16 months. Several hundred hours alongside client work. One finished product.
I built LaizyNote — a business operating system for freelancers and solo entrepreneurs: notes, tasks, projects, time tracking, contacts, calendar, documents and an AI assistant in a single interface.
I've written plenty about client projects. This is different — this one is mine. And that changes everything: the perspective, the responsibility, and the honest realization of how much a "finished" SaaS actually takes.
No code in this one. Just the parts that were genuinely hard.
Out of my own need. It bugged me that there was no tool you could start with for free, that had everything I needed — and that would later also help me run my business. The big names can do a lot, but they're bloated: you spend weeks setting up before you jot down your first thought.
I wanted the opposite: start instantly, yet have everything on board when you need it. While building, it became clear this isn't a niche — it's a gap. Freelancers juggle five tools (notes here, time tracking there, invoices somewhere else) and none of them shows how their business is actually doing.
That last part — Business Insights — is the one feature I'd single out. Because LaizyNote already knows your time, projects, clients and revenue, it can derive the numbers you'd otherwise keep in a spreadsheet: your real hourly rate (not the one you bill), the trend over years, client health, whether your portfolio leans too heavily on one client. Other tools show you what you did. This shows you what it earned you.
The biggest technical challenge wasn't a single difficult feature. It was keeping an overview: where which data lives, where it flows, who is allowed to see what — and making sure it stays that way while ten other things change.
LaizyNote runs on Firebase (Firestore + server-side Cloud Functions, hosted in the EU). Every user has a personal space but can also share team workspaces with roles and permissions. That's where it gets tricky: every single query has to know whether it concerns personal or shared team data, and whether the user may access it at all.
With other people's data you can't afford a single mistake. So the access rules live server-side — the server decides what someone may see, never the browser. Everything follows the same patterns: consistent data paths, workspace-specific caches, centrally defined rules instead of special cases per feature. It sounds bureaucratic. It's the only way a new feature doesn't quietly break three old ones.
Prototyping is incredibly fast today. A new module stands in days. What eats time is the polish — the last 10% that separates a hobby project from a product people trust with their work.
That 10% is made of things nobody notices as long as they work, and that stand out immediately when they don't:
None of these is big on its own. But there are hundreds of them, and each one wants to be found, understood and cleanly solved.
The lesson that stuck: passing tests don't prove it works in the real app. Some bugs only show up when you operate it yourself in the browser. So a lot gets tested by hand — especially anything involving real money or other people's data.
Daisy is the AI assistant in LaizyNote. You write to her in plain language and she can act: create tasks, structure projects, link contacts, summarize time and numbers. "Create a project Website with three tasks for Emma" becomes exactly those entries.
The crucial part: she asks before every write action. She shows a preview — "5 tasks will be created" — and only a confirm click executes it. AI that writes into your data on its own, without you watching, would be a breach of trust.
Building AI into a real product brings a challenge classic development doesn't have: it's not predictable. The same question can be answered slightly differently twice. The model doesn't always stick to the required format. Answers sometimes run too long mid-sentence. None of this is checkable with "is the code correct" — it has to be tried by hand, over and over. Two constraints mattered most: privacy (Daisy runs through a European AI provider, EU-hosted — fitting for a .eu
product) and cost control (usage quotas and a fixed monthly cap per user, so neither the user nor I get a nasty surprise).
The biggest surprise wasn't the technology. It was how much is needed around the product before you can bill a single customer.
A SaaS isn't just an app. It's a brand with a name, logo and voice. Legal texts, terms, a privacy policy, a service description that holds up contractually. Payment processing with subscriptions, cancellations, yearly and monthly prices, vouchers and invoices. Automated emails, backups, and processes for when someone deletes their account. Users never see most of it — but without that foundation there's no product you can trust.
I thought building the product was the work. In truth the product is one half; the other is brand, law, payment and operations.
Yes — much of it exactly the same. Firebase was the right call for a one-person project with a large feature set: it takes backup, scaling and auth off your plate so you can focus on the product. What I'd change is the planning — I'd cut some data structures differently from the start (things you only understand once you've built them wrong), and I'd budget for all the surrounding work earlier instead of treating it as "I'll do that at the end".
The result is live and publicly available, with all modules in use. Anyone can start for free — which was the whole point from the beginning.
Takeaways if you're building your own SaaS:
Full write-up (and the product) on hafenpixel.de. You can try LaizyNote for free — no credit card. Happy to answer questions in the comments.