Why I No Longer Want to Write Software for Companies That I Don't Own

I joined a new project recently after working in a mostly relaxed agile environment for most of my career. I would describe my current project as "Agile at all costs". No testing, no requirements, just go. Just do. I have to say it's created an environment where almost every developer is super unhappy and ultimately producing a sub-par product.

I want to go back to just doing good work. Obviously focusing on what the customer needs and wants, but where the engineers have a choice to do something the right way instead of just the fast way.

The problem I witness when the general goal laid out is improving "velocity" tends to be a fundamental misunderstanding of what velocity is.

Namely, velocity is speed plus a direction, and most of the activities that would normally help codify a sane and deliberate direction (i.e. rigorous requirements, rigorous design, rigorous validation, etc.) are all primed to be sacrificed at the Agile altar. What you end up with is just "speed" as a priority.

When organizations have KPIs & OKRs which set "velocity" as a prime directive what that usually means is just "speed". As long as your development cycles are shorter and shorter, then you're rewarded. In these kinds of organizations the incentives align with flailing around spastically, so long as you're doing it 1x a month, 10x a month, then 100x a month. Taking the time to be exactly right 1x a quarter, and then building on that, isn't seen as desirable.

As one would expect, there aren't an enormous number of people who are actually satisfied by doing random crap for a random cause as fast as possible. In fact, that's what machines are for, not people.

And the follow-up problem is that most Agile proponents say that "estimations aren't to be taken as timelines". But that doesn't address the fundamental issue that management always uses the engineering estimations as a method of managing their timelines.

Not to mention that "velocity" and burn down charts actually reinforce the idea that estimations are ways of tracking how much work will get done in Agile. Otherwise, tracking those metrics would be completely pointless.

So while I totally agree with this, I have some questions:

First off, do you think Agile is better than Waterfall? I personally think that it delivers a product faster, but I'm not sure if it actually has a macro improvement on the job process. I think waterfall probably addresses your developer-centric issue of satisfaction better than Agile does. I think we may have thrown away too much with the anti-waterfall hatred our field has adopted.

Second, what alternative do you propose? I think Agile is a problem, and ultimately a pretty bad process for organizing. The overhead is so expensive. The most effective and efficient my team has ever been was when we just quit Agile and wrote what we had going on up on the white board. And you know what? I've seen this work better than Agile in a wide range of settings. "Self organizing" really seems to be the key, but regardless of the self-proclamations that Agile is self organizing, it really seems to shoot itself in the foot with extra mandatory process.

Finally, it seems like all of the good solutions don't scale. Do you think this is a problem of trying to scale problem solving? I don't think I accept the argument that this is just the cost of working at larger scale, but I haven't thought of a good way to keep a team organized and still get work done without an extreme amount of organization and process (Agile) or a really involved team lead/product owner---who will slow everyone down by needing a lot of touch points.

I always understood agile to mean "as a business, have as short an OODA loop as possible, don't waste time making detailed plans until you have detailed information, and update your plans as new information comes in". This is why it gets compared with waterfall development, where massive chunks of work are planned anywhere up to years in advance.

At this point, nearly everyone who writes software is well aware that designs and specs you created 6 months ago might be out of date by the time you need them, so there's not really any need to evangelise for that point of view, and agile becomes a buzzword because everything is already agile.