Have you changed the way you learn? Maybe you started off using docs and tutorials and are now an avid book reader or course watcher? Or maybe you do less long-form content now and prefer shorter tuts/docs/screencasts?
And in an age of AI have you changed what you learn? Maybe you now only get more of an overview and leave the details for your AI tools? Or do you still aim for a good understanding of the languages and frameworks you use?
Curious to hear your thoughts!
Honestly Iâm quite lost right now, but I just cant stand books anymore when I started programming about a decade now books were amazing, but thatâs in part because I knew almost nothing, so reading was exciting. Now that I am at a decent pace, I just find alot of books to have fluff in them and its not because I know everything in said book, if I did I wouldnt pick them up, its just sometimes I donât have patience to traverse the chapters, when I really just want something to help me build intuition on the topic, rather than hold my hand. I only want to read books that get to the point
Right now, I just use LLMs mainly chatgpt & claude, mostly the former. Tbh I havenât shopped around for other models or local LLMâs, but llms just strike the perfect balance for me as I can get it to explain things without the fluff, ofc the downside is you have to infer whether the LLm response is a complete lie and as we know with elixir LLMs only have information about the most popular topics, so you still end up looking through other materials like docs etc, you also donât get the authors personal anecdote like you get in a book
I largely avoid tutorial videos , only watch conference talks or podcasts. So for me right now it is
I do wonder if there is a good LLM that can interact with a book, anyone know?
I donât think youâre alone - not because of the content but because our concentration levels have plummeted in recent years. I havenât got time to dig up the article but IIRC it said university students now have an attention span of just 3 minutes!! All thanks to the effect of endless scrolling and the algorithms social networks and apps deploy to hand out carefully crafted periodic dopamine hits (not dopamine hit after dopamine hit, but periodic - as thatâs what keeps you scrolling, and addicted!)
Good news is there is solution - a digital-dopamine detox and I really should write it up as a blog post - but in short it entails deleting all of those apps (inc youtube) and then meditating (just closing your eyes and counting to 20 until you can do it without any distracting thoughts - ideally before reading sessions). It takes around 3 days to reset your system (the first day is the hardest).
Thatâs a deceptively easy and tempting explanation to reach for. But I would argue for many of the knowledge workers itâs something else entirely: we are sick of everyone blabbering when we just want to be told âuse that special argument in that function in the library X to solve your problemâ.
I have struggled with ADHD for a number of years before I finally started beating it, and when I started truly pushing it back, I realized something (that is depressing for many): for a lot of people ADHD comes to existence because they donât want to listen to most people around them (and to most on the internet).
I like books. A lot. But with time technical books became harder and harder to read because it happened no less than 15 times to me that the true insights started at 65% of the book and lasted to about 80%. The rest was either a setup (i.e. âwhy do we need this?â) or meaningless story-telling (âI had this problem when I worked for Oracle back in 1998âŠâ).
Nuance matters. Itâs definitely not all to be attributed to the good old âkids these days have no attention spanâ trope. Most people simply donât have interesting things to say, or write bad books.
I am beginning to think Iâll start asking LLMs to summarize chapters after I finish them and observe if I agree with them. Most of the time I am after concrete knowledge. Deep technical reading is very enjoyable and welcoming⊠but not in this phase of my life and career. I need answers on how to advance with A or B, not a plethora of new paradigmae and philosophies.
Case in point: I need to start memorizing various PostgreSQL incantations because I periodically need them⊠I donât need to be told that SQL sucks and that humanity missed out on something great by choosing it. Cool, very nice, but I still have work to deliver tomorrow.
Iâd be interested in reading such a thing, so long as it takes less than 3 mins 
Iâve deleted all the social apps (was on facebook for 15+ years, instagram for maybe 5). I still have an instagram for my dog. I moved it off of my homescreen which mostly works but I have to just get rid of it. YouTube is the big problem and I actually felt mild panic when you suggested that. Iâm completely addicted.
As for your original question, Iâm still a big fan of documentation and guides. They can of course be hit or miss in quality, but sitting down and reading guides without doing any exercises or anything like that can be really beneficial. You can subconsciously pick up a lot that will come back to you when you are trying to solve a particular problem. This gets lost when you take the approach of always looking up your current problem. Youâre far more prone to asking the wrong question and potentially re-inventing the wheel.
How things have changed for me is that Iâve finally started using ChatGPT as search some of the time, though itâs mostly only when I need a refresher on something I already know. I still have no faith when it comes to things I donât know. I also refuse to pay for it so I wonât get to the point where I rely completely on it (I only get so many queries per day). Is this a good way to go about things? I dunno, but itâs what Iâve been doing.
Ya, as much as Iâd rather read a blog post (and still do) so many posts are filled with fluff, especially in the intro. Itâs been driving me nuts for years and I instinctively start scrolling to find the second header where I assume the post actually starts. This was especially bad when React came out and every single post about some small aspect of it started with an introduction explaining what React is.
This is all true. One more detail is that if you are more experienced you very rarely fall victim to the XY problem. Itâs of course always possible, none of us is without fault but I just started very often seeing people fall victim to something else: âI am so happy that I finally solved this problem, I must tell you ALL about it! Not just the thing that actually solves itâ. And I am just rolling my eyes so hard that I might finally find a brain in that air chamber I call âheadâ.
Ive actually stumbled upon this topic before and I have tried detox, for me it usually takes 1 week from quitting social media, to start seeing the benefits. Also another good method is to find the best study time and length, for me its early in the morning and I can sit & study for ~5 hours pomedoro style. I can only do this on the weekends though because of the day job
But I dont think thats my issue, my main issues with STEM books is they feel a bit like a facade/selling you a dream, reading from cover to cover wont get you ahead, its the experience of trying things out and failing. I just wish every book had like a no nonsense summary booklet or something, that I could quickly read over and then try infer information.
I would love an LLM made for books, that can summarise and give me keynotes. I saw something earlier about an LLM making up bite sized court notes for people on various topics and I tried it with BEAM vm, it was great at making questions which felt like I was actually learning something, it also had various styles of questions from crosswords, to pick & place etc
No, I have not changed. I can only learn be doing stuff myself, and the stuff needs to be a problem I genuinely need to solve. I can never follow tutorials;. I read some just so that I know what are possible.
I agree with you on this , for the longest time I thought I was just being lazy because you go on tech twitter, hn, reddit etc and all they recommend is books , I cannot recall where I heard this from but something like âyou only need 20% of the knowledge, to get through 80% of a systemâ and the remaining percentages correspond to the hardest part of a system or very niche knowledge and yes the numbers are a bit arbitrary , but it rings through to my experience where I can get up and running just looking at a cheat sheet or a crash course
Now that we are on the topic, I need to look for an LLM that can ingest a couple of pdfs and give me the cliff notes, maybe something similar to this: https://oboe.fyi/. I really enjoyed it for the bite sized learning (free tier)
LLMs are just making this worse, IMO. Those things are bad for you. Iâm quite distressed at the assumption by the majority of the industry that they increase productivity, when the only scientific studies (actual ones, not pseudo-science commissioned by the AI companies) have demonstrated they decrease productivity.
I like this article about trusting your own judgement.
Also, as this is about learning, hereâs a recent piece by a bunch of academics taking a stance against uncritical AI adoption in education.
Selected quote from the latter:
The only argument from ignorance that science permits is caution, more research, and care as appropriate actions when something is truly unknown
So most of my year has been watching people drink the kool-aid when it really doesnât seem like it will deliver on the promises, and reading critical thought to reassure myself that my instincts arenât mistaken.
People who donât want to think just found one more huge excuse not to think. Nothing new.
LLMs objectively helped me by accelerating learning and in overcoming the blank canvas terror.
You could be right! But one of the findings from the study above is that people thought they were more efficient, when they were actually less. In the long run, you may actually be worse off due to the LLM usage. Who knows - Iâd love to see more studies into the effects. But I donât think we should take things at face value or from personal anecdotes.
No disagreement here. I am also saying we shouldnât generalize every case.
I do admit that outsourcing thinking is hugely tempting. But if you can overcome that youâll get an assistant that is much better at Googling stuff than you. Just that by itself is very beneficial.
And that does not even mention the cases of ânot sure how to do X with A and B, can you help me start?â â that too carries a danger of doing things in a bad or subtly inefficient way but it does help you not drop a task and just have a starting point in a reasonable amount of time.
All in all, risks exist everywhere and I am admitting that I got tempted to not think for a little period of time, and recoiled disgusted when I realized I was doing it. But again, if you can tame that, you basically get a much better Google search and project / task generator.
I wonât turn this into an AI debate. I just wanted to point out that LLMs are probably part of the problem + share a couple of links that back it up, which is definitely on-topic for a thread about learning that mentioned LLMs multiple times 
I love my chatgpt overlord , but I had to stop using it to write my emails and fix my grammar because at one point it felt like a pain in the ass to use my brain to complete said task and the LLM gave me very robotic/perfect content
Outside of that I think LLMs have boosted my productivity, but I can see how they can become negative especially with the future generation who dont understand the struggle of studying and pondering for knowledge that alone is scary
I have managed to reduce my daily consumption of youtube by blocking the app with the digital wellbeing section in settings and also by downloading the videos and listening them like podcast with yt-dlp. However, not being able to delete the app is still a pity.
And about the topic of this thread, I am increasing my ratio of books vs docs now but mostly is docs and a little bit of blogs for emacs hehe.
The best thing I have found to ask an LLM about a book is NotebookLM. Itâs Googleâs tool that allows you to upload PDFs, websites and then query them as a whole.
The âcreate a podcast from my uploaded contentâ feature is pretty entertaining too! 
Thanks , I will give it a try, I think google is slowly winning the war of attrition
© Copyright Elixir Forum | Terms | Privacy & Cookies
