Start with the messy.

A drawing set, not a doctrine.

Think of each edition less like an article and more like a drawing set. A plan shows what we currently believe should exist. A section shows how the pieces are meant to work together. A detail shows where the thinking becomes specific. A revision shows what reality forced us to change.

We work on commercial decisions. The dangerous thing about decisions is how quickly a plausible idea can begin to feel like a fact. This series preserves the movement between the two.

What did we believe, and what earned the revision?

The drafts are allowed to disagree.

If WD004 contradicts WD001 because customers taught us something the first version could not see, the archive has done its job.

Nothing here should be treated as doctrine simply because it has been designed beautifully. A hypothesis is still a hypothesis. An observation is still an observation. A repeated pattern earns more weight. A principle has to survive reality more than once.

Do not become a design critic for us.

Use what we build. Notice where you hesitate. Notice what feels unexpectedly clear. Notice what requires explanation that should not require explanation. Notice what you ignore. Notice what makes you want to continue.

If something is wrong, tell us what happened between you and the thing we made. That interaction is part of the material.

The load-bearing idea

At different points we tried to explain the consultancy better: the methodology, the offer, why the thinking was different, why the page should feel premium. Each version could become more articulate while leaving the visitor with the same burden: what am I actually meant to do here?

That changed the design question from "How do we explain Trésor better?" to "What should happen to someone because they visited?"

That became the load-bearing idea for this version. The founder starts with the commercial decision exactly as it exists in their head. The system makes the decision visible, places the company in context, asks only for information that changes the read, then returns what should decide it, what evidence deserves weight, what we would distrust, the smallest useful test, and what would change our mind.

The website is part of the product.

We no longer think of the site as the place where Trésor describes the product. The site is the first piece of the product.

If we say we help founders make better commercial decisions, the first experience should contain some evidence of that ability. Not a claim. Not a testimonial. Not another paragraph telling you how rigorous we are. A small piece of the work.

The website therefore carries two loads at once: it has to help the founder, and it has to help us learn. If either side disappears, the structure becomes weaker.

Why the strange little scenes exist

Commercial uncertainty is not experienced as a spreadsheet. It is experienced as a state. Sometimes it feels like building. Sometimes it feels like noise. Sometimes it feels like sitting alone with one decision that suddenly seems to contain ten others.

The Builder, The Mess, and The Decision are not there to make the website more creative. They are there to test whether recognition can do work that another paragraph cannot.

If those scenes become more memorable than the decision itself, they have failed. If they quietly tell the founder, "you are allowed to bring it here before it is neat," they may have earned their place.

What this version is actually testing

Underneath that are several assumptions. We believe founders will give us a real enough decision to work with. We believe the system can identify the commercial object accurately enough to ask better questions. We believe a useful first read can create more trust than a conventional explanation of the consultancy. We believe the emotional layer can create recognition without becoming decoration. We believe the right people will want to continue after receiving value, not because the page pushed them toward a call.

Those are beliefs. Not results.

What we are watching

We are not particularly interested in asking whether someone likes the website. Beautiful websites are easy to like. That tells us very little.

Analytics can tell us where something happened. A conversation can sometimes tell us why. Neither is the truth on its own.

A useful experiment needs a way to lose.

This version should change if people consistently admire it but do not use it. If founders struggle to express their decision in the first field. If the system repeatedly identifies the wrong commercial problem. If the first read sounds intelligent but does not make the decision meaningfully sharper.

It should also change if people need a human explanation before the product becomes useful, if the emotional scenes create theatre rather than recognition, or if people request consideration because they are intrigued by the design rather than because they have a consequential decision worth working on.

Those are not edge cases we need to defend the design against. They are reasons we released it.

What we are not allowed to conclude yet

We do not know that this is the right website. We know that it is the clearest expression so far of what we currently believe the website should do.

There is no evidence yet that the current visual system improves conversion. There is no evidence yet that the first read is useful across every founder type. There is no evidence yet that this sequence is the shortest path from messy problem to useful clarity.

What we have is something better than another opinion. We now have a version that can meet reality.

The next draft begins when reality has had a turn.

We will look at what people actually did, where the interaction broke, where it surprised us, what founders told us, which assumptions survived, and what changed because of it.

The goal is not to prove that this version was right. The goal is to make the next version less wrong.