On building, one month at a time

This is a personal monthly reflection on building in consumer health, looking back on what I've been working on and forward to what's next. I'm sharing this with the people supporting my entrepreneurial journey to create more accountability (downside of currently building solo) and to increase my surface area of luck. More context here.

PS: Much of what I'm doing, including this newsletter, is new to me. Not everything I do or share will be relevant to you, but please never hold back with honest, unsolicited feedback.

February is a short month, no doubt - and yet, realizing that a sixth of the year is already over is 🤯🤯! So let’s not waste time and jump straight into it:

I’m on a mission to empower everyone to build lasting health by uncovering the essentials and making them simple and easy to act on. Everything I do is meant to work towards that mission.

The long-term plan is two-layered: software and “physical”. Currently, the focus is on the first, as the window of opportunity is closing and moats are thinning, traction can be gained faster, and the entry barriers are lower. Broadly speaking, the plan on the software side is to build the healthspan-essentials-app. 

For February, the main objective was to work backwards from the bigger vision to find the best hook/wedge to start with, ie the most interesting, narrow entry point within a specific niche audience, from where I can build layers on top.

To get there, I started with two things:

I also took an opportunistic detour, coming across a free branding competition. While not on the immediate roadmap, working through the brand brief questions turned out to be good practice to reflect on what I want to build from a brand perspective (purpose, values, audience, visual identity, etc). I was able to excite the agency through the first two rounds, though I'll likely hit a dead end on the last one, which focuses on co-promoting the journey to my audience (more on this below in “Looking ahead”).

With the vision set, the imminent question is where to start and how to continue from there? It is common startup knowledge that the smaller and more niche you start, the better. However, this doesn’t answer what to start with, which further raises the question how to figure that out? 

I’d argue that a few years (even months) ago this would have meant validating before even starting to build anything. However, in consumer more than anywhere, it’s also gotten a lot easier to build fast, creating a new way to validate by letting the user use the actual product.

I’ve decided to parallelize those two, validating on the non-building side (such as talking to consumers, looking at market data, unleashing agents on Reddit, etc) and - loosely guided by Paul Graham’s “If you're not sure what to do, make something” - building.

This will likely evolve, but currently I'm thinking of building various small, niche "products" (primarily solving for myself, then bringing to a handful of friends/users) that could compound into the bigger vision (similar to what Peter Steinberger did with OpenClaw). The idea is to create fast loops: build, validate, kill or iterate, connect to or build adjacent products. The more niche the user or use case, the better. In theory, it sounds simple. The biggest challenge for me is staying focused on one project at a time and not flying like a kite pushed by winds from all sides. Paul Graham's concept of backtracking is likely a useful one to keep in mind.

AI obviously plays a central role in all of the above. A few months ago, I would not have been able to approach this the way I am now, both because I lacked the relevant skills and AI tools were not where they are today. The biggest unlock has definitely been Claude Code (as well as the biggest distraction). As someone with no computer science degree who has never properly learned to code, it feels like having superpowers, now able to quickly turn random ideas into action. Things I played around with:

Working on the last example made it increasingly clear to me that while building has become incredibly fast and easy, there is a big difference between building for one or a few users vs potentially for many. The way I currently see how time can be allocated: 

It's become very easy to build lots of small, niche tools for yourself or a handful of users (bottom right), but as soon as you want to build something marketable/monetizable (higher bar, more edge cases, security, polish), the complexity rises sharply (top left). I'm sure this gap will narrow over time; until then, my plan leans toward allocating my time into the bottom right, stitching together small tools and iterating until something is worth scaling.

Overall, I remain very bullish and excited about the space I'm building in. The more I look into it, the more I see how crowded it is (competition is not bad per se), but also the vast opportunity. Similarly, excitement levels oscillate strongly between being extremely pumped about the outlook and letting reality settle in, seeing what a hard undertaking this will be.

Lastly, I've also started acting on two things I hadn't followed through on early enough: not letting stinginess block me from building faster (I should be the bottleneck, not AI usage limits), and surrounding myself more with other builders. To that end, a group of ETH builders recently spun out the J-Floor from SPH, further concentrating the density of people actively building in AI and robotics, offering free workspace. 

Focus. That’s what I need to focus on, literally. As much as AI is giving me the feeling of having superpowers to build 100 things simultaneously, it doesn't. I need to go deeper on fewer things to pursue the loops described above, and figure out how long these iteration cycles take and whether I need to time-cap myself.

For now, I am working on a female cycle tracking app. A what?! While this most certainly sounds unintuitive given the previously outlined vision, it is a validated space with a clear persona and a gap that my wife and I identified together, which I intend to fill. The plan is to plant a specific anchor and build out from there by integrating features and mini tools as referenced previously.

To this end, I’m looking for women to fill out this survey who tick the following boxes (sorted by priority): have a regular cycle (not on hormonal contraception), share or want to share cycle data with their partner (ideally living together), already track their cycle, have no pregnancy/kids, and use wearables.

Looking towards the horizon, there are two things I’m starting to think more about:

It’s no secret that AI is the talk of the town. The post that stood out the most was “Something Big is Happening”, in which the author shares where AI is today, where it’s going, what’s at stake, and what practical steps to take. Reading it prompted me to share the longest WhatsApp message I’ve likely ever written to my family, urging them to take AI seriously, learn to use it, and start experimenting with it sooner rather than later. The goal was not to evoke fear, but rather to spark excitement and awareness. It’s still early (eg “only” 1% of the world has read the post), but things are moving incredibly fast, and it’s easy to fall behind but difficult to catch up later. I hope the discussion will be less around “will my job disappear?” and instead “which tasks in my job will change?”, ideally with us in the driver’s seat, embracing AI and working towards automating our own roles.

The post sparked many reactions. Two follow-up pieces that resonated with me were:

Given that I've barely been able to profit from the AI wave so far myself, I often feel a bit silly preaching it to others. But I also don't care, as long as people I care about actually adopt it. And given that my self-attributed success rate is above zero (my older brother is now completely hooked because Claude Code can do his slides [who still does slides?]), I'll keep preaching it :)

And to counter the most common anti-AI argument out there ("I tried it, and it's just not that good"), here's a personal anecdote from last week's haircut in Berlin: 

As you might imagine, I'm not much of a talker at the hairdresser, but I couldn't resist running my non-representative how-early-are-we-still-with-AI survey. So I asked the hairdresser whether he uses AI and how. He had used Gemini once, ~4 months ago, to see what he'd look like with hair today (yes, bald hairdresser), but was disappointed at the result. I suggested a second try on the spot, took a picture, and prompted "Give this man hair, don’t change face expression”. He was quite pleased with the result. What he didn't know: since his last try, Google shipped a significantly better image model (Nano Banana Pro). What I didn’t know: since that haircut, Google would announce an even better model with Nano Banana 2. Things move fast!

To close with, Lovable is free on Sunday (International Women’s Day), let’s go! 😍

If you've made it this far, you're genuinely a super-supporter (or my very curious mom), thank you! If you want to be even closer to the action, here's where you can help most: