The story of building PLAPP

From sandbox to live Allegro. We did it!

On 26 October, we published our first live Allegro listing directly from PLAPP Shop — an important milestone for the project. 🎉

Just one listing? Behind it were weeks of API integration work, tests, fixes and, almost as a side quest, a “quick migration” to a new VPS. Production, as usual, had a few opinions of its own. 😄

We develop much of PLAPP with the support of AI agents. They help us move significantly faster, but they don't eliminate technical problems or replace business understanding, judgment and real-world verification.

Now comes the real test: will our beta tester at Twój Butik find this workflow useful enough to make it part of her everyday routine? The integration is still evolving, but we've taken our first step into real-world operation.

Lesson learned: AI can dramatically accelerate software development, but it cannot replace product ownership or testing with real users. A prototype proves something is possible; a product must prove it is useful.

Some progress is invisible from the outside

For the last two weeks there was less to show from the outside, while quite a lot was changing underneath. In the sandbox, PLAPP Shop connected more of the marketplace flow: publishing offers, bringing orders back into the cockpit, handling returns and updating inventory.

This is still not a production integration, but the pieces are starting to behave like one process rather than a collection of separate features.

At the same time, the rights to the PLAPP website and platform were formally transferred to PLAPP sp. z o.o. A change of infrastructure provider then started another round of work: a new VPS, migration, security, backups, agent access, the tunnel and DNS.

AI can make parts of this work significantly faster, but it does not remove the need for patience, order and persistence. Some weeks end with a new screen. Others end with less chaos and stronger foundations.

And somewhere between the next checklists, another idea has already started to form. That one can wait. 🙂

Lesson learned: AI can accelerate engineering, but it does not remove the need for patience, persistence and careful foundations. Not every milestone is visible.

Not every working route is a good route

Sometimes you can get from Poznań to Kraków through Gdańsk. You will still reach the destination. The problem is that you will travel quite a few unnecessary kilometres along the way.

A lot of our recent work on PLAPP looks exactly like that: revisiting things that already work, gathering feedback, removing unnecessary steps and finding a simpler route to the same destination.

From the outside, almost nothing may appear to change. Inside, quite a lot does.

Lesson learned: the first working route does not have to be the best one. Sometimes you only see the shortcut after travelling the longer road.

Improve the process before automating it

Automation does not always fix a process. Sometimes it only makes a badly designed process run faster. Before asking “how can we automate this?”, it is worth asking whether the step should exist at all.

Lesson learned: Kaizen is older than today’s AI wave, but the principle still holds: simplify first, accelerate second.

The first real PLN 2

Twój Sklep PLAPP completed its first controlled end-to-end transaction in the live TwojButik store.

Lesson learned: Do not test only features. Test the whole business process.

Automation is easier than trust

The more a system can do independently, the more important it becomes to define where it should stop.

Lesson learned: Mature automation includes knowing when the system should not act alone.

Humans should not be the transport layer

Automation should remove mechanical coordination work while keeping people at the points where judgment creates value.

Lesson learned: Remove coordination work, not responsibility.

The project should remember

As PLAPP projects grew, the memory of earlier decisions, failed attempts and lessons learned became increasingly important.

Lesson learned: Good human–AI collaboration needs a process that can use previous experience.

A dashboard should tell you what to do

We decided that a small-business owner does not need ten more charts. The workspace should show what needs attention and why.

Lesson learned: A good dashboard supports a decision instead of creating another analytical task.

The first store backend goes live

The backend of the first PLAPP Commerce deployment started running in production.

Lesson learned: A demo proves possibility. Production starts answering whether the system can be relied upon.

“Great tool.” And almost nobody used it.

One of our early AI agents received very positive beta feedback, but later barely entered the tester's normal workflow.

Lesson learned: Positive feedback is not the same as product adoption.

PLAPP leaves the workshop

PLAPP started around a simple principle: technology should adapt to the needs of the business, not the other way around.

Lesson learned: Technology is useful when it removes work rather than displaying its own complexity.