HNHacker News
TopNewBestAskShowJobs

ngriffiths

336 karma · joined October 16, 2024

Twitter: @ng_griff

Website: griffens.net

submissionscomments
ngriffiths··on A sufficiently detailed spec is code
I think in an ideal world, programming languages[1] would allow you to express key behaviors, interfaces, states, etc very precisely and sharpen your own understanding while writing, but importantly, also leave unimportant behaviors, structures, etc unspecified, using reasonable defaults and similar stuff.

I sort of see the success of coding agents as not just a sign that people get tricked by slop, but mainly a sign that today's languages and frameworks just aren't that good. And now it's easier to say to the LLM, "I literally don't care how the UI is organized as long as everything is visible" than to try to express that in code.

[1]: I mean this in the vaguest sense, so including DSLs/frameworks/interfaces offered by libraries too

ngriffiths··on Most-read tech publications have lost over half their Google traffic since 2024
Raises a few questions:

1. Is there still value in a trusted source of high quality, relevant information existing in an easily accessible place? To me the answer is obviously yes, whether a human writes it or an agent, and also whether a human consumes it or an agent

2. If an agent visits a page, they shouldn't generate ad revenue (I guess they don't? is that true?). Should they just be able to copy the information for free? If the answer to 1. is yes, the answer here should probably be no.

3. Does this completely break the ad revenue model that powers like the whole internet?

4. Does this seriously threaten our ability to maintain a high quality, shared collection of knowledge? What can be done?

ngriffiths··on Windows 11 Notepad to support Markdown
A lot of comments about how this is another case of useless bloat. I don't know, markdown is just incredibly useful and widespread and yet it is pretty annoying to find a good editor:

- There wasn't anything that comes with Windows that natively supports it (before now)

- All your favorite text editors don't support it natively, and plugins vary

- You can pay for a nice markdown editor but for some reason your more powerful usual text editor is still free?

- You can open VSCode, which is hilarious overkill if you just want to take some notes. Obsidian is excellent but same problem.

- Maybe something I'm missing?

Basically I think it is a great thing if I just get a lightweight markdown friendly editor built in, because I'll probably use it all the time.

...except if it immediately leads to a CVE, I guess.

ngriffiths··on What is happening to writing? Cognitive debt, Claude Code, the space around AI
> Additionally, not all writing serves the same purpose.

I think this is a really important point and to add on, there is a lot of writing that is really good, but only in a way that a niche audience can appreciate. Today's AI can basically compete with the low quality stuff that makes up most of social media, it can't really compete with higher quality stuff targeted to a general audience, and it's still nowhere close to some more niche classics.

An interesting thought experiment is whether it's possible that AI tools could write a novel that's better than War and Peace. A quick google shows a lot of (poorly written) articles about how "AI is just a machine, so it can never be creative," which strikes me as a weak argument way too focused on a physical detail instead of the result. War and Peace and/or other great novels are certainly in the training set of some or all models, and there is some real consensus about which ones are great, not just random subjective opinions.

I kind of think... there is still something fundamental that would get in the way, but that it is still totally achievable to overcome that some day? I don't think it's impossible for an AI to be creative in a humanlike way, they don't seem optimized for it because they are completely optimized for the sort of analytical mode of reading and writing, not the creative/immersive one.

ngriffiths··on Drug trio found to block tumour resistance in pancreatic cancer in mouse models
I don't think the biology is there, let alone consensus on the major ethical questions involved
ngriffiths··on Drug trio found to block tumour resistance in pancreatic cancer in mouse models
Because research on real humans and real diseases is exceptionally difficult. Clinical research is notoriously expensive, results are likely to differ from non-human (preclinical) models, and trials take forever to get started, gather enough data, and get a drug actually reviewed and approved. So even when everyone is excited by the preclinical data, there are so many barriers (both scientific and non-scientific) that getting to an approved drug is pretty unlikely.
ngriffiths··on Drug trio found to block tumour resistance in pancreatic cancer in mouse models
IN MICE. (To be fair, also IN SOME OTHER BETTER MICE).

https://jamesheathers.medium.com/in-mice-explained-77b61b598...

(mostly a joke, but I'd be in favor of adding context to the HN headline if possible)

ngriffiths··on How to Attend Meetings
One of the costs of saying no to meetings is that going to other people's (useless) meetings is a super low effort way to say "I value our working relationship." Not going often explicitly sends the opposite message.

Sometimes there is a whole set of rituals used to "prove" you actually care about the group, and the rituals only ever happen in meetings, and you cannot change them without bothering a lot of people.

ngriffiths··on Roblox is a problem but it's a symptom of something worse
Well said, my reaction was basically, what he is saying is not really for a listener like me.

> AND also was not interested in the smooth polish of PR-style transitions that would give an appearance of basic cooperation..

I agree this is a big part of it and would add that the "unhinged" look is probably just a lack of PR skill. Both sides are hostile but the interviewers "win" here by staying within the rules of the game, and they also do a beautiful job of sort of winking at the audience like "wow this dude is crazy right?"

It's impressive but sorta annoying, I'd rather listen to actual content.

ngriffiths··on Roblox is a problem but it's a symptom of something worse
Am I the only one who was really underwhelmed? I saw that it was supposedly a very tense trainwreck situation and sure, it gets sarcastic and stuff, but most of it was

Interviewer: "so I heard you were/are doing a bad job with moderation"

CEO: repeats banal PR talking point for the 10th time

Repeat.

I mean, at no point did the CEO say anything interesting about the moderation problem or what they are doing. The interviewers seem too skeptical to be genuinely interested. He explains to them that cost =/= quality and that 2016 =/= 2025 for what feels like an eternity. I was bored.

ngriffiths··on Our investigation into the suspicious pressure on Archive.today
Fair enough. It was something that impressed me when I first read about it. You hear about disputes between Apple and the FBI over unlocking phones and meanwhile banks are like "and over here are whole floors of analysts tracking suspicious stuff." I definitely agree there are downsides and not everyone is happy about the floors of analysts, but I do think they are very far away from the Overton window.
ngriffiths··on Our investigation into the suspicious pressure on Archive.today
> a private company shouldn’t have to decide what counts as “illegal” content under threat of legal action.

Immediately reminded me of patio11's amazing write up[1] of debanking, featuring banks being deputized as law enforcement for financial crimes (which is completely non controversial), and even used as a convenient tool to regulate other industries that the white house didn't like (kinda controversial).

[1]: https://www.bitsaboutmoney.com/archive/debanking-and-debunki...

ngriffiths··on How I fell in love with Erlang
I think it makes more sense in languages where you use the "let" keyword. Then it sounds like assignment, though you still have to get comfortable with the X being on the right side too.
ngriffiths··on The Silent Scientist: When Software Research Fails to Reach Its Audience
Not all engineers are in the target audience, and not all details of research findings need to be conveyed to the target audience to make a real impact. The point is if no findings ever make it to engineers (in the broadest sense), there is zero real world impact. I guess real impact is not the only goal but it's a valid one.
ngriffiths··on The Silent Scientist: When Software Research Fails to Reach Its Audience
> active science communication has been sparse in the area of software research, and those who have tried often find their efforts unrewarded or unsuccessful.

The authors suggest:

> Identify your target audience to tailor your message! Use diverse communication channels beyond papers, and actively engage with practitioners to foster dialogue rather than broadcasting information!

What I would emphasize is that many researchers just don't know how to do it. It isn't as simple as just thinking up a target audience and churning out a blog post. If you are the median researcher, ~0 people will read that post!

I think people underestimate:

- How hard it is to find the right target audience - How hard it is to understand the target audience's language - How hard it is to persuade the target reader that this work you've done should matter even a little to their work, even when you designed it specifically for them - How few people in the audience will ever understand your work well - How narrow your target audience should be

I also think many researchers want to be able to, if not as a primary career goal then at least as a fulfilling, public service type activity. Currently testing this out a bit (more: https://griffens.net).

ngriffiths··on What if hard work felt easier?
There's a certain kind of work that's productive and doesn't take much "effort," at least not in the sense of that super annoying voice in your head that says "I really should focus... I didn't get much done yesterday, gotta make up for it now! Other people are working, look at how smart and productive they are! I need coffee, just one coffee then I'll have enough energy to do this."

Used to happen all the time playing piano in a concert or just an intense practice session. I'd notice that I had been incredibly focused, shockingly focused, like my brain had decided this random sound making activity was its One True Purpose and if it didn't work hard enough it would probably shrivel up and die. And thinking back it hadn't felt strenuous, it felt weirdly calm in the moment, but I think "ease" is totally the wrong word here. It was obvious that I had been working extremely hard, it was as intense as it was calming (? somehow?), and after I'd be totally exhausted.

I think the article really means "what if hard work didn't feel incredibly annoying at the same time?"

On the other hand there's "what if work that feels relaxing isn't also nearly useless"? Because there's a great way to make hard work feel easy, which is to just stop worrying, man. Take it one step at a time, it's probably the right step. There's no rush, slow down, you'll probably get there at the same time as you would by stressing yourself out and stopping for more coffee all the time. Here, use this easy framework...

ngriffiths··on An illustrated introduction to linear algebra
I feel like it's obligatory to also drop a link to the 3blue1brown series on linear algebra, for anyone interested in learning - it is a step up from what's in this post, but these videos are brilliant and still super accessible:

https://youtube.com/playlist?list=PLZHQObOWTQDPD3MizzM2xVFit...

ngriffiths··on Asked to do something illegal at work? Here's what these software engineers did
It's funny how people's mindset shifts when they work for a company versus independent contracting, but this is another case (like salary negotiation) where it should be the same. If your client asked you to do something illegal you'd be thinking about liability and drop the client. If your employer does, you should realize not only do they not have to protect you, it's probably in their interest to turn around and blame it all on you!
ngriffiths··on The Lambda Calculus (2023)
I wouldn't be surprised if there were a much more clear and concise way to introduce it, but on the other hand, the topic doesn't seem all that useful to someone without exposure to some fundamentals, so it seems fine to me if it has a bunch of jargon. Lambda calculus programs that actually do something useful are extremely weird and hard to read and write, even very simple ones, so that's not where its value comes from. Probably a better thing to learn in connection with stuff like interpreter and compiler design when it isn't weird to say "function application" and expect you to know what that is.
ngriffiths··on 30 minutes with a stranger
Fair, but it's still interesting that people underestimate how much they will enjoy the conversation, even if it's with a random person on a bus who did not volunteer for an experiment.
ngriffiths··on A Random Walk in 10 Dimensions (2021)
> Therefore, despite the insanely large number of adjustable parameters, general solutions, that are meaningful and predictive, can be found by adding random walks around the objective landscape as a partial strategy in combination with gradient descent.

Are there methods that specifically apply this idea?

I guess this is a good explanation for why deep learning isn't just automatically impossible, because if local minima were everywhere then it would be impossible. But on the other hand, usually the goal isn't to add more and more parameters, it's to add just enough so that common features can be identified but not enough to "memorize the dataset." And to design an architecture that is flexible enough but is still quite restricted, and can't represent any function. And of course in many cases (especially when there's less data) it makes sense to manually design transformations from the high dimensional space to a lower dimensional one that contains less noise and can be modeled more easily.

The article feels connected to the manifold hypothesis, where the function we're modeling has some projection into a low dimensional space, making it possible to model. I could imagine a similar thing where if a potential function has lots of ridges, you can "glue it together" so all the level sets line up, and that corresponds with some lower dimensional optimization problem that's easier to solve. Really interesting and I found it super clearly written.

ngriffiths··on Simulating and Visualising the Central Limit Theorem
Honestly, it might partly be that I've used R somewhat irregularly and I put a lot of value in design choices that "make sense" and are easier to remember. I'm sure once you are intimately familiar with the whole base language you can be really happy and productive with it.

> I think the problem is the lack of tutorials that explain how to use all the data manipulation tools effectively, because there are quite a lot of functions and it isn't easy to figure out how to use them together to accomplish practical things.

Most languages solve this problem by not cramming quite a lot of functions in one package and using shared design concepts to make it easier to fit them together. I don't think tutorials would solve these problems effectively but I guess it makes sense that they affect newer users the most.

> Tidyverse may be consistent with itself, but it's inconsistent with everything else.

Yeah, totally agree and I really dislike this part.

ngriffiths··on Simulating and Visualising the Central Limit Theorem
For me the appeal is less that tidyverse is great and more that the R standard library is horrible. It's full of esoteric names, inconsistent use and order of parameters, unreasonable default behavior, behavior that surprises you coming from other programming experience. It's all in a couple massive packages instead of broken up into manageable pieces.

Tidyverse is imperfect and it feels heavy-handed and awkward to replace all the major standard library functions, but Tidyverse stuff is way more ergonomic.

ngriffiths··on Simulating and Visualising the Central Limit Theorem
I was definitely expecting you'd need a higher sample size for the Q-Q plots to start looking good. All the points in other comments about drawbacks or poorly behaved distributions are well taken, and this is nothing new, but wow it really does work well.
ngriffiths··on Meta's AI rules let bots hold sensual chats with kids, offer false medical info
Reminds me of a conversation I had recently about the alcohol industry. It's not so bad if your local crappy bar markets to local underage college kids. But when sketchy tactics exist and are allowed at the scale of the biggest companies in the world, you've got problems.

Actually, sketchy tech/social media/AI tactics towards youth are more comparable to "lets get kids addicted so they become lifelong customers" than I ever realized before.

ngriffiths··on iPhone DevOps (2023)
Love this. Sometimes being able to work on a phone is perfect for things that need to percolate a little while you walk around or do something mindless at the same time. I put some effort into finding a good writing and markdown rendering app but didn't realize a whole IDE is possible too.
ngriffiths··on Job-seekers are dodging AI interviewers
Personally, my feeling is that the median job application process was always this absurd, and the difference is only that the AI interviewer 1) is more obviously absurd, and 2) helps subject only the candidate to this huge waste of time, saving the interviewer from it. Which I guess also shows some disrespect for the candidate's time.

Job interviews are bad. It was always necessary to proactively seek out opportunities, network, and not just send lots of job apps... and obviously here I have to quote Patrick McKenzie who calls cold job apps "an effective strategy for job searching if you enjoy alternating bouts of being unemployed, being poorly compensated, and then treated like a disposable peon."[1]

Many people are super desperate and short on time, though, and in that case I guess you should suck it up and do it unless you think the time is more efficiently spent elsewhere. Which it totally may be, if there are tons of people getting "interviewed" and your chances are low.

[1]: https://www.kalzumeus.com/2012/01/23/salary-negotiation/

ngriffiths··on The HTML Hobbyist (2022)
> But what we can do is be the change we want. Just make my own little oasis. Find other oases and hook ‘em all up.

Perfectly said and I think this is applicable far beyond websites.

ngriffiths··on Tour de France confronts a new threat: Are cyclists using tiny motors?
Cool! Seems to be citing this nice post here: http://calmerthanyouare.org/2015/03/19/betteridges-law.html

I didn't realize the biggest reason against it is that the majority of headline questions aren't yes/no.

Oh well, it's still funny to silently answer any (yes/no) headline question with "no!" before reading.

ngriffiths··on Tour de France confronts a new threat: Are cyclists using tiny motors?
These cheating methods always seem far-fetched until I remember that getting that good at cycling is pretty far-fetched, and it's all relative to that
← PreviousPage 3 of 5Next →