Patrik
Wise word of the week
The best documentation is the one nobody has to read. — Unknown
Journal
FI Planner version 0.9.5 is out! It’s a big step towards having all features we want and you expect from the tool. This version was dedicated to adding missing events and tuning existing ones. Here user feedback and dogfooding (i.e. using the application ourselves) was critical in removing friction where it happens and closing gaps.
As an example, we kept getting questions about the difference between ‘Quitting Job’ and ‘Retiring from Job’. This is a subtle Swiss process: you can only retire (claim a pension or capital from your pillar 2) if you are employed. The alternative is to quit a job and move your pillar 2 to one or two vested accounts. Which means giving up the option of getting a pillar 2 pension. This is the only way for people who lose their job before reaching age 60, are self-employed, and people who reached FIRE at 40. But not everyone is aware of this, and it can be scary to have to manage all that money instead of getting a pension.
I considered improving the documentation for those two events. But the better solution was to make it explicit (almost self-documenting) in the application by presenting all options in a single place. I introduced a new event to model ‘Leaving Job’ and made it a required input to decide what happens to the pillar 2: either it is moved to a vested account or into retirement. It’s an event only in the application, behind the scenes the system still has the two low-level ‘quit job’ and ‘retire from job’ events.
In the case of retirement, there is still a question: the split between pension and capital. I added a slider to select the split between the two. This was one of those “wow” moments in the app development. We are so often faced with the “pension or capital” question, and the answer is often a wall of text with many considerations (don’t get me wrong, you really need to understand the consequences). But here the question can be answered by simply moving the slider and quickly seeing the impact it has on one’s financial evolution. It felt almost too simple to be true.
I think we are getting close to reaching the feature set target we defined for v1.0, the public unrestricted launch, which is a mix of must-have features and automation of processes that make the whole application function with little intervention from our side. Right now there is still some work to do for each customer. Our short-term roadmap is:
v0.9.6
- Automation of processes, in particular user account management. Connect with Kit for e-mail management, improve Stripe integration for account management, in particular automatic synchronization between FI Planner and Stripe.
- Proper logging of user actions for analyzing how the app is used. I know, it’s embarrassing that this is not done yet, I would say this is probably a victim of prioritization as we started getting users and feedback and wanted to react timely.
v0.9.7
- Add more cantons: SZ, LU, TG (handling user requests, this will increase our coverage from 74.5% to 84.6%)
- Update tax (some data for VD and VS has changed)
We will continuously evaluate the user feedback and the requests for missing features. Once those drop, we will probably have reached the sweet spot for v1.0 and opening up the app. I think we are getting close.
How did I use AI this week?
Most of my requests are tiny. It’s often for getting a second opinion or doing what AI does well (wordsmithing):
‘check the release notes for grammatical or spelling errors and if needed rephrase each bullet for clarity using a simple friendly non-technical language but avoid sounding like tone-deaf corporate speech.’
Another area is when I know what I want, but I don’t know the tool well enough. I love accumulating encyclopedic knowledge (I could probably list ‘reading wikipedia’ as a hobby of mine) but I don’t want to read through 40 pages of documentation for something I’m likely not going to need again. A recent example was to clean up our Git setup that had accumulated lots of historical debt. We had a backend and a frontend repository. I then created a deployment repository for pushing our releases to Digital Ocean, using two git submodules to include the backend and frontend.
Before long, I was only developing in the deployment repository, because most of the changes touched both the frontend and the backend. Submodules proved to be quite a pain in this configuration, having to push each change to 3 different repositories. Cursor came to the rescue, creating a plan to merge all the repositories into a single one. It involved some convoluted and rare commands to rewrite history, port the tags, and ensure everything was consistent. This took probably 1-2 hours. In the pre-AI times it would have cost me a few days, probably going down a few dead-ends while doing it.
And last but not least, when I look at my Gemini chat log, I realize how it’s replacing search for many requests that require a bit more research than just looking up a reference page. Here’s an extract from my recent chats:
- Fan Strategy for Room Cooling
- Cheap VoIP Proviers for France
- Peppermint: brown leafs, cause and solutions
- Breakfast and jogging: Timing and ideal meal
- Fix LinkedIn Link Preview Image
- UDM-SE for 10Gb/s Fiber 7
- Salat with bresaola and mozzarella
- Insolit Venice: museums and secrets
It has become so deeply engrained in my way of researching things. One thing not visible in the list is that I often had requests that mixed languages, where I started writing in English and pulled some technical German word, or added Italian or French words. AI had no problem with that and would just accept it as input easily.
Marc
The build
Last week was better because I found my Deep Work mode, thanks to Walter (more on that below).
It may seem very little to you, but I managed to:
- Onboard 20 new waitlist members
- Run our regular Discovery survey (results still to be analyzed and explored)
- Add better and clearer documentation for the input of your 1st pillar data
- Add og:image everywhere on our website for better rendering and SEO
- Review the huge list of our v0.9.5 (thanks Patrik for all the crazy work done during the past weeks!)
- Restart my work on pricing (see “The edge” below), and discuss its technical implications with Patrik
It’s not much, but I’d have done only 1 and 2, had I not taken back my life (well, at least my todolist) in control.
The edge
We heard all your feedback, and it’s clear in my mind now: we’ll offer a one-shot pricing solution (read: non-subscription style).
The details are being finalized while I write this, which is why it’s the uncomfortable thing right now.
The goal of this new pricing alternative is dual:
- For the users: have a way to try the product without committing to a recurring subscription
- For FI Planner: have more users using our products, to get even more feedback that we can integrate
I’m still convinced it’s too early to offer a “Freemium” package, as it will bring too much noise compared to what we can handle. And also, I see this pricing solution as ideal for the “Growth” stage, not for when you’re fine-tuning your Product-Market Fit.
Another thing that makes this topic fit into “The edge” is that I have the fear of having too many pricing points with 3 options, or of making a mistake, and then we need to create another pricing model AND handle the former onboarded users on a past model… anyway, fears, but let’s cross that bridge once we get there ;)
Note to a friend
Having a strong personal productivity system is gold!
I still feel very energized while it’s already 15h here, all because I didn’t get disturbed by “just one quick check of my mailbox”, and just focused on my task at hand. Pure Deep Work day, I love it (check Cal Newport’s book of the same name if you don’t know the concept).
Keep experimenting
I know it’s the heart of a startup, and once you stop experimenting, you start falling behind. But I can tell you that theory is cool (and true), but practice is hard! At the moment, I feel like I want things to settle and that we call it our final product/pricing/website/whatever. And it’s exactly the opposite that we need to do: more exploration, more experimentation, full steam ahead! That’s counterintuitive. And hard. And I love such challenges thrown at me by my own very human nature :-D
Tool of the week
Claude. LLMs. AI. Walter (my homemade Chief of Staff that I detailed in last week’s Journal entry).
Gosh, I mean, I feel I’m finally unlocking the full potential of AI for my own personal productivity. And yes, I know, I’m maybe still in the honeymoon phase, and in 3 months, I will tell you that this Walter thingy was cool, but XYZ is the new thing, even better! We’ll see.
All I know is that I’ve finally found back some mental peace, and some mental space to enjoy true Deep Work sessions. This feels good.
Ah, and also, AI helped me optimize how I cool down our home with fans (those were fun discussions with Patrik on Slack this week, up until we tried the range hood trick -> it didn’t help much on my end though…).