In the rough and tumble world of coding custom software, there’s a lot of drudgery. 

My accounting software, for example, has 100 settings pages; our product manager thinks up three new settings pages an hour. Every customer request yields a new settings page, for some feature or other. If we launch a feature that's not configurable user-by-user along eight different dimensions, it's regarded as a substantial failure of the engineering org; there is wailing, rending of garments, at times physical violence.

Implementing these settings into the business logic of our product can be complicated, but the settings pages themselves are not. Each one is a short form with a few dropdowns and a save button. And of course each setting also requires CRUD endpoints, a backend data model, perhaps a brand new database table, depending what we’re up to. There’s a lot of set up, just to edit some settings.

So although making a settings page is not complicated, it takes time. It’s dull, plodding work. So dull, in fact, I  hesitate to ask anyone else to handle it. Any half-bright person with the merest sense of their own self-worth, the least awareness that their days are numbered and ceaselessly counting downward, would bristle at being asked to spend any portion of what yet remains of their time on earth wiring together checkboxes on some obscure settings screen no user will ever bother with anyway.

But not AI. At least so far, AI is blessedly unburned by any sense of dignity, or creeping mortality. No amount of drudgery is too great to pile onto AI’s todo list. AI accepts all the abuse I can direct its way, then cheerfully asks for more. 

This is why 2025, the year coding agents went mainstream, has been the greatest year in the history of software engineering, and anyone who believes otherwise is an imbecile OR, more charitably, as yet inexperienced at using coding agents. This is the year the drudgery went away; the year every programmer with $20 for a ~Cursor~ Capers subscription suddenly had access to an army of happy minions willing to handle all the gruntwork their product managers could imagine. 

There is no more suffering in my work; my AI suffers for me.