Posted by tylerdane 18 hours ago
Literally an hour ago my 7 and 9 year olds came to mom and I and said, “thank you for disabling YouTube. We are having so much fun.” I know this sounds like a fake Donald Trump kind of story but I swear to god it just happened.
How learning and doing aren't exactly the same and that you need to get back to it many times rather than doing a lot at once.
It's ofc nothing new and the same principle as for example spaced repetition.
I had a student come to me with essentially the same problem over two years and each time I helped her she was in refusal to listen as she stressed herself to just make it work now. Her problem was that she never took the time to do the basics and rejected any learning opportunity as it stared her in the face.
You get results over time if you dedicate yourself to just doing the thing. For many subjects there is no shortcut, no way to walk the path without actually walking it. Every time you encounter an issue there is a learning opportunity. Use it.
You follow a tutorial to do something, feel happy about it. Then you start a new project to put your new skills to good use and... Blank. No idea where to start, no idea how to proceed.
It's so important to build stuff, using references is fine, but following tutorials is not the way forward! You have to work on your own without the training wheels.
In my experience most people can do this, if they think about it a bit — identify the thing they want to learn and find a tutorial for it. Which is amazing, really; this sort of meta-knowledge is a remarkable human concept.
For me it is. Even in my domain where I’m an expert and it’s fun, it only is if I’m working on something interesting.
Sometimes distraction is the main issue when it comes to having ideas.
I get stuck into this mentality of "I need to learn and master X, Y and Z before I can even start building my dream"
Would be much better served by just building whatever and learning the skill
Once you're a little more confident (you know a bit, but not much) I suggest to modify the tutorial as you follow along, that makes the tutorial harder and gives you small challenges to overcome while still giving you general guard rails.
Then as soon as you're dangerous enough to be let loose you should pick your own projects that are slightly above your skill level. Maybe try different approaches if you're unhappy with the first result.
When I wanted to improve my comic drawing skills ca. 2009 I started drawing and publishing a daily webcomic strip for a year. That really helped.
But tutorials remain useful even if you're advanced or a pro. E.g. if you use blender a lot and a new feature comes around watching a tutorial on it is a very efficient way of getting up to speed. Of course you will watch tutorials differently from a beginner, you will pick up on different things etc.
The best way to learn is a serious project with a deadline, but if you have that deadline it will make you wish you had watched some tutorials when you had the time. Source: I teach this kind of stuff at the university level for 6 years now.
You absulutely CAN meaningfully pickup things in a day or two, especially with modern AI agents. 3D modeling is a good example, it is not that difficult! It takes some preparation not to be blocked, and good hardware, but when you actually start it goes fast.
You need a concrete goals, not some nebulous plan to learm one hour a day for years.
Do you people have to mention AI in every single subject.
In the case of 3d modeling, it did initial research, prepared software, prepared a few prototypes to kick start, prepared validation checklist, and found some tutorial videos for me.
I am learning a bit of 3D modeling in Blender so I can mod games that I like (just for private use), I do get stuck sometimes on the silliest things and Blender docs don’t help, but neither did LLMs tbf when I tried to troubleshoot issues with them. I wonder how I can make it a bit less tedious.
My goal was to create 3d shapes out of math curves. LLM wrote bunch of scripts, that generated 3d models for me.
Usual problem with LLM is that it needs good grip and traction on problem, to actually work. It is fine with text and code, and images, not so much with 3d objects.
Not sure whst is your workflow, but perhaps give llm ability to see rendered game without your help. It is a problem with integration and automation.