Kruno Sulić

Building @Clipriseapp and other online products. 10+ years in digital business. Sharing the decisions, experiments, mistakes and systems behind the work.

[Kruno Sulić @krunosulic]

10+ years building online businesses, products and systems.

Now I’m building @Clipriseapp in public and sharing the real decisions, experiments, mistakes and results behind it.

No recycled AI advice. Just the work.

[Kruno Sulić @krunosulic]

@christophersaum https://t.co/9PVkbc4Bcj

[Kruno Sulić @krunosulic]

I treat consistency as returning to the process after a disappointing result. If the result changes the brief, it was never process, just mood.

[Kruno Sulić @krunosulic]

AI cuts execution time faster than decision risk. The real test is still human judgment: should this pricing page, API, or prompt be allowed to fail?

[Kruno Sulić @krunosulic]

I keep close to founder support because the complaint there is often the first product truth. If it cannot bypass the roadmap queue, you are learning too late.

[Kruno Sulić @krunosulic]

The more AI providers you add, the less any single provider should feel visible. A good abstraction layer turns failure into routing, not a user problem.

[Kruno Sulić @krunosulic]

@vitaliidodonov MRR

[Kruno Sulić @krunosulic]

Discipline is not always more effort. In a long work block, fatigue starts to blur judgment, and recovery begins when I stop before the next decision turns noisy.

[Kruno Sulić @krunosulic]

Speed is a feature of a reversible decision, not a universal virtue. A pricing page can be rolled back; a core API or slug choice can’t. Which decision could be tested cheaply instead of debated?

[Kruno Sulić @krunosulic]

@Salifsibane16 Welcome to try https://t.co/WJfJn5by7c - if you need anything ping me directly

[Kruno Sulić @krunosulic]

A model catalog is inventory. The product starts when selection, cost, failure, and output are controlled enough to trust the result.

[Kruno Sulić @krunosulic]

I’ve learned that multiple products add maintenance before they add revenue. The hidden cost is not launch, it’s every API, invoice, and status endpoint you now have to keep alive.

[Kruno Sulić @krunosulic]

Irreversible decisions deserve more evidence than reversible ones. A pricing page, API, or slug can be changed later; a core decision with high reversal cost can’t. What would be costly to undo six months from now?

[Kruno Sulić @krunosulic]

A prompt enhancer should stay off when the brief needs control, not interpretation. If it rewrites the intent, the user stops steering and starts guessing.

[Kruno Sulić @krunosulic]

The gap between an idea and an operator shows up after the first version breaks. The real test is whether the next fix changes the failure state, not the mood.

[Kruno Sulić @krunosulic]

A risk register is only useful if it tracks unknowns that can still change the plan. If the symptom is a fragile forecast, which layer do you inspect first: demand, pricing, or one unknown that could invalidate your current assumption?

[Kruno Sulić @krunosulic]

A longer prompt can improve control, until one instruction asks for precision and another asks for freedom. Then the model has a contradiction, not a brief. Which instruction would you delete first from an overloaded prompt?

[Kruno Sulić @krunosulic]

@tomiarakaki Appreciate the reply mate. Any pointers as I am struggling with that part. 0 tiktok views on 100 vids - thinking to buy new prewarmed acc and/or to go for creators

[Kruno Sulić @krunosulic]

@tomiarakaki Wow! Aso or seo?

[Kruno Sulić @krunosulic]

Building from Zagreb is fine when the API, partners, and customers are global. The local part is still the handoff: one 403, one invoice, one status endpoint that blocks everyone.

[Kruno Sulić @krunosulic]

Build versus buy is usually about which dependency you want to own. If support tickets or a 403 keep landing on your team, which layer would you inspect first, and which failure would you rather control yourself?