HNHacker News
TopNewBestAskShowJobs

axegon_

3,418 karma · joined November 9, 2018

email: ping@axegon.com
submissionscomments
axegon_··on It's probably time to rethink “Building in public”
The way I see it, those are just a spin on "I did X for Y months and here's what I learned" articles. Clickbait at best but mostly those are just articles that fill void with words. Whenever I stumble upon those, I am never really buy into them. I am a huge believer in the idea of the American dream but in a functioning economy, this can only work a handful of times. Over the years I've attempted creating a startup and bootstrap it on my own or with a small team of friends and the one thing I've learned from those attempts(see what I did there...) is that it's practically impossible, even if you have the right idea and you get the timing right. Arguably I didn't get the timing right on a few occasions and I was way too early for the party. Had I waited for a few years, those might have picked up but the truth is, I sincerely doubt it. At this point, whenever I land on one of those indie gigs, I instinctively roll my eyes and click away.
axegon_··on Tech Jobs Have Dried Up–-and Aren't Coming Back Soon
I can't say about the US. That's not so much of an issue here on the other end of the pond, if anything people are increasingly distancing themselves from the 20-30 age group. I've had my experience with them and a large chunk of them are insufferable. A notable example - one had a complete meltdown when he was told that go is not a low level language(and being the one that said it, I was his biggest enemy) :D

It pains me to say this, but I agree with you on the subject of AI. I think people are wildly misusing it and creating yet another bubble. It is good at digesting and sifting through data but it's terrible at providing good solutions and people are using it for the latter. I've been actively using it for months at work specifically for sifting through large, completely unstructured documents and pull out the relevant data from them and it has done wonders. So much so that the management has invested in some beefy GPUs for the occasion. And by beefy I do mean BEEFY, coming from someone who owns 2 * 24GB VRAM nvidias for personal usage. I also feel like the day when new models become a rare and unexpected occurrences is not far.

axegon_··on Tech Jobs Have Dried Up–-and Aren't Coming Back Soon
Clickbait. Yes, there is a shift but it has little to nothing to do with AI. While there are entire companies getting hyped about generative networks and LLM's(read "managers thinking they will get a fleet of devs for little to no money"), those are the exception and will go extinct sooner or later - I can name a few off the top of my head.

There was a huge boom around Covid, sure, when infrastructure was clearly not ready to cope with the demand so as many people had to be dragged into the equation. There were also companies aggressively hiring tech people partially because of the demand but also to establish a monopoly. And the rise of one-online-course developer programs was the cherry on top. That was bound to collapse and Musk's campaign over at Twitter gave everyone the confidence that there's no need for anyone but a few managers and prompt "engineers". And we are reaching the "find out" part for all players involved.

If you are a tech worker with skills, then you have absolutely nothing to worry about: AI can only write trivial code which normally would be thrown at an intern - stuff like "write a function that takes in a hashmap containing key-value mappings and a second hashmap to swap out the said keys", which you will likely still need to tweak to make it safe. But for any serious or critical code, if you are relying on a LLM to do the logic for you, you are basically walking around blindfolded, holding a loaded gun pointed at your head. Sadly I know a lot of people who actively do that exact thing.

The other part is that the market is over-saturated with products that no one needs. How many more social medias do you need? None. How many more trading platforms? None. How many more messenger apps? None? Dating apps? None. Yet new, "revolutionary" ones pop out daily. Guess how many of them fail? Their marketing campaign was "blockchain" 5 years ago. Now it's "ai". My coffee machine needs neither of those, who are you trying to sell this to?

When it comes to the tech sector, people should understand that this is not a bottomless pit and it all boils down to a finite amount of people that will use your product(consciously or not). And by implications you need a finite amount of tech workers. The jobs have not dried up, it's partially normalization after the Covid hype and to a lesser degree managers thinking AI will do their tech work - the same people who have become Musk cult members. I used to work for one and guess what - the company he founded in 2017 with very noble intentions and received a really good funding is collapsing cause they are unable to roll out a product. Those are bound to fail either way.

As for courses - there are talented people who have only done a bunch of those, but they are the extreme exception and realistically most of the people who enrolled in those were simply there because they thought that's a cheap ticket to a 6-figure salary.

axegon_··on Old Computing Books
My first thought was "I wonder if the C programming language would be on that list" and sure enough... Worth noting that, although the book is 45+ years old, it's aged incredibly well (I know, there have been multiple "patches" applied since the first edition). But considering how books that came out 5 years ago are wildly outdated today, while this one is just as relevant as it was when it first came out is a huge achievement. Major props to Ritchie and Kernighan.
axegon_··on The Subprime AI Crisis
I find this to be a bit exaggerated but there are a few fair points. I think people are starting to see that ai is not as good as the demos and flashy videos makes them believe. Since the day chatgpt came out, my biggest problem with it has been the fact that you will ask it a mildly difficult question on any subject and it will confidently give you a response that's complete bs. And I'm starting to see people coming to the same conclusion as time goes by, despite the advancements. Ai has a place in this world even with it's current limitations and shortcomings, no doubt about it - there are cases in which it does a nearly perfect job - I'm currently building an AI rig for my home lab because it has a ton of use cases even outside my daily work(where I do use llms for processing unstructured data).

The second issue I see is that data is becoming widely unavailable-everything is getting blocked, crawlers are getting cut off immediately, 1000 API calls cost as much as a mid-sized flat in a European capital - the complete opposite of what the internet was supposed to be.

The third issue is the fact that people are gladly using chatgpt to answer questions on stackoverflow for example. We know how badly llms start performing when you train them on their own data... Not to mention the considerable spike in critical bugs in open source projects over the last two years-there's a good chance it a lot to do with it.

The fourth is social media-there are already tons of examples where troll farms are no longer paid workers in some dump in siberia but are in fact powered by chatgpt out some other service. And to think that I laughed at the dead internet theory when I first heard about it...

axegon_··on Telegram founder charged with wide range of crimes in France
I love it how fast people are jumping in to defend a guy who is by all accounts guilty because "well how is he supposed to know illegal stuff is happening on his platform?". He knew and everyone behind telegram knew. Well read this: https://eur-lex.europa.eu/eli/dir/2000/31/oj

Telegram has legal obligations and we all know they haven't fulfilled them-you can easily verify that on your own: you don't need to be a government agency to do that - open telegram and use the search function, hell even the auto suggestions lead you to a ton of illegal stuff.

axegon_··on French prosecutors say Telegram CEO freed from custody, will appear in court
Shame.
axegon_··on Why is the Oral-B iOS app almost 300 MB? And why is Colgate's app even bigger..?
I asked a similar question a while back on here regarding a toothbrush with an esp inside. I think one of the answers I got captured it nicely: https://www.youtube.com/watch?v=TQ4W7yB9Mow
axegon_··on 80% of AI Projects Crash and Burn, Billions Wasted Says Rand Report
To be honest, I am really frustrated with what is happening: the hype train killing something which in principle could be a good thing, as usual. In the second half of the 2010s, it was blockchain: Payments - blockchain is the solution. Logistics - blockchain. World hunger - blockchain. Cure for cancer - blockchain. 75% of all job offers from startups were blockchain-related, and admittedly, I worked at such a startup, which, from what I'm able to gather, is a few months away from total collapse.

With vision models in the late 2010s, I was seeing AI winter 2.0 just around the corner - it felt like this was the best we could come up with. GANs were, to a very large degree, a party trick (and frankly, they still are).

LLMs changed that. And now everyone is shoving AI assistants down our throats, and people are trying to solve the exact same problems they were before, except now it's not blockchain but AI. To be clear: I was never on board with blockchain. AI - I can get behind it in some scenarios, and frankly, I use it every now and then. Startups and founders are very well aware that most startups and founders fail. But most commonly, they fail to acknowledge that the likelihood of them being part of the failing chunk is astronomically high.

Check this: a year and a half after ChatGPT came about and a number of very good open-source LLMs emerged, everyone and their dog has come up with some AI product (90% of the time it's an assistant). An assistant which, at large, is not very good. In addition, most of those are just frontends to ChatGPT. How do I know? Glad you asked - I've also been very critical of the modern-day web since people have been doing everything they can to outsource everything to the client. The number of times I've seen "id": "gpt-3.5-turbo" in the developer tools is astronomical.

Here's the simple truth: writing the code to train an AI model is not wildly difficult with all the documentation and resources you can get for free. The problems are:

Finding a shit load of data (and good data), which is becoming increasingly more difficult and borderline impossible - everyone is fencing their sites, services, and APIs - APIs which were completely free 2 years ago will set you back tens of thousands for even basic data.

As I said, the code you need to write is not something out of reach. Training it, on the other hand, is borderline impossible. Simply because it costs A LOT. Take Phi-3, which is a model you can easily run on a decent consumer-grade GPU. And even if you are aiming a bit higher, you can get something like a V100 on eBay for very little. But if you open up the documentation, you will see that in order to train it, Microsoft used 512x H100s. Even renting them out will set you back millions, and you can't be too sure how well you would be able to pull it off.

So in the grand scheme of things, what is happening now is the corporate equivalent of pump-and-dump. It's not even fake it till you make it. The big question on my mind is what would happen with the thousands of companies that have received substantial investments, have delivered a product, only for it to crash the second OpenAI stops working. And even not so much the companies, but the people behind these companies. As a friend once said, "If you owe 1M to the bank, you have a problem. If you owe 1B to the bank, the bank has a problem." In the context of startup investments, you are probably closer to 1M than 1B. Then again investors are commonly putting their eggs in different baskets but as it happens with investments and the current situation, all baskets are pretty risky, and the safe baskets are pretty full.

We are already seeing tons of failed products that have burned through astronomical amounts of cash. I am a believer in AI as an enhancement tool (not for productivity, not for solving problems, but just as an enhancement to your stack of tools). What I do fear is that sooner or later, people will start getting disappointed and frustrated with the lack of results, and before you know it, just the acronym "AI" will make everyone roll their eyes when they hear it. Examples: "www", "SEO", "online ads", "apps", "cloud", "blockchain".

axegon_··on Arrest of Pavel Durov, Telegram CEO, charges of terrorism, fraud, child porn
If I have to dig through third party clients in order to trust a system, then it's clearly a shit system. Signal > anything else, especially telegram, which can burn in hell for all I care.
axegon_··on Dual-screen laptops make more sense with this spiral notebook-like hinge
Well I'm the one that patched the driver for the bottom display brightness control for kernel >= 5.8. There's a good community around these laptops so long term suffering is pretty unlikely :P
axegon_··on Dual-screen laptops make more sense with this spiral notebook-like hinge
I've been rocking an Asus zenbook duo(first gen from like 2020) for a few years now. Well... Since 2020. It has to be said, in the case of the zenbook, there's a lot of compromise in terms of user experience. The small screen is useful on the go but I do use a desktop computer 95% of the time. I'm ignoring the fact that the linux drivers are a bit of a pain in the ass. The bigger problem is the fact that although you gain some additional screen area, the keyboard is a bit awkward but nowhere nearly as awkward as the trackpad. It is practically unusable so you need to carry around a mouse with you.

Setting those things aside, I have mixed feelings about adding more hinges to laptops. In my experience the hinges have always been the Achilles' heel of laptops. Also any additional screen adds up a fair bit of weight which seems trivial until you end up stranded on an airport for 20 hours and have to figure out how to take a nap on the bench while your stuff is secure(looking at my dumb younger self who decided to buy a 17 inch Sony laptop at the end of my first year in university). And the extra screen means your battery life will suffer. I mean yeah, I have 4 monitors on my workstation and when it comes to work, they actually make sense: one for a terminal and logging, one for actual code, one for documentation and one to manage chats and music. But when it comes to laptops... I'm not so sure anymore.

axegon_··on Arrest of Pavel Durov, Telegram CEO, charges of terrorism, fraud, child porn
:%s/selling groups/selling drugs/g
axegon_··on Arrest of Pavel Durov, Telegram CEO, charges of terrorism, fraud, child porn
The difference is that this is not an isolated case on telegram(you said it yourself: "some amount", which implies "limited"). At the same time, you can literally open up the app and with 0 effort find everything they are accusing them of - drugs, terrorist organizations, public decapitations, you name it. They also provide the ability to search for people and groups around you, and I am literally seeing a channel where people are buying and selling groups "800 meters away" from me and another one for prostitution, which is also illegal in my country. Meanwhile, see their TOS[1]. They have not complied with any of the reports or requests from users (and governments by the looks of it) to crack down on them. While 1:1 chats are theoretically private and encrypted(full disclosure, I do not trust Telegram or any of the people behind it), telegram's security for public channels and groups is absolutely appalling and they are well aware of it - they just chose to look the other way and hope they'd get away with it. You could have given them the benefit of the doubt if those are isolated("some") instances, sure. But just as in the case of Kim Dot-I-support-genocide-com, those are not isolated cases and saying that they had no idea is an obvious lie.

2000/31/EC[2], states that providers are generally not liable for the content they host IF they do not have actual knowledge of illegal activity or content AND upon obtaining such knowledge, they take action and remove and disable access to that content(telegram has been ignoring those). Service providers have no general obligation to monitor but they need to provide notice and take down mechanisms. Assuming that their statement are correct, and they had no idea, they should be in the clear. Telegram provides a notice and take down mechanism. But saying that there are channels with +500k subscribers filled with people celebrating a 4 year old girl with a blown off leg in Ukraine and no one has reported it in 2 and a half years after it was created is indeed naive.

[1] https://telegram.org/tos/eu

[2] https://eur-lex.europa.eu/eli/dir/2000/31/oj

axegon_··on Claude's API now supports CORS requests, enabling client-side applications
I see three problems with this:

1. From a product perspective, this is like going to a restaurant to get dinner but having to bring your own kitchen utensils, food and cooking your dinner yourself.

2. Anything running in a browser is inherently insecure - what's the guarantee that the site where you're pasting your key doesn't have some incredibly stupid security flaw and your key gets leaked?

3. Even if there are no vulnerabilities, you're still pasting your code in a random form somewhere on the web. All it takes is an ajax call or a websocket and someone, somewhere has your key.

axegon_··on Show HN: I Made a Website for Problems
In principle this is a good idea but it's definitely a double edged knife for the users. From what I understand, your idea is to rely on some level of anonymity. Let's assume your security is 11/10 and nothing gets leaked - great. But people are very bad when it comes to this though: they assume that a semi-random username somewhere will shield them, which couldn't be further from the truth. You need to share 3 seemingly trivial facts about you, and people with good experience will be able to sift through that and figure out your identity. I have a case from about 2 years ago: a guy shared the country where he studied and a place where he worked. A few people instantly connected the dots and figured out who he was. Those of us that did just laughed about it and moved on with our lives(his existence really couldn't possibly be any more sad than it was). However eventually others figured out who he was and the fact that he attacked homosexuals constantly, while complaining about his diarrhea and describing in detail his wife's pregnancy (and being the son of a media oligarch)... You can imagine how that all blew up in his face. While he can only blame himself for this, there could likely be cases when you might also be held responsible. What I'm trying to say is, be careful. Good luck.
axegon_··on What you learn by making a new programming language
Likewise, I've done that more times than I'd like to admit. And I took it a step further with an LLM like a month and a half ago fenced behind a number of json and yaml instructions, containing conditions and validators. It works like a charm but yeah... Oops, I did it again.
axegon_··on Ask HN: Resources about math behind A/B testing
Many years ago I was working for a large gaming company and I was the one who developed a very optimal and cheap way to split any cluster of users into A/B groups. The company was extremely happy with how well that worked. However I did some investigation on my own a year later to see how the business development people were using it and... Yeah, pretty much what you said. They were literally brute forcing different configuration until they(more or less) got the desired results.
axegon_··on OpenCV and the Wild Kingdom
If I understand correctly, I feel like OpenCV and computer vision is a huge overkill for something like that. Here are a few thoughts:

1. A microcontroller with a LoRa module and a motion sensor will get you a long way and it will consume little to no power. That said, you also need to make sure you have a device to either read the signal or someone in a radius of 4 to 6km is hosting a lorawan gateway. If not, I'm sure people will be grateful if you set up one yourself - to my mind it's worth it and I happen to be the only person that hosts one and covers my area of the city. Just make sure you don't enable any of the other gimmicks the microcontroller might come with - wifi, bluetooth or whatever. For instance I have a bunch of these [1] and they seem like they would do the trick for you if you pair them with a motion sensor. I need to warn you though, the documentation is appalling.

2. If you do wanna go CV - you can but it will never be low power and you will need to figure out some connectivity. If it's a raspberry pi or some other SBC with GPIO, you could of course hook up lora to it. As for the CV itself - opencv will get you there or you could train a very small yolo-like neural network - if you control the environment/lighting and they are static, you only need a handful of sample images to train it and you will get great accuracy. Like even something in the very low double digits will do the trick. Imo that would be overkill and your best bet is still the first option.

[1] https://www.lilygo.cc/en-bg/products/lora3

axegon_··on Math is running out of problems
If there's one thing that history has taught us, regardless of which field, it's that anything we manage to answer raises at least half a dozen other questions. Saying that math, of all fields, is running out of problems is one of the most absurd statements in this day and age. I was blown away by the capabilities of dumb SVMs in university and today SVMs look like a child's play - just over a decade later.

For as long as humanity has existed, whenever we feel like we've reached our peak, something happens and completely shatters our understanding. Say we have prime numbers and their occurrence is completely unpredictable - the same way Aristotle was convinced that objects come to a rest because they get tired. It's not a specific branch of knowledge, it's an outrageous claim like many have already pointed out. Could it be that math is running out of problems? Yes - in the same way that the universe might vanish tomorrow. Both of those claims are equally absurd.

axegon_··on Math is running out of problems
I have only one [1] thing to say here

[1] https://en.wikipedia.org/wiki/Charles_Holland_Duell#Everythi...

axegon_··on Ask HN: How do you go about finding a job?
I deleted my linkedin years ago: I found it utterly useless. i won't lie, my social media game is as bad as they get but linkedin seems like the least effective option. I have not gotten the standard "while your credentials and experience is very impressive blah-blah" bollocks message from linkedin applications. So far it's been either local job boards or angel.com.

Recruiters are a mixed bag. I've had wonderful and horrible experiences(and almost nothing in between). My current job - 10/10. The one before - same - the guy was awesome and was the reason I took the offer(against my better judgment) and ultimately was the only person I liked in the company and the only one I miss. The job before that - uugh... I joined the company before the HR did and had it been the other way around, I would have turned the job down - she was absolutely unbearable. In the grand scheme of things I agree: HRs are atrociously insufferable.

axegon_··on Ask HN: Is there any software you only made for your own use but nobody else?
Plenty as a matter of fact. Two large ones off the top of my head.

First one: cloud services are expensive and come with a lot of overhead. However I do enjoy the idea of lambda/cloud functions a lot but I didn't want to rely on cloud services for them, especially since I have a ton of hardware to self-host stuff. So I built a service which piggybacks on DIND where I could add templates for such lambdas for a few languages(just rust and python cause that's mostly what I use) where I can allocate some resources and build tiny containers. In order to avoid wasting resources, they are all off by default but through the magic of what I build(management and something like a reverse proxy + container management for DIND), I can call an endpoint, which will check if there is a container to handle the requests and proxies the request and response to and from it or it will spin up the container, do the same and shut it down after a pre-defined amount of time. You could argue that I could have achieved the same with minikube or k3s but this has proven to use a lot more resources compared to a single rust binary I can run on anything x86 with docker on it. The downsides are that I can't distribute it across multiple servers so it only runs locally. But that fits my needs perfectly. There might be some people that would be interested in it but I never got to open sourcing it so I'm the only user. I've been planning to open source it "next week" for the past 2 years or so.

Second one: I love data. Specifically unstructured, raw data and I've been collecting a ton of articles from the internet over the past 2 years or so. The ingestion involves a pipeline that automatically processes them and extracts data/makes them a lot less unstructured: events, locations, timestamps, what has happened where and all that jazz. It's a long list of ml models, data transformations, validations, 3 llm's fighting each-other and a bunch of other stuff. Kinda feels like a child to me so I don't want to open source it. I know that I could hypothetically convert it into a product and make some money off it but my past entrepreneurial experiences have been very unsuccessful, I have limited time and lack motivation. So at the moment I'm simply enjoying the data getting spit out on a grafana dashboard such as this [1].

[1] https://i.imgur.com/fAoQghk.png

axegon_··on Aeon: OpenSUSE for Lazy Developers
I am although I am on the fence. As a developer tumbleweed provides me with everything I need - new releases, new kernels, new packages, etc. Mind you I use Leap at work and I gotta say I am really disappointed. There are some new packages which I can only get running inside docker containers which is a huge pita. I might end up going back to fedora 10+ years after I decided I had had enough of it.
axegon_··on US Bans Kaspersky Software
Brilliant! I have to say, I'm really impressed by how much the US is stepping up it's game, first with TikTok, now this. I wish Europe would do the same. Like yesterday.
axegon_··on Protecting Children's Safety Requires End-to-End Encryption
Not a whole lot to discuss here, I fully agree. This goes beyond children's safety and into general privacy(which the vast majority of people are totally clueless about, contrary to their beliefs): no amount of encryption can compensate for bad privacy habits. People tend to think that the bottomless pit that is the internet and the billions of users will keep them hidden, which couldn't be further from the truth. And speaking of the truth, all people are surprisingly unique in their own way and there is no such thing as "average". I've been fiddling with this idea for a while and I am putting it to the test in my spare time and so far the results are astonishing: it's far easier to doxx someone online as a consequence of a personal fact that they've publicly shared than some data leak by a private company or even a government institution. Leaks make matters worse, sure but the pile grows exponentially and not linearly(which seems to be what most assume): People get freaked out about their number being posted somewhere online while not realizing that by saying something along the lines of "I studied in London in the early 2010's, then in Edinburgh until 2015" and "I lived in Paris between 2017 and 2019" and someone wants to find their identity, they've willingly helped them narrow down the identity to a handful of people - likely in the very low double digits. Add one more random fact, such as "I have a Volvo xc 70" and you've basically put a pin on yourself on the map.

Most people have probably never considered this. It may sound like a cliche but it's far more important to educate people as well as their children from a very early age about those things than to assume that outsourcing the problem to technical solutions would somehow solve it - it won't.

axegon_··on Ask HN: What happens when I click "request for quote" on your SaaS?
My current company is not a SaaS but I have worked at two in the past and both did the exact same thing: schedule a call with a sales person.
axegon_··on The Raspberry Pi 5 Is No Match for a Tini-Mini-Micro PC
Well... Yeah... I have a bunch of raspberry pis from back when they were cheap + a raspberry pi 400 I won at a hackathon a few years ago. There are things about them that I love(the older ones to be more specific): low power and great thermals. But ever since their prices skyrocketed, I have completely abandoned the newer ones and opted out for what me and my friends refer to as "cubes". Optiplex micros and similar ones. They are a tiny bit more expensive but you have expandable storage and memory and in lots of cases upgradable cpu if it comes down to it. And most importantly x86, which I take over ARM any day of the week, especially when I have to dive a layer or two deeper - it's just more convenient having all the additional tools.

However I have one or two projects that I keep kicking down the road due to time constraints which involves a rasprerry pi zero w2. The reason being is that it's compact, packs a decent punch and is extremely easy to power it from an 18650 battery and keep a low profile.

I'm fine with the current prices of the raspberry pis but it's hard to justify using one of them if you don't need GPIO pins or something that uses little power. My advice to the raspberry pi foundation is to stop trying to push performance and focus on features. I'd be the first in line to get a new raspberry if it comes with a battery hat and LoRa built in(the project I was referring to really). Here's another one(which ironically already sort of exists): the milkv duo. There are lots of cases when you don't need a full operating system 24/7, which would waste a lot of resources to do one-off tasks every now and then. Ideally you should be able to boot the OS from a sensor or some sort of signal, which is hooked to a very under-powered micro-controller only when you really need it, let it do it's job and then power it off. As we stand, the raspberry pi foundation largely offers the same product in different form factors with some arguably marginal upgrades, which most people don't really need.

axegon_··on Cost of self hosting Llama-3 8B-Instruct
Up until not too long ago I assumed that self-hosting an llm would come at an outrageous cost. I have a bunch of problems with LLM's in general. The major one is that all LLMs(even openAI) will produce output which will give anyone a great sense of confidence, only to be later slapped across the face with the harsh reality-for anything involving serious reasoning, chances are the response you got was at large bullshit. The second one is that I do not entirely trust those companies with my data, be it OpenAI, Microsoft or Github or any other.

That said, a while ago there was this[1] thread on here which helped me snatch a brand new, unboxed p40 for peanuts. Really, the cost was 2 or 3 jars of good quality peanut butter. Sadly it's still collecting dust since although my workstation can accommodate it, cooling is a bit of an issue - I 3D printed a bunch of hacky vents but I haven't had the time to put it all together.

The reason why I went this road was phi-3, which blew me away by how powerful, yet compact it is. Again, I would not trust it with anything big, but I have been using it for sifting through a bunch of raw, unstructured text and extract data from it and it's honestly done wonders. Overall, depending on your budget and your goal, running an llm in your home lab is a very appealing idea.

[1] https://news.ycombinator.com/item?id=39477848

axegon_··on Ask HN: What was your most humbling learning moment?
So much this. But there is a fine line somewhere in there. I have seen and admittedly worked on such projects where time constraints simply do not allow you to polish all bits and pieces. That's perfectly reasonable during POC/early development stages and if anything, I encourage it: Writing hundreds of tests when all requirements change three times a day is incredibly counter-productive, slow you down and eventually burn you out. It may happen that crappy is much better than shiny, polished and over complicated if the project itself is not going to scale any further than it already has. And once you get a more complete picture of what the end goals are, then you can go back and gradually start doing things "the right way". But I've also been in another extreme. Take my old job for instance, which despite my at the time cognitive dissonance, I hated to a large degree because of this: relatively early stage with ever-changing requirements, endless dependencies, brutally unstable setup for development where deploying a single pod in minikube was a terrifying prospect because everything was hanging by a thread and rebuilding the cluster took hours. That was made even worse by dozens of forked open source projects that were patched left and right to fit the needs, lagging years behind the original projects, wild dependencies between repos, no real observability over what was going on, the version control was catastrophically packed with auto-generated code(like 80% or more) which was made worse by the fact that everything was git push --force so they don't have to deal with conflicts. Imagine having to explain that this practice should be avoided in nearly all cases. In a nutshell imagine crappy code and infrastructure which pretends to be and is sold as an enterprise grade software. I guess cognitive dissonance was a common theme in the company since everyone was under the impression that everything is perfect. Which couldn't have been further from the truth.
← PreviousPage 11 of 34Next →