3 ms·
The loop of programming "skill" is the path a craft takes as it moves to mass production. To me, much of programming is building custom automobiles by hand. E
by bmh100 12y ago
The loop of programming "skill" is the path a craft takes as it moves to mass production. To me, much of programming is building custom automobiles by hand. Eventually, tools and processes will become standardized to the point that true mass production is possible. Notably, powerful frameworks and changing technology stacks continue make problem solving skills and learning ability valuable. But what happens when our frameworks are so powerful, so stable, and so well tested, that we simple do not need craftsmen? The industry will move to low-skill "coders" (assembly line workers), line code supervisors (assembly line supervisors), framework engineers (factory engineers), framework component engineers (machine/robotics engineers), and programming process managers (operations managers).
- nly 12y ago... and the occasional guy/gal who still welds together scrap and builds kit-cars at home.
- thewarrior 12y agoThats a terrifying future for a person like me who has just started programming. But get this , programming can never be fully automated. There will always be one guy who has to sit down, understand the requirements, convert into a logical structure/business rules and then design a proper solution. Maybe a powerful framework will then allow this one person to realize what he could have otherwise done only with the help of 3 - 4 other junior developers but until we develop true AI programming won't be automated.
- k__ 12y agoWell, the boring programming jobs will be automated, but for them you didn't need a CS in the first place.
- thewarrior 12y agoI'm a coder at an outsourcing firm in India .. So gulp , I am a bit scared :P
- k__ 12y agoMaybe you should be :D Joke aside, I thought India was getting more high-tech lately. My impression was that this outsourcing would stop soon anyway because the wages getting higher and India needs their coders for themself.
- thewarrior 12y agoWages are actually lower when adjusted for inflation when compared to what IT workers used to make in the early 2000s. Also theres increased competition from Ukraine / China. But its hard to find better work in other sectors and the less said about the standard of education here the better. I really enjoy coding and I'd say that I'm better than 75 % of the coders here. But thats hardly enough to get into a place where they do the real nice work. We even have to sign an agreement whereby we're forced to work for a particular company for years and can't leave without paying a hefty fine. Atleast that never happens in North America. Good jobs are hard to come by and I'm thankful for this job.
- vezzy-fnord 12y agoThat sounds like a very distant possibility to me, bordering on a pipe dream. We're at a far cry from having the tools that enable us to do this equilibrium of power/stability/testing. The likeliest scenario is that we will keep iterating over the same flat tire that sort of looks like it reaches equilibrium, before it inevitably collapses under its own weight and has to be restarted, again and again. I do not see this pattern going away.
- Karellen 12y ago> Eventually, tools and processes will become standardized to the point that true mass production is possible. I like to think that. However, to strain the programming-as-craft analogy (which I think is a good one), I don't think we're yet at the point where we even have a good selection of well-made hand-crafting tools yet. We've probably got the equivalent of a reasonable axe, hammer, and nails by now (matching these to current day compilers/linkers, interpreters/vms, and editors/IDEs is left as a futile exercise for the reader) and we might have the vital core of some other tools, but I don't think we have well made selection of saws, planes, drills/bits, squares, clamps, levels, etc... yet. We're getting there, fast, but considering the craft is only about 170 years old[0] we just haven't been at it that long. [0] Starting in 1843 with Ada Lovelace's method of computing Bernoulli numbers on Babbage's Analytical Engine.