Back from the Void with reflections from AI
Wow! It’s been a long time since I have written anything here. Mostly, because I have gone back to working full time. Life happens.
I won’t bore you with the centuries worth of history of what has been happening in the space in these… less that two years. Suffice to say, that it was a lot.
But I do want to make a couple of remarks.
Is (gen)AI good or bad?
I won’t try to discuss this much, and if you are not happy with that, you can close the tab. I can give you a few (old) links though. There are lots of vocal people on the internet with thoughts on this. They are much more informed than me and they can make much more nuanced arguments:
2025-06-04 - I Think I’m Done Thinking About genAI For Now by Glypgh
2025-05-02 - My AI Skeptic Friends Are All Nuts by Thomas Ptacek
2025-05-28 - I am disappointed in the AI discourse. You can listen to Steve talk about that blog in: Broken AI Discourse with Steve Klabnik (Software Unscripted podcast) and in AI Discourse with Steve Klabnik (Oxide and Friends podcast)
2025-06-11 - TDD, AI agents and coding with Kent Beck. I agree with Kent’s sentiment on this podcast:
In 52 years of programming, this is the most fun I’ve ever had.
And with this post on x.com:
I’ve been reluctant to try ChatGPT. Today I got over that reluctance. Now I understand why I was reluctant. The value of 90% of my skills just dropped to $0. The leverage for the remaining 10% went up 1000x. I need to recalibrate.
I would be remiss if I didn’t mention ASDP: The podcast Episode 239: Claude-Poisoned Dev Sipping Rocket Fuel. I like this podcast because Conor is of course a fan of these tools, and he sometimes asks Ben (probably on Episode 246: Not High on AI?) who bluntly says, and I am paraphrasing of course, that he likes to work on things that LLMs can’t do. I am with Conor - these tools do feel like rocket fuel. Does it make me sad that an LLM can spit out better code than me, faster than me? Well, we are not all working on embedded systems like Ben :) - lots of people, like DHH says (again, paraphrasing), are just moving data from web page to databases. Massaging information.
So yes, there are lots of moral and ethical implications. Yes, they acquired illegally, or semi-legally lots of training data. Meta for example is still fighting this in the courts, Anthropic settled, etc.
Yes, these tools can be dangerous. AI psychosis is now a thing. I would never imagine of taking any moral or ethical guidance from them. They are trained, besides lots of pirated books, on the whole nasty internet. I believe, and still notice, there is nobody really there. It is a fancy autocomplete with LOTS of bells and whistles. I don’t really care about what they have to say outside spitting out code and summarizing information, etc… I acknowledge that is a selfish and very narrow point of view.
Conclusion
If you listen to the folks at Oxide and Friends - just choose by the title, you’ll see that they are an incredible technical focused crowd, building the software for their computer company - LLMs are widely used by them in various effective ways.
Of course, there are many other examples.
Tools like Claude Code are incredible enablers and accelerators, as I hope to show in the next few posts. Things that people would never build because of time constraints now can flourish and exist.
You might have hear the term “scratch your own itch” when referring to building small tools, or making particular contributions to open source projects. I expect that today it is easier for these things to happen. I was never fully satisfied with the time-tracker apps out there. I wanted something a bit better than a Google Spreadsheet. A couple of years ago, it would have taken quite a lot of time. Now, in a few days I built the small tool I wanted… and happily use it every day.