Ian MacDowell

I'm a director of engineering, mostly engineering operations for Unkork.

Unkork is a no-code SaaS platform, basically do enterprise level applications. Well, we enable an operating system that allows our customers to build applications for their customers. So banks and insurance regulated industries. There's a wide variety of, well, highly regulated industries for the most part. I

No, they're building more like point of service pieces, right? So a regional bank building out, you know, an application for just that specific branch, let's say. So it's a no-code drag and drop WYSIWYG platform. You know, you're just dragging things onto a canvas and building out forms. And then those are all tied back into isolated APIs and databases.

It's meant to be a little bit 50-50, where somebody without a lot of technical expertise could build a fully functional web application that is also compliant with all the regulations and standards that they need to comply with. So accessibility,

SOC 2, everything else with that.

Right now we have about 20 engineers.

It's fluctuated and we've had some layoffs, but...

A little bit of both. A little bit of both. You know, AI is definitely coming for us and that's, we're in a big pivot and transition right now. We've got a new CTO, Dave Ferrucci, who's famously the guy that did the Watson program. So he's leading the charge on bringing us up to date with a fully AI compatible and compliant new way of operating through these things. So bringing vibe coding to the enterprise.

Messy.

Especially as somebody in charge of the SDLC.

We're seeing a really rapid transformation into an AI first SDLC.

So really bringing those AI tools right into the middle of how engineers are cooperating, collaborating and building together.

It's changes every week, it seems like, in terms of the tools available to us. You know, switching between models and seeing, you know, vast improvements in time, but also vast increases in cost, like the new Claude46 Opus model is five times as expensive as 4.1.

So navigating those pieces while trying to maintain something that can actually go out and be repeatable. I think that's the biggest challenge we have just industry wide right now is, okay, you vibe coded something. You know, you did with your agent, you made an application or you made part of your application and then I'm taking another piece. And with my agent trying to marry those two things up and have consistency across collaborative channels is a challenge.

I've been evangelizing something called, you know, atomic context, trying to borrow what we've done previously with atomic design systems where you can allow, you know, a large group of designers to basically design with the same look, tone, and timber. How do we do that with AI native prompting?

Right? How do we bring all those pieces together so that as we're collaborating, we're also building something that looks like it was done by a team, not just individuals making their own individualistic parts. So I think that's where an atomic context model sort of comes into play. So what are the facts we're working with? What are the opinions? What are the guardrails? What are the contradictions? What are those smallest repeatable units that we can sort of put into place and build contextual layers up so that those all get included with the prompt that you're giving and the prompt that I'm giving so that everything that comes back has the same collective cohesive set of facts and attributes to it.

Well, I think you end up with an end goal in mind. Right? It's how you get there. I think for the longest time, execution in terms of engineering is what really mattered. Right? The better you were at that execution, the more validity your ideas had.

You know, I've worked almost 20 years now. So I've been in very design led organizations and then very engineering led organizations. And there's always that tension between those two groups. Right? Of like a designer is always thinking about the best case scenario and an engineer is always thinking about the worst case scenario. Right? What are all the edge cases that I need to consider as an engineer that I have to capture and make sure I have in place. But on the design side, you can really just think what's the best path somebody could take to get to the resolution of the problem that they're trying to seek. You can really just stay in that lane now where you're thinking about the idealistic path and allow the AI to worry about the edge cases. And as long as you're telling it and prompting it to worry about those things and you're giving it this sort of a common atomic context that helps shepherd where it needs to stay. Now you can be a lot more creative and fruitful and not worry about the expense of the engineering side because that's actually no longer the bottleneck. You know, the idea that comes first can be quickly executed upon as long as you're being really definitive in how you're explaining and describing the solution that you're looking for. You can become much more focused on a problem rather than worrying about the context of the solution.

Time.

So for me right now, the biggest bottleneck is probably time.

And just in terms of having too many ideas at once and trying to get them all out at the same time and then staying up to date.

You know, there's always a better model that's coming out next week. Right. And how do you re-engineer your workflow and your SCLC to take advantage of that?

Feeling behind is really prevalent in the industry right now.

And then you've got the old guard that wants everything to sort of stay the same and really pull back on that AI piece. I think the speed of ideas is outpacing the rest of the industry's ability to accept and move forward with them. So it's the organizational change much less than the technological one.

I think it's the thing I'm most excited about right now is the art of the possible.

For so long it has been your idea is worthless. It's only if you can execute it and make it a reality, then it's worth something. And I think that's that paradigm is completely flipped to where now the execution is insanely cheap and quick. But it is truly the idea and the way that you can look at a problem and define it that makes it accessible to everybody. It makes just a really exciting time in terms of the fact that anybody can produce just about anything if they have the wherewithal to do it.

The thing that's keeping me up at night is definitely the jobs,

just the industry as a whole. You know, when everybody when a designer is now a 10 X engineer, what does that mean for everybody else? I think you've got a high degree in the industry of oversight of these AIs. So your high mid levels and your seniors and your staff engineers, they're going to be fine for this foreseeable future. But everyone that's a junior, I mean, these these coding tools are eating your lunch. You're never going to be better than them. And then how do you how do you ensure that the A.I. is producing the right thing if you yourself don't know what the right thing looks like?

I mean, this has changed our approach because we haven't been.

We've people have left and we have not backfilled because we haven't needed to. We've got a real skeleton crew compared to what we had a year ago, two years ago.

But through these A.I. pieces now, everybody's producing just as much, if not more, with fewer people.

So so I started Unicorn. Well, I bought the domain Unicorn Graveyard dot com back in 2015. And the thought process behind it was actually I was working for Argo Design at the time. We had other startups that we were working with. They failed because startups fail all the time. We were fine, but they failed. And it hit me like, what do what do you do with all the swag, the IP, the computers, the desks, everything else when a startup dies? So I was looking at domains and Unicorn Graveyard dot com was open. I was like, oh, perfect. I'll buy that. Never had the time to do anything with it. So it sat for years. And then I've recently gotten into 3D modeling, 3D printing and decided to make that the full brand. So it is more of like a 3D lifestyle slash modeling brand that we're trying to build up. We've got a couple thousand followers here and there. It's growing pretty quickly, actually.

So far, yeah. And it's working. Ironically, used AI to build a business plan at the onset because my whole thing was like, I don't want to print and sell models on Etsy or anything else like that. What I'd really like to rather do is just be a designer and design one thing and then scale it. So how do I scale creativity? It gave me an entire six month business plan for growth and opportunities. And I've been following that. And now we're at like 3000 MRR.

Well, I'm at 3000 MRR. It's just me. But in 10 months time and that's that's been pretty cool. And it's the goal would be to fully replace my salary. And if my trends hold up, which they currently are, that is likely within 18 to 24 months.

I'm selling models. So I have a Patreon that people subscribe to. They basically get to download and have commercial rights to then sell the models of mine that they print. They're very unique aesthetic. And, you know, I'm doing something called color stacking, where I essentially take layers of color and make these bands within it to increase contrast and design elements, introducing color without taking erroneous amounts of time and wasting filament.

I have this business without AI. I really would not. It has accelerated so many workflow pieces for me that as a single person operating, I can do it all by myself in my off hours. There's no way I would have been able to do this without the help of AI and just about every single facet from writing copy to design ideas to iterating through, you know, just different aesthetics. The speed that it gives me to, you know, experiment with different ideas means I can do anything I want across this business.