Color

Indigo
Red
Green
Teal
Blue
Purple
Rose

Mode

Light
বাং

What we research

Six areas, each one with something released behind it.

Languages and compilers

Programming in Bengali rather than English, and what it takes to make a general purpose language usable by people who never learned to code in a second language.

Shunno Programming Language

Machine learning and deep learning

Models trained for real, narrow problems — conversation, classification, and prediction — rather than benchmarks, and small enough to run where our clients already are.

Britto AI, Fishtock

Web frameworks and engines

How much of a full stack framework is actually needed for a small team to ship — worked out by building lightweight MVC, template, and database layers and using them on real projects.

S PHP Engine, S Template Engine, S Database Explorer

Developer tooling

The time between an idea and a running project. Scaffolding, task automation, and project-aware command line tools that remove the setup nobody wants to repeat.

Shemul CLI

Desktop and platform software

What a browser or a document viewer needs to be when it is stripped back to the parts people use, and how far a cross-platform build can be pushed on modest hardware.

Jaal Browser

Privacy and the open web

Consent, data handling, and the rules a site has to follow — implemented as something a small site owner can install rather than a policy document they have to interpret.

CookieCons, ProjectPress

Process

How research becomes a product

Nothing leaves the centre until it has survived all five steps.

  1. 1

    Question

    We write down the problem in one sentence, along with who has it and what they do today instead. If that sentence will not hold up, the work stops here.

  2. 2

    Survey

    We read what already exists. Most questions are answered somewhere, and finding that out early is the cheapest result a study can produce.

  3. 3

    Prototype

    We build the smallest thing that can be wrong in public — a library, a model, or a command line tool — rather than a finished product nobody has tested.

  4. 4

    Evaluate

    The prototype is used on real work, by us first. What it costs to run, to learn, and to maintain is measured before anything is promised to a client.

  5. 5

    Release

    What survives is documented, published with its source, and then kept — updated and supported for as long as people depend on it.

Bring us a question

If there is something you need answered before you can build it, tell us what it is. We will say plainly whether it is a study, a product, or something already solved.