Patrik
Wise word of the week
Software has high fixed costs and near-zero marginal cost — you build it once, then each additional user costs almost nothing. — Joel Spolsky (on software economics, paraphrased from his early essays)
Code is leverage. — Naval Ravikant
We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run. — Roy Amara (Amara’s Law)
Journal
So far I’ve reported about engineering topics. But I’m also managing the accounting of the company, and a company is a business and money matters.
This company is atypical. I work with a few startups, and they are constantly worried about financing: having enough seed money to get started, and additional capital to grow faster. This is a problem we don’t have.
This company started on existing foundations. I wrote the initial version of the software as a pet project in my spare time over the past years, to test some financial scenarios and to learn programming in Go. So we got this pretty much for free. And my expenses were covered by my day job.
I now have a bit of financial buffer which allows me to try launching this product without having to pay myself. I can wait for the company to be profitable. Obviously there is a deadline, but no immediate urgency. We can pay ourselves a salary at the end of the year once we know how much revenue we had.
Similarly with Marc, who has his day job covering his expenses and can ramp up this business in his spare time without needing a salary right away.
Once the company starts turning a profit, we will obviously pay ourselves a salary for our hard work, but for the time being we can grow this business without killing it immediately with our respective salary demands. I know this is a luxury and a privilege not everyone has, but this is our ‘unfair advantage’.
Building a software product (as opposed to dealing with projects) is a very powerful thing. The cost is fixed (it’s implemented once), variable costs are negligible, income is proportional to the number of users. So either it’s a complete loss or a beautiful win.
Since we both work from home, we really don’t need an office nor special material. Our costs are limited to the services we use, here’s the breakdown for May 2026:
- DigitalOcean (cloud provider): 24.86 USD
- Google Workspace (email, calendar, 2 licenses): CHF 15.13
- Senja (feedback manager): 29 USD
- Kit (mail management): 59 USD
- Netlify (web hosting): 57 USD (soon 19 USD, because we removed the option to commit from different GitHub accounts)
- Cursor (AI tool): 20 USD
- GitHub (code hosting): free
- Notion (intranet): free
It’s kind of ironic that running our software is the cheapest item on the list… there is some room for optimization, but it’s already a tight ship.
How did I use AI this week?
I’m still finishing working on the current release, adding a few functionalities. Cursor generated some interesting code, but I think I let too much through. I was overwhelmed by the amount of code it generated and made a few sloppy code reviews (“it seems to work…”).
I paid the price by spending the last 4 days doing refactorings to tighten the code. Cursor generated code in a very defensive way, checking anything that can possibly go wrong. It also failed to understand the architecture and principles I was using (blame me: I failed to tell this explicitly).
So, I got code in the wrong places that needed to be reconciled. And also a bit of code that was actually not needed. Cursor added many checks because it tends to work locally: it makes sense in a structure to check for consistency. But it failed to understand two things:
- structures may have an invariant, it doesn’t need to be constantly checked once it’s established
- some of the libraries already checked those conditions, so no need to check the same before calling
Bottom line: in most of the cases, AI is a clear boost, the work gets done quickly. But when it fails, the cost of fixing it is large. I think it’s an improvement, but accounting for all costs, it’s an improvement, not an order of magnitude better.
Marc
The build
Price is a product feature. That’s what I learned during the past years, reading books from famous entrepreneurs and pricing experts.
Price shouldn’t be an afterthought.
Yet it’s tough to play with, as there is so much emotion tied to it…
Gosh, can I really price our product this much? I know it’s very valuable, because it’s so hard to do on Excel… and I don’t talk about maintaining it with new regulations… for each canton! But still, won’t my blog readers hate me or call me anti-frugal?
Or:
Come on, let’s make it CHF 1/month, so I don’t have to have difficult conversations (mostly with myself in my head…)
But when you step back, and take the pricing question from a rational standpoint, it can be a great ally.
Pricing a product at its true value, you only get clients who feel strongly the problem you’re solving for them. And it cuts both ways:
- Price it too low, and you attract impulse buyers: “hey it’s cheap and I wanna try this thing from MP whom I like a lot”. And they use it once and forget about it forever.
- Price it at its true value, and only the people who genuinely feel the pain sign up… and stick around to actually use it.
And for context on what “true value” means here: a one-off financial plan from a traditional advisor like VZ runs around CHF 2'500, for a single, static snapshot. FI Planner is roughly a tenth of that, for unlimited simulations you run yourself, whenever your life changes.
Now, don’t get me wrong: making FI Planner affordable matters to us, and we’re actively exploring different plans to get there, all while keeping the business sustainable for the time the tool takes to build and maintain. But for now, the price is itself a feature: it lets us develop FI Planner at a manageable pace (read: not 100'000 users overnight) and stay focused on delivering the exact value that the clients who feel the problem are looking for.
The edge
I feel uncomfortable with my pace at the moment.
Indeed, one commitment I made to the MP blog, myself (and informed Patrik, or did I? please confirm Patrik! :-)), is that Mustachian Post is my first priority (i.e. publishing one article a week no matter what).
The other commitment I made to Patrik and myself is that FI Planner is my key project (on top of the “standard” blog), so no new book writing, no new “just this small tiny other side project”.
But if you couple this with (excuseeees!):
- Family
- Work
- Unexpected life events
- Sickness
- Etc.
You end up with:
- I would like us to onboard new FI Planner clients 50 or even 100 at a time (although I know it’s counter-intuitively better for our product, as we can iterate it gradually, vs. a one-shot golive)
- I would like to ship 10x more frontend improvements than I currently do
So basically, the infamous: I lack time.
I joke a lot with Patrik that “when I will be FI next year (if all goes fine…), I will have waay more time!” And Patrik always tells me that this is a dream, as his days still have 24h only…
Anyway… the good news is that I found an arrangement at work that should free up some time, so hopefully I will feel less frustration about it.
Note to a friend
If you want to be better at pricing (for your startup, your corporate, your SaaS, whatever), read one book: “Monetizing Innovation”.
Tool of the week
Claude Fable 5 is really better at thinking in more detail about the problem you throw at it. One example is on my blog where it will not stop at my instructions, but try to think about the blind spots I may have.
The last example: as I was optimizing a page speed, Claude did spot a mistake on a (quite) old blogpost where a link was broken.
Want me to fix it while I’m at it?
Sure thing dear Claude!
And it acts like this on all my issues for a week now.
But, because there is a but… Claude Fable 5 burns your tokens twice as fast as it did with Claude Opus 4.8…
Which ended this morning in me recharging 50 USD of usage credits… all burnt by noon already…
So, after one week of (productive!) fun, I’m gonna go back to Claude Opus 4.8 as my default model, which I didn’t max out since I subscribed to Claude Max by the way.
And I hope that the Google price war (which started earlier this week) will make Anthropic revise its decision to make Fable 5 usable only via “Usage credits” (vs. subscription), and that it will decrease its price too ;-)