I dont think that is really true. There are many open weights models you can run yourself, including state of the art models like deepseek. Right now you its still expensive to run them at a reasonable speed, but for instance a $9500 mac studio can run deepseek at a reasonable, if not spectacular, speed.
The day to day of a software engineering job has changed and will change more depending on how advanced llms get (and all bets are off if we do get to AGI). However, I don't really think you will see a massive drop-off in employment for two reasons. First, competition is going away. If your developers are x times more productive, so are your competitors developers. I can see a bit less need in non-product companies if more domain experts are able to build their own tools, or generalized agentic tools eliminate some of the need for custom tooling.
However, I think there is a second thing that is often overlooked here. It seems the angle is always 'oh companies can replace developers' but no one seems to consider that developers can replace companies. I think you are going to see small teams of very skilled people replace able to make amazing products. There are limits to what llm's can do on their own but a skilled engineer who masters the tooling and can unblock the models and knows the techniques to keep the models productive as the codebase grows will be able to produce amazing things far beyond what they could ever produce before. I think you are going to see an explosion of new, smaller companies. I think if you are an engineer you shouldn't be gloomy, you should be excited.
I think category A is too broad. I enjoy building systems, not typing code. You can still get that joy having an LLM do the coding. You don't need to let it do the systems design. Also, some of the building process is the system I care about it, and some of it is tooling for the system I care about. Those tools are means to an end and I am happy to defer the entire process to the LLM, and those are stand-alone usually which is right in it's wheelhouse.
This is just wrong. I don't know how else to say it as this point. If you are doing something novel, you need to be a lot more specific in what you tell it to do, but it can still put it together 5x faster than you can do manually. It's like saying you can't do something novel with junior developers. On their own? Not really. Guided by someone experienced who can do the systems design and has the domain knowledge to be the product manager? Yes, you definitely can. And it's not different with advanced LLMs. It's even better than having junior developer's working for you because they don't get tired or bored. They won't slack on writing docs or tests. I have written novel systems by having LLMs do the grunt work, and they have better docs and tests than I have ever done before because it's willing to do the things I don't find as interesting without complaint.
I love this book but yes, you really need to understand when it makes sense to apply these patterns and when not to. I think of these kinds of architectural patterns like I think of project management. They both add an overhead, and both get a bad rap because if they are used indiscriminately, you will have many cases where the overhead completely dominates any value you get from applying them. However, when used judiciously they are critical to the success of the project.
For example, if I am standing up a straight-forward calendar rest api, I am not going to have a complicated architecture. However, these kinds of patterns, especially an adherence to a ports and adapters architecture, has been critical for me in building trading systems that are easy to switch between simulation and production modes seamlessly. In those cases I am really sure I will need to easily unplug simulators with real trading engines, or historical event feeds with real-time feeds, and its necessary that the business logic have not dual implementations to keep in sync.
I find the SqlLab in apache superset to be very good, and I have duckdb as a data source (anything that supports SqlAlchemy works). It works very well. To be honest, when I first saw the screenshot, I thought it was SqlLab. I haven't actually tried the duckdb ui, though.
I don't think it's really an appeal to authority. No one is saying it must be true because this guy says its true. However, when a very respected figure in software engineer culture like Steve Yegge gives an opinion, that is more noteworthy than when random joe schmo gives his opinion. The fact that you don't know who he is means it's not interesting to you. Clearly, it's noteworthy to others.
I haven't got to trying claude code yet, but absolutely with cursor and windsurf you can have the agent be reading the output of what it writes and runs and it can fix things it sees. You can also have it review code. It also help some times to have it review in a fresh chat with less context. I really think a lot of people on HN are not really pushing on everything that is available. Its magic for me but I spend a lot of effort and skill manifesting the magic. I'm not doubting other people's experience really, but wondering if they are giving up too fast because they actually don't want it work well for ego reasons.
Well I have also worked on systems of multiple millions of lines, well pre-llm, and I sure as he'll didn't actively understand every aspect of it. I understood deeply the area I work on and the contracts with my dependencies as well the contracts I provide. I also understand the overall architecture. We'll see how it goes if my project grows to that point, but I believe by clearing documenting those things, and overall focusing on low coupling, I can keep the workflow I have now, but with context loading for every session. Time will tell.
In general though, its been a lot of learning on how to make llms work for me, and I do wonder if people simply dismiss too quickly because they subconsciously don't want them to work. Also "llm" is too generic. Copilot with 4o sucks but claude in cursor and windsurf does not suck.
My experience has gotten better by focusing on documenting the system (with ai to speed up writing markdown). I find reasoning models quite good at understanding systems if you clearly tell them how it works. I think this creates a virtuous circle where I incrementally write much more documentation than I ever had the stomach for before. Of course this is still easier of you started greenfield buts allowed me to keep claude 3.7 in the game even as the code base is now 20k+ lines.
I have the same experience but I am still 5 to 10 times more productive using claude. I'll have it write a class, have it write tests for the class and give it the output of the tests, from which it usually figures out problems like "oops those methods don't exist". Along the way I am guiding it on the approach and architecture. Sometimes it does get stuck and it needs very specific intervention. You need to be a senior engineer to do this well, In the end I usually get what I want with way more tests than I would have the patience to write and a fraction of the time. Importantly since it now has the context loaded, I can have it write nicely formatted documentation and add bells and whistles like a pretty cli, with minimal effort. In the end I usually get what I want with better tests, docs and polish in a fraction of the time, especially with cursor which makes the iteration process so much faster.
I think this is what the kids call "copium". To be honest, when people think like this it makes me smile. I'd rather compete against people programming on punchcards.
This is that old finance chestnut. Two finance professors are walking down the hall and one of them spots a twenty dollar bill. He goes to pick it up but the other professor stops him and says "no don't bother. If there was twenty dollars there someone would have already picked it up"
Yes, people arbitrage away these anomalies, and make billions doing it.
I think what ozim is trying to get across and I agree with, is that "it's who you know" does not imply that its just a good ol boys network for it's own sake. Working with people you know and trust is normal. There is no way you can suss out in an interview loop or a client meeting what you can from a recommendation from someone who has with direct knowledge through experience with that business partner or employee. Managing relationships is part of the job and part of career building. For the career part, I have found that just being competent and nice goes a long way, without having to do aggressive networking.
I don't think that is true, or at least I have not seen that anywhere. My understanding is this bill removes any kind of non-competes. It was already basically unenforcible to have an unpaid non-compete in New York (through common law)
It's not usually a deal breaker because of how common it is, but it does matter. You are going to be disadvantaged against someone who can start sooner. I fully support this legislation. Non-competes are anti-competitive, period.
Not really. It's widely understood that NY won't enforce a non-compete if its not paid. It's very common in the hedge fund world to get paid your salary for a year to do nothing, even for junior developers. It depends on if the company really wants to hold you out or not. They will release you from the non-compete if they don't want to pay.
On the other hand, in the hedge fund world, bonuses are a big part of comp but generally only base pay is paid out, so in reality you might be say 150K to 250K while your comp in expectation is much, much higher. For a junior dev maybe your bonus is .3x to .6x base but for someone senior, your bonus might be 1x to 5x base or more depending on where you sit in the organization. Therefore sitting out still costs you a lot of money.
Thats a good point but also it means you will see particular areas be more dramatically affected by this. For instance, I have noticed big price buts in some ski areas where I have though about buying a small second home or condo.
Homeownership has never been anything close to 100% of the populaion. Historically I think its about 65%. It's probably more like you need a median income to afford a 25th percentile home. I don't have the actual data on that but it's my guess based on the fact that 65% of the population owns a home (which is verifiable).
I don't even think it requires that in a strict sense. No one knows how long the cycle time is or if you are will reach a new long-term mean or not.
Another important point is its not the case that all the hiring is LIFO. In fact we know that many old-time googlers were let go. I am sure many of the more recent hires have stuck around.
"The average person has more business sense than the CEO of Zoom and its entire board of executives who allowed rampant hiring because I could've guessed during COVID, despite the surge in demand given all of the work-from-home during lockdowns, my company most likely won't grow forever, therefore I won't triple my staff over the next 3 years and then need to just end up laying them off"
I don't mean this to be snarky but this is like saying its stupid for a department store to hire in the lead-up to Christmas. They are going to have to lay many or most of them off after. Couldn't they see that Christmas wouldn't last forever?
Think of it this way - if there is a huge market/demand expansion, and you don't meet that need/demand, someone else will. Even if you think that much of that expansion is temporary, you will have fallen behind competitors in technology and market share. When the boom ends, they lay people off and you don't have to, but you are still behind now.
I am really curious what the person writing this article thinks people will take away from it is. You failed three oral exams but were let through anyway because meritocracy has gone of out of fashion. Ok. I guess the article is just a slice of life piece for this magazine but its almost like it was generated by a right-wing ragebait factory to drum up racial tensions and anti-academic sentiment.
You are right but calling this 'obfuscate and save face' is a pretty uncharitable take. "let's take this offline" is just a polite jargon for 'this is off-topic or inappropriate for this venue'.
I generally find hacker news interpretation of business jargon as uncharitable and unnessarily hostile. Not sure why such an innoccous expression like 'circle-back' sends people into fits of rage. There is good reason to be circumspect about heavy project-management overhead as its often misapllied, but that really has nothing to do with this jargon.
> What if I also told you that memorising facts is not the same as learning?
It's interesting, I believe this but recently have come to a little bit of a modified conclusion. I might not fully grasp a particular concept, but the brain is clearly a very powerful generalizer. Feeding in facts, even if they aren't fully connected for you yet, can help your brain make that connection. I think the key is to not just accept that I will memorize x, y and z but that i will keep thinking about it as I ingest new data.