Long-form posts and the Know Thyself newsletter — startup finance, building iOS products, and the parts of entrepreneurship that do not make it into the success stories. By Erdem Gülen.
Long-form posts published here, in English. Roughly weekly.
The duplicate payment, the supplier email with new bank details, and the contractor still on payroll all arrive before the finance team does. The five cash controls a ten-person company can run on one page — two people between an invoice and a payment, bank-detail changes confirmed by phone, cards with owners and limits, payroll reconciled by someone who did not run it, and a bank reconciliation someone actually reads — what they cost, and when to add the next layer.
Sort an early-stage budget by size and the top rows are all people; the rest is noise with a line each. Why hire timing, fully loaded cost and ramp are where the forecast error lives, why every planned hire needs a trigger rather than a start date, what the sequence of hires says about the real strategy, and why the round should be sized from the hiring plan — not the other way round.
Every app left in the store is a standing claim on future hours — the OS cycle, dependency churn, support, and attention — and it looks profitable only because the developer prices their own time at zero. The ledger I keep for each app, the three tests that decide which ones stay, and how to retire one without hurting the people who still use it.
The expensive AI mistakes in finance come from putting a probabilistic model where a deterministic one belongs. The four jobs where a language model earns its place in a finance stack, the four where it does not, and the propose–validate–approve pattern that keeps every number reproducible, traceable, and signed by a person.
Most founders write to investors only when they need money — exactly when the email is least useful. What a monthly update that actually gets read contains, why the flat month is the most important one to send, and the failure modes visible from the receiving end.
Founders picture a CFO arriving and building a model in a week. What actually happens: an audit before anything else, a first reporting pack, the model built on data that has already been checked, and an operating rhythm — plus where both sides most often go wrong.
Indie app pricing gets decided in the last hour before submission and never revisited. Why the subscription-versus-one-time choice follows from your cost structure, why trial placement moves revenue more than price does, and the lifetime-value maths almost nobody computes.
Most founders track runway as cash divided by last month's burn — a number that is already wrong the moment something changes. Why runway needs to be a live model with a cash bridge, committed-versus-discretionary spend, and a burn multiple, not a single division problem.
Most startup data rooms are built to look complete, not to survive scrutiny. What an associate opens first, the documents that kill deals when they are missing, and how to build a room that speeds up diligence instead of inviting it.
Bolting a copilot onto a legacy finance stack is not an AI-first finance function. What actually changes when controlling, reporting, and forecasting are designed around AI from day one — and why most "AI adoption" in finance stops at the chatbot.
Early and growth-stage startups keep buying finance software that half-fits and calling it solved. What forward-deployed engineering — small, in-house tools built around the actual business — looks like, and when it beats another SaaS subscription.
A CFO who uses ChatGPT occasionally is not the same as an AI-native finance executive. What the distinction means in practice, why it now shows up in fundraising outcomes and burn multiples, and what to look for when hiring.
Almost every founder hires financial help later than they should — usually mid-raise, when the numbers have already become the bottleneck. Six concrete signals that the finance work has outgrown the founder, what a fractional CFO does that a bookkeeper does not, and when you genuinely do not need one yet.
On-device AI is harder than calling an API, and worth it more often than people assume. What actually changes when you process photos and video locally — the economics, the constraints nobody mentions, and the cases where you should still use a server.
Know Thyself — essays on startup finance, investment and entrepreneurship, published in Turkish on LinkedIn. Subscribe.
Yapay zeka dünyayı değiştirirken Türkiye'nin bu dönüşümdeki pozisyonunu ve kaçırılan fırsatları değerlendiriyorum.
Read on LinkedInBaşarılı girişimcilerin masalsı anlatıları arkasındaki gerçekleri ve başarı faktörlerini inceliyorum.
Read on LinkedInSaaS şirketlerinin takip etmesi gereken temel metrikler ve etkili dashboard oluşturma stratejileri.
Read on LinkedIn20+ yıllık deneyimimle startup finansmanının teorik ve pratik yönlerini karşılaştırıyorum.
Read on LinkedInSAFE anlaşmalarının ABD dışındaki ekosistemlere uyarlanması ve hukuki pratikler.
Read on LinkedInSimple Agreement for Future Equity'in temelleri, avantajları ve girişimciler için pratik kullanımı.
Read on LinkedInGirişim sermayesi dünyasının ironileri: Eski girişimcilerin yatırımcı olması ve üniversite terklerinin başarısı.
Read on LinkedIn