Florian Myter

The through-line

The recurring preoccupation is defensibility and where it actually lives once technology stops being scarce. The argument is that "Building a moat from a tech perspective has become really difficult, while go to market is still something very human — you're still selling to people" [3], and everything else follows from it: speed over polish, validation before building, and a growth machine assembled from loops, conversion and activation rather than from a clever product alone. The second, quieter through-line is honesty about what is not working. Growth is described as linear, churn as "wreaking absolute havoc on growth" [1], and the viral loop as unsolved even after a win [2]. The numbers are published as they are, including the unflattering ones.

On why the moat is go-to-market now

AI code generation has collapsed the technical barrier, so defensibility has to be built somewhere a competitor cannot copy from a repository: the go-to-market machine, which stays human and stays hard [3]. This is not a reason to slow down on product but a reason to be clear-eyed about what the product buys you. The practical consequence shows up in what actually gets measured and reported: conversion of website visitors, activation, churn, acquisition mix [1][2]. The pleasure taken in the mechanism working unattended is real, "It's always nice, like, had somebody in South Korea while I was sleeping started using the tool, then converted to a paying customer" [3], but the mechanism is the thing being built, not the code underneath it.

On churn and linear growth

Churn is treated as the binding constraint rather than an annoyance. Monthly submissions grew linearly, and the diagnosis is that acquisition improvements are being spent offsetting losses at the other end: churn is "still wreaking absolute havoc on growth", with the offsetting work coming from "getting increasingly better at converting website visitors and activating them" [1]. The response is to ask publicly for help from people who have solved it, which is consistent with the broader habit of treating an unsolved problem as something to name rather than to narrate around [1].

On viral loops and reporting your own numbers

The viral loop work is a case study in incremental, measured change. The badge on all Weavely forms generated in the free tier was redesigned and a CTA added at the end of free forms; the loop's share of new sign-ups moved from roughly 5% to roughly 11% [2]. The verdict on this is deliberately split: the viral factor is still 0.04, "so we definitely haven't solved this yet. But hey, I'll take a 2.3x lift any day!" [2]. A relative improvement is worth banking without letting it stand in for the absolute result, and the update itself was prompted by a single person asking a follow-up question, which is a fair description of how small the audience for honest operating detail usually is [2].

On shipping fast and 'good enough'

The instruction is blunt: "For the love of God, just break stuff as fast as you can" [3]. The target is the test-and-retest culture of academia, and the mindset shift academic founders need most is around 'good enough' and failing fast, shipping things that are only good enough to learn from [3]. This sits directly against the instinct to make the technology correct before exposing it to anyone.

On spin-offs and building the wrong way round

University spin-offs are structurally disadvantaged: "As a spin-off, I think by default the only way you can build a startup is the wrong way, namely you start out with technology and then you try to reverse engineer a problem onto that" [3]. The root cause is distance, academics sit too far from real-world use cases with paying customers, so the problem gets fitted to the technology after the fact [3]. The corrective is cheap and comes before any building: validate the hypothesis by talking to people, using paid interview platforms at €50 to 100 per interview, where five people in the right demographic already yield major insights [3].

On Belgium, global ambition and regulation

The Belgian market is small and segmented by language, and over-focusing on it makes internationalising later genuinely hard, which is why startups there need global focus from day one [3]. That constraint is framed as an advantage rather than a handicap, because it forces global ambition early instead of allowing a comfortable domestic plateau [3]. On the standard complaint about EU over-regulation, the position is that founders are worrying about the wrong thing: problem-solution fit is the top problem, regulation only bites at scale-up stage, and if it ever genuinely blocks you, move to the US [3].

Takeaways

• Assume the technical moat is gone. With AI code generation, defensibility sits in the go-to-market machine, which is human and hard to copy [3].
• Acquisition gains do not compound if retention leaks. Linear growth can mean conversion and activation improvements are being consumed entirely by churn [1].
• Report the relative win and the absolute miss together: a 2.3x lift in loop-driven sign-ups, from ~5% to ~11%, alongside a viral factor still stuck at 0.04 [2].
• Small surface changes carry loops. Redesigning the free-tier badge and adding a CTA at the end of free forms doubled the loop's share of new sign-ups [2].
• Spend €50 to 100 per interview and talk to five people in your demographic before building anything; that alone produces major insight [3].
• Ship at 'good enough' and fail fast, especially if you come from academia: "For the love of God, just break stuff as fast as you can" [3].
• Build for global markets from day one if you are Belgian; the domestic market is small and language-segmented, and late internationalisation is painful [3].
• Do not let EU regulation dominate your worry list. Problem-solution fit is the real problem; regulation matters at scale-up stage, and relocating to the US is the fallback [3].