Coding is kind. The dichotomy is based on feedback availability, quality, and delay. While the environment does change over time, it continues to be kind: 99% of the time, you're able to hit compile and see instantaneous, accurate feedback. Doesn't matter which language, framework, or standard you're using (except maybe Malbolge). Even debugging is kind save perhaps for heisenbugs, but that's a bit of a stretch to declare the whole field a wicked environment.
To a first-order approximation, the length is 1/4 wavelength per leg. However, that assumes an ideal conductor in free space. We rarely have the luxury to suspend an antenna a few wavelengths from the ground, water, buildings, and anything else remotely conductive. The way the antenna interacts with those things in the near-field affects its impedance and makes it behave electrically longer.
The common formula you'll see is: Total Length (in feet) = 468 / f (in MHz). If we suppose exactly half wavelength and do the unit conversions, we would expect L (ft) = 492 / f (MHz). So why do the calculators use the shorter length? It's an empirical compromise. Taking 5-10% off total length is generally what's needed to account for things in the near-field. The 468 number has been repeated enough that it's stuck. In practice, I almost always cut dipoles for a full half-wavelength, hook them up to an antenna analyzer and then trim them down. With as cheap as hookup wire is, I'd rather not take the risk of being too short and having to field solder a splice (not fun on ARRL Field Day).
There's not really a general formula for finding length. The physics is nothing more than Maxwell's equations, but many of the deviations from an ideal dipole come from interactions with the environment. It's difficult to measure and/or predict how the environment will behave, so you're often better off building the antenna and then adjusting it in place. And so we end up with empirical rules of thumb like L = 468 / f.
Antennas are one of the harder topics for amateurs, for sure. The theory is well-understood, sitting somewhere at the intersection of EE and physics. I'm lucky enough to have a strong background in both fields, but there's a clear lack of curriculum for amateurs without that background. This website at least seems useful for building intuition, so hopefully it helps you some.
There are a handful of products for pets at least. I'm personally familiar with Whistle. It's collar-mounted, works fairly well for a large dog in a rural area despite spotty cell coverage (which the software doesn't handle terribly gracefully). Amusingly, the monthly subscription is only a third the price of the product in the OP.
For livestock to be a viable market, the monthly subscription would have to go down substantially. In the US, your large market segments are poultry, pork, and cattle. Chickens just aren't worth enough. Hogs are $75-150 live (with thin profit margins) and are a pain to catch, though I'm not familiar with escape rates in commercial operations. Cattle are worth 10x that and do escape occasionally, but they're not terribly difficult to find nor catch.
Perhaps there's a market for tracking hogs for small farms and cattle for commercial grazing operations, but I'd be surprised if tech could do much to disrupt the existing low-cost solution: ear tagging and a phone call when your neighbors/police find a cow on the road.
http://blog.rongarret.info/2009/10/catalog-of-wealth-creatio...
From this comment tree on the original discussion of pg's "Frighteningly Ambitious Startup Ideas": https://news.ycombinator.com/item?id=3687080
The whole tree is definitely interesting in addition to the blog post linked. Also, a follow-up post a few days later: http://blog.rongarret.info/2009/10/wealth-production-mechani...