Hi {{first_name|dear builder}}!
My friend and fellow entrepreneur Andrew replied to last week’s email with a question that has been keeping me up at night too:1
I am curious for your take on building software. I spent the holidays learning Claude code, and I am starting to see "software for one" as the new paradigm. Like you did with your website, I'm building tools and systems for myself with negligible overhead and efforts. It is truly insane. At the same time, I've been thinking about building software to sell to clients as complementary to our content IP. But I get this nagging feeling that it is a waste of time because eventually people are going to have exactly the software they need for themselves, and monolithic software will become outdated. What do you think?
In the same way ChatGPT was mindblowing in 2022, coding with AI today is a “truly insane” experience. All it takes is ten minutes chatting with a coding agent to have your world shaken about what is now possible. Here’s a representative example of someone making a Trello-like app in 20ish minutes - worth checking out if you have not yet been hands on with these tools:2
But while it is easier to build software than ever before, it is also scarier to build now because the whole landscape is in motion. If it’s this easy and cheap to build your own custom software, what’s stopping each and every company from building their own “software for one” as Andrew put it? Why would they pay someone else for their software, and then - gasp - also pay a bunch of consultants to configure that software?? This is a big question with big implications.3
The most common response is trying to evaluate if organizations can make their custom software truly enterprise-ready, sustainable and secure. But I’m not sure this is the most useful question. This is a technical challenge, and we can expect the AI coding tools to get better and better at handling it.
More interesting: the adaptive challenges at play here. The leadership professor Ronald Heifetz has found from decades of research that when companies fail, it is usually because leaders have misdiagnosed an adaptive challenge as a technical problem.
Technical Challenges:
The problem is known, easy to identify, and often recognizable based on experience
Similarly, the solution is known based on experience and expertise
The locus of responsibility is with experts and authorities
Obstacles are usually resource-limited, i.e. time and money
Adaptive Challenges:
The problem is often unknown or hard to identify; tied to deeper patterns or dynamics and requires learning
The solution is unknown, which again requires learning
Responsibility is with those affected by the challenge (stakeholders), including authorities
The obstacles are less tangible—hearts and minds, values, loyalties and relationships4
Up until now, software makers have focused almost entirely on solving technical problems. So much of business software serves IT or finance department needs first, with everyone else an afterthought.
Think of medical software emphasizing billing codes rather than holistic health notetaking. Think of sales software oriented around “configure-price-quote” rather than ongoing customer relationships. And because it is fresh in my mind - think of synagogue management software with robust deferred revenue tracking but lacking in pastoral care features or a decent interface for program registration.
Perhaps this was by necessity: these were the most pressing problems to work on (or seemed that way at the time). Perhaps it was by pragmatism: the IT or finance departments were the ones doing the software purchasing. At any rate, we appear to be entering in era in which AI is increasingly able to eat up those technical problems for breakfast. And I’m here for it!
So to bring it back to Andrew’s question: I think it is an extraordinary time to build software - IF your software is attempting to help solve an adaptive problem, not a technical one.
The new ease of software creation allows thoughtful makers like Andrew to better help clients with adaptive challenges, because Andrew can make software that expresses his team’s special way of seeing the world. Andrew can incorporate into his company’s software their philosophy, content, and workflows. His clients won’t be able to replicate that themselves - they can only access the same AI everyone else has, plus their own internal knowledge. When it comes to adaptive challenges, we need help from those outside our own system to help us understand ourselves clearly.
Software has been eating the world for a long time, and it is poised to devour it as small teams and individuals are now able to create their own software for exactly their own context and preferences. I absolutely think “people are going to have exactly the software they need for themselves, and monolithic software will become outdated” as Andrew is wondering about. And I think that is a healthy, democratizing development - more tools to the people.
But none of that is going to solve the adaptive challenges we all face at work. And this is where I think the real opportunity lies.
So Andrew, yes: build your software! But build it as a frame for your approach, not just as a technical solution. The value is in how the software embodies your team's particular understanding of your clients' adaptive challenges.
What does that look like practically? Your software should use language that reframes how clients think about their work. It should surface insights at moments when they're most needed. It should encode workflows that reflect your insight about what actually works, not just what's technically possible.
Your clients can't build software that helps them see their own blind spots. They can't build software that challenges their assumptions about how things should work. They can't build software informed by patterns you've observed across dozens of similar organizations.
The monolithic software makers will struggle not just because everyone can now code, but because most of them were only ever solving technical problems in the first place. I’m hopeful for a wave of new software that helps people do the much harder work of changing how they think, work, and relate to each other.
Building to help people with their adaptive challenges is never a “waste of time” - it’s one of the most valuable things you could work on.
Tool of the Week
Last week I highlighted Lovable, this week Replit - a similar competitor but the one that we’ve chosen to use for Welcome Back. We found it to be more robust in terms of getting closer to a deployable application rather than staying at the interface level - BUT that comparison shifts all the time as these tools are neck and neck with each other. Try them both!
Surprising Male Vulnerability Moment of the Week
I was struck by this post from Andrej Karpathy, one of the co-founders of OpenAI, and the person who coined the term vibecoding (less than a year ago). Even one of the world’s top experts on AI feels behind!
1 Literally…I wrote most of this post when I was awake at 4 AM one night last week after giving my 9 month old a bottle
2 Hi mom!
3 Including my ability to pay my mortgage