yegge wrote about this ~20 years ago, with a heaping helping of his usual snark directed at the OO fads of the time: https://sites.google.com/site/steveyegge2/age-racecar-driver
mwcampbell··on On the nature of computing science (1984)
What are the simpler, more fire-and-forget, approaches that you have in mind as alternatives to Kubernetes?
mwcampbell··on Tweaking Stunt Island's 30-year-old 3D engine
Is there still slow video RAM on machines with unified memory, like Apple Silicon Macs?
mwcampbell··on The Death of the Stubborn Developer
I still think @mpweiher's original response to the release of GitHub Copilot is spot on: https://blog.metaobject.com/2021/06/don-generate-glueextermi...

That is, rather than use an LLM to generate boilerplate code from an ambiguous natural-language prompt, use a high-level programming language to precisely, concisely, and directly express what you want.

mwcampbell··on Writing Tiny Desktop Apps in C and GTK3 (2022)
The best reason I can come up with to mimic Cocoa, rather than Win32, is that macOS users are the most... discerning about native versus non-native apps. So a Windows-first app on macOS will certainly be noticed, but that's not as true for a macOS-first app on Windows.
← PreviousPage 4 of 34Next →