The foundry

A small shop
for large specifications.

Brand New Level is run from Weesp, next to Amsterdam, as a network rather than an agency: a small core that pulls in specialists across Europe per problem. It exists to build products, not to sell hours.

Where and what

Pampuslaan 59
1382 JM Weesp
Netherlands

Founded by Kiril Nachev. Tech made in Europe, with a preference for privacy-respecting defaults and licences that do not surprise anyone later.

The shape of the company

What a foundry is,
and what it is not.

A foundry makes things and puts its name on them. It is not a studio that makes things for other people and puts their name on them. Both are honourable; only one of them is this.

We own the products

Popcorn, PageWebPro, Kerf and hAire are ours: the roadmap, the engineering and the consequences. Customers buy a product, not a project.

We build the hard middle

Colour engines, machine-file encoders, render pipelines, protocol surfaces. If a problem can be solved by wiring two SaaS products together, it is not ours to solve.

We work as a network

A small core, with specialists across Europe brought in where the problem needs them: printers, carpenters, colour people, hardware people. The person who knows the domain is usually not on the payroll, and should not have to be.

We are not an agency

Client work lives at Brand Level, deliberately under a different name so nobody has to guess which one they are buying.

We are small on purpose

Four products from a team this size only works because the products share a spine: one design engine, one web platform, one attitude to formats.

Method

How a product
gets started here.

Not a process diagram for a pitch deck. This is the actual order things happen in, and the point at which we are allowed to say a thing works.

01

Find the retyping

Look for the place a human is being used as a converter between two systems. That is where the product goes.

02

Read the specification

Properly, and all of it. Where no specification exists, get a corpus of real files from someone who has them.

03

Build the engine first

The hard part before the interface. If the output is wrong, a beautiful editor in front of it makes things worse, not better.

04

Prove it physically

Put it on a press. Load it on a machine. A format counts as supported when something real comes out correct, not when the tests are green.

05

Then make it pleasant

Hide every one of those specifications behind words the user already uses. The best outcome is that nobody ever learns what BTLx is.

Evidence policy

What we are
allowed to say.

Marketing usually flattens four very different kinds of statement into one confident voice. We keep them apart, on purpose, and label them the same way on every page of this site.

Verified

A test or a machine says so

Something in the codebase asserts it on every commit, or a physical machine produced the right result.

“Every BVX operation the decoder understands re-encodes to exactly the values it was read from.”
Measured

We counted it, once

True at the time of writing and likely to drift. Stated with a hedge word so nobody quotes it as a guarantee.

“About 3 KB of script on a published page.”
Reported

Somebody else states it

Including us, elsewhere. Attributed to its source rather than absorbed into our own voice.

“Deployable from mid-2026. The figure hAire publishes.”
Not claimed

We would like it to be true

Superlatives, firsts, and market-position statements we cannot evidence. They stay off the site until they can be checked.

“World's first CMYK RIP editor.” Not on this site.

If you find a sentence on this site that does not survive that test, tell us and we will change it. The offer is not rhetorical. The address is at the bottom of every page.