Roadmap

Where it’s going, and what’s still wrong.

Direction, not dates. From 0.5.0 on, development concentrates on the molecule sandbox. The other scales stay in the app as they are.

  1. Now

    0.5.0, “the model reacts”

    Released 18 September 2026 · Windows

    • Bonds form while others break, so atoms can swap partners and react.
    • Molecules build up one atom at a time; oxygen and hydrogen make water when heated.
    • Geometry error down from 3.3 % to 2.5 %, with no loss of speed.
    • Saved scenes reopen exactly where you left them.

    Before that: 0.4.0 and 0.4.1 (11 September) added ions you make with + and −, a launcher, and a crowded scene running fourteen times faster. 0.3.0 introduced the molecule workbench.

  2. In development

    Built, measured, not released

    EXPERIMENTAL

    GPU molecular mode

    An optional mode running a machine-learned model on an NVIDIA GPU. In tests it cut the average bond-length error from 2.5 % to 0.71 %.

    Not yet validated for reactions or ions, so it stays switched off for now.

    Double bonds that resist twisting

    In the development build: cis and trans molecules no longer flip into each other at room temperature.

    Updates without the installer

    In-app updates are built and tested; moving existing installs over is next.

  3. Next research

    Chemistry that holds up

    • Reactions that behave, so chlorine, fluorine and oxygen stop reacting too readily.
    • Electrons that move between atoms, for real charge transfer.
    • Checking the hard cases against independent quantum-chemistry calculations.
    • Nitrogen’s fourth bond, so nitric acid can hold together.
  4. Community

    Early testers, and people who know chemistry

    Early-access builds by email, bug reports from different Windows machines, wrong chemistry reported with a reference, and feedback from teachers on what students should see.

    How to get involved

Open problems

The honest part.

These are known, measured, and not fixed yet. If you can help with any of them, write in.

N₄ falls apart too easily

A nitrogen tetrahedron breaks apart far more easily than theory says it should.

Some radical barriers are wrong

Chlorine, fluorine and oxygen react with hydrogen too easily when heated.

No spin, no hydrogen bonds

So oxygen and water don’t yet behave quite like the real thing.

The worst geometries

Carbon monoxide’s bond comes out 7.6 % too long.

Write to Arda

Send a message