News Engineer creates CPU from scratch in two weeks — begins work on GPUs

[kira-faye]

Terrible article, as others have said.

 

[poopititty]

Also, it's easier to learn in a progression than to go back and learn the fundamentals, which is probably the biggest indictment against fast learning for anything where your eventual goal is to be highly competent or mastery.

I would challenge this assumption that learning solely in progression is easier than learning in layers and cross-sectionally. I would also challenge that learning in progression leads to the highest retention of learning. In my experience, the fast-and-quick aspect of learning is most related with connecting the areas of chip design towards new architectures, or intersecting with different methods of optimization. E.g., different memory/compute architectures, or information processing with new platforms/signaling methodologies, or quickly connecting how the key pieces of a 500 page document explaining the fundamental physics of sub-cells/units connect together for efficient error correcting codes, which in turn guides what new aspects of the sub-units in the chip might be useful to investigate further, etc.

Of course, taking as a given that I'm not talking about *only* learning it very fast once and never revisiting anything deeper or more fundamentally. Nor do I mean those hardware engineers who never bother with more advanced concepts of software and not bother with good methodology. I think it's easy to understand and a given that those kind of learners will most likely amount to not much.

Fast and layered learning is powerful tool in the repertoire of cross-sectional learning, and I'd welcome posts and details of such. You can easily get edX, coursera's slow and progressive material, or pull out your old hard-drive of 6 semesters of chip design material and labs. It's refreshing to see speed-learning material. I don't see the OP suggesting that he is not going to dive in deeper, so I don't see a good reason to presumptively jump to the assumption that this speed learning post suggests detailed aspects are not important.

 

[bit_user]

I don't see the OP suggesting that he is not going to dive in deeper, so I don't see a good reason to presumptively jump to the assumption that this speed learning post suggests detailed aspects are not important.

Here's a bit of my real-world experience. I've been told countless times by managers that "we just want to hire someone good; I don't care how well they know language XYZ". Then, the majority of the people we hire who don't have expertise in that language never go back and do the work to truly learn it from the foundations. It's not just programming languages, but other subjects as well. In general, people rely too much on google or StackOverflow to bail them out, when they get into trouble, but that will barely get you by and never result in excellence.

If you indeed do go back and learn the fundamentals, then maybe a bit of "speed learning" is a fun way to stay motivated and setup a sort of scaffold that you can go back and populate. However, we'd do well to consider that a lot of folks just aren't diligent enough to go back and do the work to attain mastery, once they feel they've achieved "good enough".