Going back to CS and IT, I found myself talking many times along the years with good software developers coming from a totally different education path (graphics, philosophy, agricultural science, etc) and they had blind spots in areas that I had been exposed to at least a bit at university. Examples: SQL, one of them knew perfectly well how the ORM of the day worked (some JS one 12+ years ago) but not why; networking, one of them knew only to enter the SSID and password and everything else was magic.
The issue is then, that those scripts are probably discarded after one use by the user, partly because the LLM doesn't make them generalized enough and focuses on solving the one task at that time. Partly because, when the issue next arises, it seems cheap to the user to just have the LLM do it again.
I think these models could be trained/fine tuned in a way to anticipate future issues or create general enough solutions but it would imply lowering their own use... so there might not be an incentive for that.
Let me just write a small Python script to decompile a section of this DLL so I can verify the constants for this function matches what the reference said.
With the images, I'll write a small Python script to compute the motion error using the Lucas-Kanade method[1].
Almost literally what Claude (Opus) casually has written as it chipped away at a couple of my problems.
[1]: https://en.wikipedia.org/wiki/Lucas%E2%80%93Kanade_method
When I found and configured a local LLM to handle files, it took 6 min. 45 seconds to process two files --- I need to handle almost 5,000 in Q4.
When running locally, it took the LLM 405 seconds to process two files.
If the task is filling in tax forms from a spreadsheet then yeah, you can ask an LLM to write a script for that.
If the task is proving thousands of math theorems or the task is translating specs into working software, then yeah, I guess you do need an LLM.
For each file which was uploaded create a move command in a .bat batch file which will rename the matching file using the current filename of the uploaded file and use the amount, and date of the check, the Invoice ID, PWS ID, and COMPANYNAME in all UPPERCASE run together with no spaces from the PWS population line using a filename like to:
$<check amount>_<Account ID>_<date of check as YYYY-MM-DD>_<Invoice ID>_<PWS ID>_<COMPANYNAME>.pdf
Do not explain how this is to be done, open each file, examine it, and create the appropriate command, returning the contents of a .bat file with each command on a separate line which may be run on the files in that directory to rename the files from the name as uploaded to the generated filename with the command on its own line.It's still to this day one of my favorite reads. By the time I read it, I knew about a lot of the possibilities within it already. But it does exactly this: it skims across the top of dozens of topics, and it makes you just aware enough of... all sorts of areas of CS.
If you're curious, you might dig deeper - at worst, you know these all exist.
[1]: https://www.amazon.com/New-Turing-Omnibus-Turning-Excursions...
https://www.amazon.com/New-Turing-Omnibus-Sixty-Six-Excursio...
P.S. As sibling noted, the new price in that Amazon link is completely insane. Getting it used for a few $/€/£ is well worth it though
I remember that the book uses a basic or python-like pseudo code and that their neural network implementation has a couple of very subtle bugs which prevent a direct translation to python. The surrounding explanation was correct, but the code wasn't, and so it (presumably unintentionally) forced me to really understand what was happening in order to get it working. I forget the bug, something like a dropped negative.
The information density of the book is just so incredible. Each chapter is like 6 pages long and somehow that's enough to go from zero to a working single-layer neural network, or boolean logic circuits, or error-correcting codes, or 50 other things. I remember it being almost addicting to work through.
- Practical Electronics for Inventors.
- Electronics All In One For Dummies.
Hope folks would chime in with more, I would love more recommendations here.
https://nostarch.com/secret-life-of-circuits#content
And there’s “Practical electronics for inventors”
https://www.mheducation.com/highered/mhp/product/practical-e...
But electronics is not that large of a field and understanding the mathematical models is kinda required to design stuff as components can be rather fragile and not have a lot of room to play around.
This guy is great because he doesn’t talk, he just makes stuff:
I got through most of Art of Electronics, and then found this, and wished I had just gone through this series the whole time. AoE is much more of a reference than a good learning-from-scratch resource.
https://archive.org/details/newturingomnibus0000dewd/page/n2...
After primary and secondary education – which were very broad – I had expected to learn in depth about something narrow when I went to university. A few years into a computer science programme, I was at first a little disappointed when I realised I had still only taken what could reasonably be called introductory courses in various sub-studies of computer science.
But as I got out into the workforce, I've noticed it's been useful to have tried a bit of everything and to have the vocabulary to talk about the basics in each of the sub-studies.
This is basically the opposite of the last 50+ years where vertical knowledge (being a savant in a certain very specific field) was seen as the most important thing to have in a super specialised world.
This of course is just my perception, which needless to say, may be wildly off
for instance, if you use a shitty library X in a project, and you ask your frontier model "please change A -> B" the model will just do that, it will not tell you "well maybe you could also swap the whole thing with Y" because that's not its job - it's job it to change A to B.
(I agree, they may not be good at making strong long-term decision still, but that may also be an issue of inadequate context a lot of the time)
Humans have a practically infinite "context window" (active memory), we make associations and vaguely remember that thing we heard about that might be useful here, let's explore this thread ("you're absolutely right!"). Agents, being powerful as they are, don't do that, they pattern-match on the task and use the obvious solution, but don't activate their latent knowledge until you point them in the right direction.
The obvious solution might be the best one in 80% of cases or if you don't really care to get the best results (and it's absolutely fine!), but if you do care in your particular case, you have to babysit your agents and really understand if they're doing what you had in mind.
The sweet spot is a true expert in a domain driving the LLM and keeping it honest, whether that be via agents/skills or reviewing work. In software dev, there’s huge tranches of work at junior/mid level where LLMs can just go, and anything where there is a clearly testable goal it can iterate on faster than a human expert, for example recent progresses in maths proofs.
But, I don’t think there’s an LLM out there that is an expert at everything. At least, not yet.
Just knowing what's possible, having been exposed to a bunch of ideas, libraries, tools, and, yes, reading too much Hacker News - seems like it's allowed me to ask good questions, propose interesting experiments, and drive my own knowledge forward faster with LLMs.
I don't know that it's a particularly unique skill, but from personal experience there seems to be a spectrum of ... creativity? Freedom of thought / expression? And people who are kinda stuck in their ways seem to get less out of LLMs than people who have some idea of what they want and even some direction in mind, but just need the details made frictionless.
I think I'm not doing the best job expressing what I'm on about, so here's an example - I have a little livecoding app[0], which I developed (vibe-coded) for a conference talk I gave in October[1]. It's been a lot of fun to play with and has formed a good surface for lots of different experiments that never made it into the codebase - just throwaway ideas, etc.
So when all the hype about Jev was going a month or two ago, my experiment was "can Jev sit in on the jam?" And what I did was hook it up to the app - not to the "musical context" information that other agents (agents in the Elixir sense, not really the 'llm agent' sense) in that app receive, but rather to the raw signal info that was coming out of the master audio node. And then Jev made a prediction about what note to play, and then that note was sent through WebMIDI to a VST on my machine so that Jev was almost like an outside "guy" soloing over the app's music.
(Jev was a pretty boring soloist, because playing the "most likely" note every time is not especially exciting, but that's beside the point here).
Maybe Claude / Codex could have come up with that idea without me - surely the models know about all of those different tools / options already. But I'd bet that if I hadn't known what to ask for, I'd have gotten something less fun.
So, idk. Does that mark me out as some kind of special generalist who will forever be ahead of the permanent underclass because he reads too much HN? Probably not. But it's at least a useful disposition in the LLM era.
[0]: https://github.com/epiccoleman/tripulator/ (AGPL, go nuts!)
Either you fill it with vertical knowledge (which sucks because now AI wipes out a lot of specialists), or you fill it with horizontal knowledge.
Vertical knowledge is the deep silo of learning that comes from doing more advanced techniques to deal with increasingly difficult problems.
Horizontal knowledge is knowledge about process and workflow: where does the thing created by vertical knowledge fit on the bigger picture? What systems are integrated from left to right to keep you successful?
It may be that a successful developer is now a specialist in general technology.
Silos between specialists were always talked about being broken down but the truth is breaking down a silo means asking your specialists to do more work. You don't get to be a developer anymore, you must also take on operations for successful devops.
It may be that AI finally allows broken silos to be successful because AI removes a lot of friction that occurs when context switching between silos of specialists.
Then the question becomes: who is still creating quality with all these new tools?
Because at the end of the day, everyone wants maximum quality for the cheapest cost.
Maybe the course promises more than it delivers. It can still be very valuable.
For some things, mainly practical applied skills, that is the case, but for a whole class of others it is definitely not. Theoretical foundations are one such area. Whatever you do not learn in college you will never pick up in the job. I have may colleagues who were weak on complexity theory or database theory or things like security fundamentals or number theory or even probability theory and they have never managed to bridge that knowledge gap compared to their colleagues who were strong in those things right out of the gate. Make sure to deep dive at uni, that is likely the only chance you get. (No, watching a bunch of youtube clips does not cut it.)