I'm a seeing-eye dog for a computer
claytonwramsey.com
claytonwramsey.com
What I have found super useful actually is having models make a interactive 3d viewer in which I use move / highlight / paint (soft body painting directly onto the geometry for issues and different colors mean different failures). This gives a much better way to communicate the physical relationships and positions that are hard to get across in a labeled screenshot.
Its for sure still a lot of manual work so the "seeing-eye dog" description definitely holds. But I have found that after a couple of examples with the extra context the model gets much better at handling the problem and becomes useful.
He realises the current woe of things but doesn't realise he's been arguing with bot farms and teams of people hired just for this reason - to sow doom, arguments and engagement.
Around 10 years ago I noticed this happening on trending topics of Twitter - it wasn't that the opponents were stupid because they disagreed, it was that they were simultaneously intelligent and stupid in the way they spoke in a way that I realised I'd never seen in genuine people, making me realise it was probably different people or bots under one account.
If I realised it 10 years ago it was probably happening for at least 15. He wasn't better than these people he was falling into their trap
I can also highly encourage people to keep notes and do some basic OSINT. The effective internet is smaller than one might think, so that proves useful time and time again.
If something seems _odd_, I suggest looking up _why_ it might be odd. Gathering context, essentially.
Like "Okay, this guy is weird. Aah, okay, LinkedIn says that he works there. Okay _now_ that makes sense".
Based on that, you can then decide how to approach the (previously failing) interaction.
__
HN is a very easy place for that, because, to give somewhat concrete examples, if someone is shilling for something and seems unreachable for common sense, LinkedIn usually tells you that their salary depends on that.
And people play very open here, because this is treated as a business networking event.
Everything else that I care about is in private Signal groups or an equivalent of same with anywhere from 15 to 200 people, and adding any new member means they need to be vouched for by an existing one. And at least 55-75% of the people know each other face to face through industry vertical specific events, trade shows, conferences and similar.
I can't even imagine trying to engage with people on something like x/twitter or other social media. Not knowing whether the 'person' you're replying to is somebody's LLM authoried bot farm.
Tbh, my experience with _that_ has been worse than with a random sample of internet background radiation, as people seem to have no idea how to filter correctly and who to vouch for and why.
_Especially_ with people they know via work, conferences, etc. Spaces that do not filter for "genuinely relevant contributer to a conversation" at all, but encourage faking that.
At a certain scale you know who the people are that have super-admin privileges at very specific ASNs and they are well known in the industry. Basically much as somebody in academia in a journalism school on the west coast would know who is the department head of the school of journalism at a major university in Boston or NY and recognize their name.
Now within a Signal group of 150 people adjacent to the neteng teams at large to medium sized US/Canadian ISPs, it's entirely possible that somebody let an uninvited rando in, but we don't discuss anything super sensitive (that's for more 1:1 messages). A person who has been lurking for weeks or months and suddenly starts chatting in an inauthentic way would be found out very quickly.
One of the things that ISP technical operations groups do very well is filter for non-authentic participant in something, because sales persons getting into groups of technical discussion with the angle of selling some new thing is a perpetual problem and is a well known factor to deal with.
Also sales people generally don't do well when a complex operational-related technical matter is put in front of them in a fast paced environment where they can't engage their engineering team to get an accurate reply. Ask anyone that's ever tried to get their direct cellphone number removed from a Cogent IP transit sales person's CRM system....
I know the kind of space you are referring to. I'm in those too. But my point is that they regularly disappoint me.
Which is usually worse disappointment than with randoms on the internet, because the people should know better but don't.
With these tight knit spaces you get all the social dysfunctions of people liking each other not for merit but because they've been around each other for long, and that's - for me personally - worse than the public alternative, as you suddenly do not argue with logic, but with the tribe and social cohesion.
Which might be a winning strategy, but also might not be.
I have a theory that a good part of the filtering for authentic humans has already been done here, which gives me a higher level of confidence that I'm not metaphorically shouting into the void by conversing with somebody's LLM or social media influence apparatus.
My mental health improved significantly. I actually feel valuable at work. I feel pride in contributing to my local community. And the paycheck isn't actually worse.
One of the great things about not having ads on rangerovers.pub is that Google is just not interested in indexing it, because it doesn't make them any money. It shows up here and there in search results, sporadically, especially if you know a username or phrase specific to the site, but other than that if you want to know about it you need to be told about it.
I was never a fan of web forums but they are a nice compromise between really oldschool mailing lists and FAANG content-to-eyeball transformers.
I guess there's at least 3 or more active and currently developed software packages with open source licenses, the flaskbb that site uses, simplemachinesforum, the classic phpbb, probably some others I've forgotten about... The LAMP stack dependencies for all seem very mundane and very lightweight.
I'd love to write a backend for it to expose the forum as NNTP ;-)
On a site like HN, blocklists are very effective at keeping people off your feed.
Not sure why that makes any difference - I've had plenty of arguments on Facebook with people who are perfectly happy to spew racist and/or conspiratorial bullshit while having their full names, their holiday photos and photos of their children and/or grandchildren attached to their identity. The weird thing is that these people are 99/100 times incapable of actually having an argument, they either start insulting you straight away or say some variation of "if you don't like what I'm saying then leave" or actually majority of time "what does a foreigner like you know about this topic".
The real names policy has yielded zero of the promised effects imho.
I have no interest in posting under my real name. My surname is too unique, and I don't want employers to find me online
Us older ones are still fighting the last war. If only we explain it properly, maybe the bots will understand!
I'm going to be thinking about this perfect summary of my frustration of the last 20 years all day
I don’t understand your point at all in this context. What’s “Aumann Agreement” in the context of a discussion about values?
So if you’re not reaching a shared conclusion then you’re not exchanging information, you don’t have shared axioms, or one of you isn’t rational.
If it didn’t land with you, it was just a poor choice of phrase and while I care to attempt to explain it, I don’t care to attempt to defend it.
Values are not different from other positions that people might hold. I believe values and principles are just lossy compressions of positions about specifics because we cannot express them all. When you poke at the edges you find they are fractal. So “what’s your estimate for random variable X?” And “are open borders good?” Are not really different questions so much as questions about different levels of specificity.
My incomplete self-note about this here https://wiki.roshangeorge.dev/w/Principles_Are_Compressed_Im...
I believed they were all bots until I actually interacted with such a person in real life. It's terrifying.
Paid, lmao I wish. Nobody gets paid to repeat propaganda, generally.
Not all of them, and I'm not saying they target me. I noticed it more in popular engagement sinks - comments on popular instagram videos, replies on trending topics on Twitter, the main subs on Reddit.
There absolutely are engagement farms doing this for multiple reasons, and it at least seemed to me that most of the interactions in those places were not genuine.
I've had the same kind of argument you'd swear was a bot with real people I know offline. The best explanation I could come up with was that they got radicalized online, was shameless about arguing in bad faith, and was willing to say heinous things in a group text, but refused to continue the discussion when I challenged them to do so IRL when I next met them - their electronic words towards me were particularly foul. Polarization, online echo-chambers, lack of consequences, and shamelessness have conspired to crater the quality of discourse and provide plenty of fuel for bots and psy-ops operators to fan the flames.
"Spot the bug in this code" when the code can be looked at and pattern-matched against bugginess is something they seem really good at.
Some parts of debugging, like "Here is this logfile, what do you think is going on?" are also surprisingly good.
It's that thing kind of in the middle -- I know it when I see it honestly is the best way I can put it into words. An example from recently, I'm receiving some bad data on a network message parser. Immediately I don't know whether it's a my-side or their-side thing, but I know if I try and just vaguely describe the behaviour to the LLM it will start churning tokens.
My current approach to problems like this is -- I need to tell the LLM what it needs to do to give itself the data it needs to solve the problem. My first reaction now isn't "It's not working, there's a bug, it's not doing X". It's "Okay, this isn't quite working properly; I need you to add some debug logging around X, Y and Z so we can figure this out". That tends to avoid spirals and get me out of the situation much more quickly.
The seeing eye dog analogy is pretty apt actually. I would love to see some transcripts from the author if they are able.
Edit to add: I think the 'thing' I'm alluding to might be -- if I have trouble expressing the buggy behaviour clearly in words, then I know it's probably going to be a fair few back-and-forths with the LLM to get something; the harder I find it to concisely describe, the more risk that it'll fall into a pit. Doubly so if I offer up a hypothesis which turns out to be wrong.
LLMs generally only debug systems by reading the code with whatever information you give them in a prompt. The image in the article is meta-prompt - the prompt is whatever comes from the vision model the AI happens to use to 'understand' the red circle annotation. That won't work. To successfully debug what's going on it will need much better state information. Has the 'shelf' been explained to is? Is the contrast and lack of shadows in the image messing up the vision model? Why isn't the 'lid' in the image? And so on.
LLMs are clever but they're not magical. Treat them like a naive junior dev. Give them enough data about the state of something to understand it properly.
For LLMs, this is likely to be disproportionately effective as well: especially because they don't really build up a persistent view of the codebase, they're generally re-reading it each session, and they tend to be surprisingly good at predicting the behaviour of code.
(That said, knowing where and how to gather more evidence to make things clearer is a pretty core skill in troubleshooting, so it's generally good advice anyhow)
Sometimes you can do that mentally and fix the code. Often your fix will be right especially in a relatively simple part of the code. However, equally often you'll fix a different problem (or something that wasn't a problem at all), and the original bug will remain but you'll believe you corrected the issue. This is why you should always replicate a bug to understand it, and why you should always add a test whenever you fix a bug to prove you actually fixed it as well as preventing future regressions.
This is exactly how I debug things. When I first started working I was forbidden from using the debugger “you won’t have a debugger in the field”. Being able to read the code, spot the bug and tie it to the symptoms; that’s a skill. A damned important skill.
But lots of bugs can and should be found and fixed by inspection of the code, inspired by a reasonable observation.
Depending on your process needs, maybe write a test to confirm the fix.
You are doing those junior engineers a disservice with this message.
I'm usually trying to go in the other direction with them - you don't need to be able to reproduce it yourself or see the log files to narrow down or find the cause of an issue.
Logs/etc are just additional tool that make it easier.
Lost track of how many juniors I've seen throw their hands up and say they can't progress because they don't have logs/clean repro
Not the parent, but I think their approach is the correct one even though their line that you quote "You can't debug something..." should be everything instead of something.
The issue with your juniors is them throwing their hands in the air without actually reading the code when they hit a wall, and not them trying to reproduce and inspect the bug first.
What's the issue with telling them, "Obviously if you can't reproduce it and inspect it with dev tools, read the code."?
I've won bets off devs who blamed me cause they said they couldn't reproduce the bug consistently when they tried but it would happen randomly in production. Mind you, the bet wasn't even about who was at fault. The bet was that I could reproduce it if I actually tried.
You are complaining about juniors in a role that's known for the motto "it works on my machine" of course they're going to whine the moment it takes some effort to reproduce something.
At the end of the day, how are you going to claim you fixed an issue if you can't reproduce it to begin with? The senior has to now read the code more carefully cause there is no proof the junior actually fixed anything.
In addition, if you find the issue in the code -- unless it's a really unusual race condition -- you have what you need to construct a reproduction scenario. That's also part of the coaching.
[1] At least, I did until the place I work switched everyone to vibe coding. Now I give terse responses to their LLM-generated PRs -- which usually misses the forest for the trees -- since I know that they're just going to feed my words back into the machine.
It's a good approach if every run of your code is very slow or very expensive, but if you can iterate quickly by running the code it's way more effective to try to reproduce and narrow it down that way
I think you've been doing your juniors a disservice
I am teaching the more fundamental skill: how to (learn to) think through the problems and the systems in a way that lets log files be a useful tool instead of the only option.
Oh, I wish. "This call doesn't return, how strange! Let me fire off a bunch of small test script one after another, with a timeout of 10 minutes each, to narrow down where the error might be!" In terms of tokens, it's maybe even efficient. In terms of time it isn't!
I call this "thanoscoding" for two reasons:
1. "Fine, I'll do it myself"
2. In the past I found I have to "snap away" the mess the LLM made in order to start afresh from a known good state (generally with git reset). But that was 1-2 generations ago when it comes to models. GPT6 Astra probably does things right the first time, 90% of the time.
I've been using a few different "smart" LLM to work on an analysis, parsing, search and correlation tool that ultimately deals with a 5.5GB on disk (with indexes) mariadb database that has its origin as a federal government department's 905,000 row plain text CSV file.
There are a ridiculous number of data entry errors and just plain weird fuckups in the data origin that don't seem they will be ameliorated any time soon, so automating the drudge work of cleaning it up and rectifying it into something usable is a textbook case for this. Very pleased with the results so far.
Also.. uh.. isn't this.. good? I thought AI was to steal all our jobs.
___
Beside that, kinda weird self-description.
Isn't the computer executing your commands and you're just filling in where it cannot do that?
Being that dog implies that the computer is in the driver seat.
I mean it's supposed to be a joke I guess, but I read it as one that leaks internal metadata which seems to be incorrectly calibrated.
> Also.. uh.. isn't this.. good?
<!!spoiler ahead!!>There is a TV show called Person of Interest <!!spoiler ahead!!>, where Machine (AI connected to Internet and CCTV networks) has no legs or eyes, so when it needs to go and check something not covered by CCTV feeds, it gives instructions to a real person. In the show it is called an ‘analog interface.’
But it isn't. That's my point.
I told the clanker "hey do that", and like the intern/junior it emulates, it eventually says "boss! Help! I can't do this alone".
It is I who is in the driver seat.
From the original goal point of view yes, as it was boss who has initiated whole breakfast procedure. But from an execution standpoint it is intern who gives its boss a job of a grocery store run. He could give same job to anyone else, boss as a persona is irrelevant here.
Coding was never the reward. Acting as a translator for a machine is far worse than allowing the machine to solve the mundane parts and leave you with bigger building blocks to play with.
After 25 years I have to confess I hated being a software engineer. It felt like grinding in a video game.
Now I can finally create and innovate at the speed of thought, and I'm very grateful to have this technology now.
Serious question - what do you do for fun? I find fishing with friends enjoyable, and using my hands tidying up the old place I bought. It’s not innovating software I’ll never use but I’ll never tire of writing the same old ASP that delivers my clients the results they’re after.
Yes, when I think of OpenAI and Anthrophic and Google and Meta and any AI labs and their intentions, i cry a single tear for how these great instituitions are working so hard to bring freedom for humanity.
AI owned by a few companies is more likely to put the majority right back to serfdom. You're a privileged fool to believe that freedom is the likely outcome of the current stampede.
With a username that appears to cheer for a sociopath¹, I doubt reason will convince you.
¹ https://futurism.com/artificial-intelligence/sources-sam-alt...
(Like, Meta’s intentions? lol)
I hope you don't consider NWA a racial hate group given their name.
Also, I don't think you can cite a subjective take as objectively correct by including it as a footnote. Why do you need to medically diagnose him to say what he's doing is fucked up?
Why does he have to be literally the devil for outrage to work?
He could also not be a sociopath, but still his actions will be his actions, and it is okay to criticize them. Next what, you're going to link to an article that says Sam Altman is also a homosexual, and we must be morally outraged at that next? Because that is where your direction is going, instead of arguing about the ill effects of AI.
I think anyone who genuinely thinks AI is a wonderful thing for everyone probably enjoys the privilege of its use, reducing their own workload, without it (yet) affecting their employment, or their local environment, and possibly even lacks the empathy to comprehend negative affects on others. Obviously I was mistaken for thinking you were as such.
Putting words in others' mouths doesn't give credence to your points, the only homophobia here is coming from you. Straw men are not your friends.
Sociopathy in the context of running a powerful organisation is entirely relevant because making decisions that affect many, without really caring about the consequences for them, is generally regarded as a bad thing.
I think (this is subjective opinion from my experience so take it as such and not something objective) but a lot of your want for a better world is also linked with your desire to shame and ridicule people. That is not a healthy outlook towards life imo. Not that what you think is wrong morally (i am on your side actually as you realised) but rather the relationship with the moral values. And its bad not because your outrage is fake or stupid but because that relationship is not good for the self eventually. I can think OpenAI is doing a lot of wrong things, I think Sam Altman is sketchy person who is definitely not a morally good agent but I don't need to hate him or get triggered by him. I also think if he just disappears tomorrow, there will no hit to the broader forces which we should actually protest against imo.
> that relationship is not good for the self
I’d expect to come across, or read in my eyes, differently. Maybe more sympathetic, gracious, or kind, or I can’t think of the right word. Guess I see some bones one could throw to the interlocutor considering you probably share much of the same worldview.
Thanks for provoking some thought! I’ll come back to this one for another read and reconsider outrage & healthy outlooks on life.
Giant corpos hold every sliver of your so called freedom, you dont innovate, you repeat what others created before you. You use a tool that shackles your thoughts and creativity in a never before seen way, what you perceive is an illusion of liberty that is no present. Without others you are nothing and they can take it away any seconds, you are an addict, not an innovator.
I feel very much like a luddite..or some artisan of a bygone age. The guy above laments how he never enjoyed coding, this is an interesting sentiment, I know quite a few people who enjoy the craft itself, myself included. The ability to form words that have meaning, that create something from nothing. Sure, the words are just a tool, but its something _I_ can master, not some abstract wish machine that may change its functionality tommorow. Obviously one can argue that the computer itself is in this case the one that creates and not me, its not magic that just works with me, but the machine I got obeys me and me alone..another reason why personal computing is important.
It is true that closed weights models are a big issue. It is also true that LLM-generated solutions usually drift towards a median.
But that is not the dead end you think it might be.
Most coding work is repetitive boilerplate, and most typing is just.. well.. typing. Miserable work I too did not really enjoy. What I did enjoy were the end results, and that was just a necessary step to get there.
Now that's less the case than it was before we had LLMs, and for that I too am glad.
___
I think the article headline might've primed you (and me, fwiw) to reading the comment you're replying to as passive. But if you just look at the words of it, that might not actually be the case.
Astra is already good at taking screenshots and acting on it (part of the agi claims).