Vibe coding looks quick on social media. Today I was on day 12 of my 30 day AI experiment, and I want to give you the version that includes the slow parts.

The experiment is simple to state. I am taking a business concept through to profitability in 30 days, using AI for 90% of the workload. I am working with two team members, and each of us has taken a different role. I should be clear about the outcome. I do not know yet whether it will make a sale. What I do know is that by day 30 we will have a software package ready to sell, and I hope we will be ready to take the first subscribers.

Where we are on day 12

At the start of the week we had 29 modules left to build for the product architecture. I worked out that I needed to complete three or four modules a day to finish on time. Yesterday I completed four. Today I completed three, or close to it. We are on track.

The modules are not equal. Some are quick to box off. Others take real time, thought and set-up. That is why I track the count each day rather than relying on a feeling that things are going well.

What I built today

I worked on the financial side and on task and project management.

The transactions module lets you sort and search through every transaction, and you can make amendments without going into the main database. There are about 8,000 records in it. We pulled them in with no effect on the speed or function of the web app. AI metrics sit on top of it. It took a couple of hours.

The accounts module shows the bank balances for every account the data comes from. It also has a live feed with an alerting system, so we know if a feed stops working. That took about an hour.

The third piece was external contractors. The web app is complex and has a lot of data in it. It works on mobile, but someone who is not comfortable with technology will find it hard going. So I built a Slack route. A contractor, once onboarded, asks an AI Slack bot for their to-do list and sees it in a Slack channel. They post updates on each task by typing messages back and forth. They complete work and attach proof from the same channel. Their work feeds into the main task and project management system.

That one took about three hours so far. I expect another hour of fixing before it is fully operational.

Where the time actually goes

The pattern is clear after twelve days. The large pieces of work go quickly. A module I expected to take days came together in an hour or two.

The time goes into the work around the build: - error checking - bug fixing - correcting issues - the user experience - the small automations that make the product easy to use - health checks and customer support - the architecture that keeps the software stable

None of that shows up in a seven day app video. All of it decides whether the software works for a paying customer.

Why this matters if you think you missed the boat

Some people see content about simple ideas turned into money-making apps in as little as seven days. I am not saying that is impossible. For something truly simple, it is possible.

Something comprehensive and detailed is different. Vibe coding makes it much quicker than it used to be, but it still takes time, effort, strategy and planning. Anybody can now create an app with AI. Creating one that works and does what it is supposed to do is a different outcome. Plenty of people will produce poor software and try to monetise it, and customers will find out quickly. I see that as an opportunity for anyone who is willing to do the careful work.

The link to the streak

This is day 12 of 30 on the experiment. It is also consecutive day 2,159 of my daily running streak. The two have the same shape. A target is split into daily units. You check the number each day. You do the unit whether it feels exciting or dull.

The modules work like the kilometres. I need three or four a day. I have run 16,577.44km in barefoot-style footwear, which leaves 23,497.56km to complete a lap of the world at 40,075km. Neither target gets done by a burst of effort. Both get done by showing up with the daily unit, including on the days when the work is fixing small errors.

The reason behind the running

I run and vlog every day to raise £1M for children's causes. The more people who see the vlog, the more money we can raise for it. I do not pretend the experiment and the mission are the same thing. The experiment is a business project and it carries real risk. But the discipline I use to run each day is the discipline I use to finish the modules, and I want to be honest about both.

What I am taking from day 12

The learning for me is that the scale of the work is not where the effort sits. The effort sits in the detail, the checking and the stability. I am confident we will reach day 30 with a product that is ready for sale. Whether it makes a sale is still to be determined, and I will report it either way.

If you are building something with AI, tell me where your time is going. I read the comments and reply to anything relevant. If you believe in the mission, subscribing, sharing and commenting helps more people see the vlog. I will be back tomorrow with day 13.

Watch the full episode: https://youtu.be/q2AstT9vvmY