AB

I used to think that religion is a problem, but recently I realized that religious people are the real problem. Religion is fine, providing we see it in its context. But religious people? They sometimes scare me!

What used to be an asset has overnight turned into a liability. Those developers who are willing to roll with punches will stay on top of the game. The rest will go back to farming, making sourdough bread, etc.

I keep hearing how AI will never be able to replace skilled software developers. I guess it really hinges on the definition of "skilled developers". Right now the skillset is drastically shifting.

Science is eager to prove itself wrong.

Religion is eager to prove itself right.

Similarly, the ushering the age of of AI-generated software did not inaugurate the "on demand" software. It will take some time before me mature to the point of being able to use the "on demand" software technology.

The invention of the printing press did not inaugurate the era of "on demand" printing. It took a long time before the printing technology reached the maturity to enable the "on demand" printing.

Wow, ICE agents are now doing door to door milk delivery!

AI gave me accurate, but silly answer to my silly question:

Just because you don't know how to use AI to generate good solutions doesn't mean everyone else in the world also doesn't know how to do that.

Turns out it is literally impossible to do those two distinct things at once. Something's gotta give.

Once I discovered TDN, everything cleared up for me.

In the earlier days (my pre-TDD days), I was always trying to accomplish the impossible task -- write code that will function as expected while at the same time be structured as expected.

I find it very beneficial to severely segregate the concerns related to writing code from the concerns related to discovering what is the code actually going to model.

I find that the more I'm doing Test Driven Navigation (TDN), the more I'm starting to discover things that would otherwise remain hidden in the background.

Religion tells people what they want to hear. Science tells people what they don't want to hear.

No wonder religion is more popular than science.

Learn how to write reliable tests. That is the number one skill that will keep you afloat into the future.

I know of a number of mission critical businesses who deploy code changes many times per hour. Do they have teams of Marvel superheroes who are capable of opening and reviewing PRs in a matter of minutes/seconds? No! They have teams of skilled engineers whose focus is quality tests.

I have less confidence in human reviews than I have in my tests. Of course, the clincher is this -- the quality of tests. Many (most?) software developers treat tests as an afterthought. Some even scoff at tests. Few understand that test are way more important than the implemented code.

I always start from a failure. Then, I seek methods that will make the failure go away. Whichever method does that the quickest way is the method I'm going to adopt.

It is actually very simple. Seek quick feedback (ideally, instantaneous feedback).

We're all much more focused on a fully qualified success when engineering our systems, and leave the failure handling as an after thought.

Therein lies the biggest impediment to adopting AI into the software development discipline.

The only way to effectively leverage AI when developing software is to always start from a failure. Adopting the "let it crash" mindset seems utterly counterintuitive to 99% of software professionals I've had the pleasure to work with.

The biggest problem with current misunderstanding of the amalgamation of AI with software development teams is in failing to understand that failure is the first class citizen, it is a centerstage of software development.

I applaud the acumen of AI-resisters. They know that without serious ramping up, they will be up the shit creek without a paddle if they would agree to use AI.

AI is a double edged sword. If you are not sufficiently competent, it will eat you alive.

An AI's job is to take the failing test and make it pass by generating some code that will satisfy the expectations encoded in a failing test. It is the job of a human developer to create a failing test.

In the same way, I don't feel I own the code generated by AI. If there is a defect/bug in the generated code, my responsibility is to tighten the prompt/failing test. I never feel like I must roll up my sleeves and fix the code generated by AI.

I never felt that I own the binary code generated from my source code by the compiler/interpreter. If there is a defect/bug in the compiler, I do not feel responsible for fixing it. I pass it on to the vendor.

I really don’t understand why does machine written code need to be human readable. Help me out here.

Look, guys, this is not my second rodeo.

Question: I sometimes here the phrase "giant solution in search of a problem", but I'm not sure what that means. Do you have an example that would help me understand the meaning of that phrase?

Answer: Spec-driven development with AI.

Here is an idea: organize an AI-resisted coding campaign to align all software professionals who are proudly resisting using AI for creating code.

I think it's wrong that only one company makes the game of Monopoly.