I run a software studio where AI writes all the code and nobody on the team is a developer. Bespoke software for SMEs. Sharing what works and what breaks.
I run a fully AI powered software development studio I.e. AI writes ALL the code. None of our employees are software developers (but all are technical e.g. actuaries, engineers etc). We mainly focus on bespoke software for SMEs. We have grown revenue 30% MoM for the last 6 months. AMA!
@oldstackjournal Do you think we will get to the point where ANYONE could do this? I agree for simple workflows, but what about things like ERPs or bigger systems needed to run a business? And will they want to actually manage those systems?
It's weird that someone can spend months learning to prompt AI into writing code, and never once think about where that code actually runs. Infrastructure knowledge is what separates builders who ship from builders who stall.
What part of it broke you first: hosting, databases, or production failures?
@omkarships Focus on distribution and the new feature will come from customers asking for them, otherwise you basically building a product for yourself
@GohilHardy Different view from someone whose studio runs on it: our AI spend is a rounding error next to one developer's salary. Prices could 5x and the model would still hold. We are on subscription licenses, not API btw.
@omkarships Same conclusion from running a studio on it: we went deep on one tool and only add another when a client problem forces it. Most of the those crowd are optimising the toolbox instead of shipping. Which two are you using?
@GohilHardy Agreed, although I have become very used to interacting with Claude Code. Something drastic would need to happen for me to need to switch to codex
I run a software studio and we do not hire software developers.
AI writes all the code. The team is actuaries and engineers.
When the code is handled, the scarce skill is understanding a client's problem precisely enough to specify it. Rigorous technical professions train exactly that.
Revenue is up 30% month on month for 6 months.
The next wave of software studios will hire actuaries before developers.
AI writes 100% of our code. No one on our team is a software developer.
What we do have: actuaries, engineers, and 30% month-on-month revenue growth for 6 straight months.
I post about what actually happens when a non-developer team ships real client software with AI:
1. What works
2. What breaks
3. The business numbers
What is the one thing stopping you from shipping software without a developer?
@oldstackjournal Hey! I run a software studio where AI writes all the code and nobody on the team is a developer (actuaries and engineers instead). Building bespoke software for SMEs and sharing what works and what breaks. AI coding + shipping for real clients is my whole feed.
@omkarships Depends who's buying. I sell bespoke software to SMEs and those buyers trust a .com (or their local domain) far more than a .ai. If your customers are other builders, .ai is fine. If they're normal businesses, .com. Who's the customer?
@GohilHardy Agreed, bad output is bad output. But given the current level of SOTA models, bad output usually comes down to the user's spec. Both things are true: criticise the tool, then check what you asked it for.
Here are 5 things I've learned building software without being a programmer:
1. The gap between idea and product is smaller than you think.
The tools exist. The barrier is mostly mental.
2. AI doesn't replace your thinking. It amplifies it.
The people who struggle are the ones who hand the wheel over entirely.
3. Community is the real unfair advantage.
Most breakthroughs I've had came from a conversation, not a tutorial. Find/create your network.
4. Shipping something imperfect beats planning something perfect.
No exceptions.
5. Accessibility is the direction everything is moving.
The future of building belongs to anyone willing to learn how to ask the right questions.