Hello, "Theodore Tso" <[email protected]> wrote on 24/07/2026 at 16:06:24+0200:
> On Fri, Jul 24, 2026 at 11:40:30AM -0500, Pierre-Elliott B??cue wrote: >> > That's something I find very interesting with AI-assisted software >> > engineering: it puts specification, documentation and testing back at >> > the center of the picture. >> >> It has always been at the center. > > It may have been at the center of *your* workflow, and if you're > lucky, your employer's mandated workflow. However, for many > developers, and for many startups, the temptation to "move fast and > break things" is incredibly strong. > > To the extent that this incentivized people to create more unit tests > and define interfaces between components (which is generally required > for good unit tests; otherwise you may be forced to create integration > tests, which take a lot longer and burns a lot more CPU time, more > energy, more water, more CO2, etc.) Right. Even when having to go fast, the job still is about specification and architecture. But indeed, doing a great job is far harder in constrained timeframe. >> I'm also really terrified by the cognitive decline that seems to arise >> with these tools. > > I'm worried only if any tool start out being cheap (for example, > $200/month for $2000 worth of AI tokens), and then it suddenly becomes > more expensive, people who started relying on any tools, only to have > them suffer a rug pull, it might be highly problematic. > > On the other hand, it's unquestionable that the fact that calculators > are incredibly cheap (a four function calculator runs for maybe $3 > USD) has meant that there has been a decline in the ability of people > to do math in their heads. I think this comparison has some merit, but is not a good one anyway. Giving up a specific skill is not the same as giving up your ability to think, learn, and criticize the knowledge you're receiving. These are fundamental human skills, which take time to develop and master. The fact that now, people have access to potentially unlimited knowledge and reasoning ability even before having developed their ability to check/criticize/ponder makes them both vulnerable to a rug pull, but also to be fed garbage without ever realizing it. > As another example, compilers have *definitely* resulted in people no > longer as good at writing assembly language. There were some Unix > old-timers on a mailing list recently reminsicing about how way back > when, it was *assumed* that human-written assembly would always be > more better and more efficient than compiler-generated assembly --- > and these days it is pretty much self-evident that compilers do a > better job than what humans can do. Part of this is because how much > more complex modern CPU's have gotten compared to a PDP-11. But I > think it's more about people being less proficient at certain skills > that are no longer as important, and less about some more general > claim about "cognitive decline". Considering the vast field of applicability of Generative AI, I fear that it's not about losing certain skills. Smart people will get and be smarter. The others. Oh well. I wish I'm wrong. -- PEB
signature.asc
Description: PGP signature
