(this is a transcribed version of a post that I dictated - so take this as more of a stream of thought “talk” rather than a blog - but I liked how it ended up style wise so it stays here. But hey, I do need to prep for a talk…)
Have you sometimes felt like you’re talking to the void when you try to convince someone to use a new tool (AI or not) and they just don’t “get” it? I certainly do at times. It feels like we’re living in different worlds. We keep talking about “the future of AI”, “how it will impact engineering”, etc etc, and … well.
It’s my fault. Not the other person’s. The fact they don’t get it – isn’t something they did wrong. It’s on me for not showing them what can be done, meeting them in their workflow. As practitioners, we need to remember this gap exists.
I was a big skeptic for a long time – many years – until a certain point, when I was like … “holy shit”, I can do <X>! I still am a skeptic, but that’s a long story.
I don’t remember the first time I really did it – but here’s an illustrative example from a past blogpost of mine that is around the same time frame (early/mid last year). I’ve always wanted to make games. I’ve tried many times and struggled. I finally got to make one on vacation mid last year following a tutorial. I used AI (Copilot – wished I had agents at the time) to get through all the graphics code, and then I could finally spend time working on game design, the parts I enjoyed – it brought back the fun in programming. I didn’t care about pretty graphics - just wanted a guy on a screen, move up when I press up, etc etc. In this particular case, the code did not matter.
So I had this holy shit moment, right? And I thought I’d show some demos, get other people excited, and try to have them see it. But… that didn’t work.
A thing a coworker told me a few months ago really stuck with me. He said
> You guys keep showing these amazing demos of agents doing ABC, but I’m just trying to do this stupid migration and it just doesn’t work
And he’s right. There’s a huge gap between showing this fancy demo, and getting things work reliably, well, and useful enough that someone wants to use it in their day to day. We also don’t want to just show off crazy stuff – I want to meet folks where they are and integrate into their workflows first. I was able to eventually help that person – but after a lot of work to sit down, understand what they were trying to do, and figure out the right tools they needed. And to be fair, there is a learning curve with AI - it took me a year of practice (and multiple years of failures) before I got better. It's like a programming language - you have to invest in it before it gets better. I wish at some point it gets to the point where it "just works".
This happened again on Thursday. I sat down to hack with a coworker on a migration. I was that asshole, saying “hey, we prefer not federating work, can you try doing it yourself” and suggested he try AI to do some parts. He had! but it hadn’t worked out. So we sat down together.
As he explained his goals and what he was doing manually, I started taking notes – but also, feeding in 2 prompts to claude code. By the time we finished his walk through, claude code had also arrived at the same conclusion. That was his “holy shit” moment. We did a few more deep dives to make sure it wasn’t an accident, and now we’re not federating that work.
I've seen this quite often now. It’s when you see “oh man, this can actually do something – doesn’t need to be the biggest thing, but it saves me a few minutes. and that’s worth it”.
The problem is, once you've had that moment and you've been using AI for a while, you forget how magical it feels. The extraordinary becomes routine. What seems obvious to you is mind-blowing to someone else.
It seems like we’re building this for the 1% of the 1% - the cutting edge folks, big tech companies, the “future”. But what about everyone else? That’s not most engineers. Or students. or people starting out. We have a responsibility to make technology accessible to everyone. This isn’t particular to AI – Jean Yang has a great blog post on this.
We need to make all we do accessible to everyone. Sure, we can and should inspire. But we should make sure we bring everyone along – tell them what the future is, meet where they are, and help build that moment.
So, now to inspiration…
What excites me most is that we're approaching a world where the marginal cost of software approaches zero. If you know how to write code*, you can get software for just the cost of tokens - and those costs are getting vanishingly small.
(speaking from my own experience - yes there’s a lot of tools that let you do this if you don’t know code – but I’m only speaking to my own experience).
As an example. I've been struggling with prioritization lately - too many things to do, too much mental energy spent figuring out what's next. So I built myself a personal assistant. Every morning, it checks my calendar, scans my GitHub, reads my priorities doc, and tells me exactly what I should focus on. But it also does some reflection/advice, pretending to be a wise manager.
I demo’d this last week. It had some funny insights:
> First thing you should do is prepare for the demo/talk you are doing
> Review this PR
> You need to commit the PR for this script you are showing off
> You’re double booked right after this meeting, pick and choose your battles
> You’ve said <X> is a P0 for 3 weeks in a row but you haven’t done it. Is it really a P0?
A bunch of coworkers asked for the script - so I shot it off. But to me, I was sad that so many people were surprised by this. To me – this is just natural. We have all the tools (MCPs) for this. It took me an hour or so of prompting to tweak it to my needs. That’s it.
I brought this up on Twitter and folks had a similar reaction, with folks offering to pay for it. I might do that as my next side project on the pile… but more on that later.
It makes me sad that these aren’t accessible to everyone. I'm building more tools like this all the time. The bottleneck isn't coding anymore - it's literally just thinking of what I want to exist.
As another example, from winter last year. I noticed many friends struggling to find jobs post-layoffs while other friends were desperately hiring. In under 10 hours, I threw together a job board to connect them. I’m quite confident saying that this project wouldn’t exist without AI. I blogged about it before too. It’s not that the project was hard or impossible. I just didn’t have enough time to sink into it otherwise.
I have 3-4 side projects in flight. I hope to share more soon when I have time to polish them off. But they all illustrate the same thing. The bottleneck is no longer typing. It’s ideation.
(fun fact: on Thursday a coworker saw me at my desk – asking me why I of all people have RSI gloves on. Shouldn’t the AI be doing all my typing? But – it does! it lets me focus on what I cherish. Writing blogs, slack messages, emails – focusing on talking about the technology and the work, sharing best practices, rather than “just” typing out code. Writing, though, is sacred to me. That’s mine.)
Back to the oh shit moments and implications for the future.
On Friday I had another major "holy shit" moment with something I'm building. (Can't share details yet, but soon. If you’re at $DAYJOB, expect to hear more about this Monday - you know where). After a couple of years experimenting, these still happen to me. It’s brought the fun back in programming.
But I’m still worried. We're creating a world of haves and have-nots. Those of us who've figured out how to wield these tools are building personal assistants and custom software on a whim. Meanwhile, others are giving up because their AI can't reliably write a file.
The era of personalized, custom software has arrived, but only for some of us. How do we bridge this gap? How do we get everyone from "this doesn't work" to "holy shit, I can build anything"?
That's the challenge that keeps me up at night. We can't just focus on the frontier. We need to bring everyone along.
If you made it this far, thanks for reading. I'm curious - what would you build if the cost of software was essentially zero? And more importantly, what's stopping you from building it today?
PS: You may also like my friend Maisem’s post – it’s along similar lines – we were chatting and ended up writing similar things in parallel – I spoke to my experience with software engineers, but he’s really looking to change the world and make it easier for everyone.
PPS: Between the time I wrote this and the time I published it, I saw Jean Yang's latest post on why AI will create a lot more develpoers. It's great and you should read it.