Skip to main content
AIDEOLOGY

Sydney, New South Wales

Difficulty that reads the player, not the spreadsheet.

We build the adaptive systems that sit underneath a mobile game: difficulty that responds to how someone is actually doing, and level generation that stays fair when it stops being handmade. Then we ship our own titles on them, because a system nobody has shipped is a slide deck.

Status board

Where this actually stands, in numbers we cannot round up.

Most company sites lead with a metric chosen because it flatters. These are the four that describe the position, including the two that do not.

0

Titles released to a store. Nothing has shipped yet.

2

Systems in the engine, described below, both in active development.

28 May

2026. The date the company's ABN became active. That is the whole track record.

0

Security or quality certifications held. See the note further down.

The systems

Two problems, taken seriously enough to be slow.

Both are ordinary problems in mobile games. Both are usually solved badly, in a way that looks like difficulty tuning and is actually monetisation pressure. The distinction is the whole reason the company exists.

01

Adaptive difficulty

In development

A model that estimates how a player is coping from what they do in the round, and adjusts the next round toward the edge of their ability rather than away from it. The hard part is not the model. The hard part is that the same machinery, pointed the other way, is a system for engineering frustration until someone buys their way out.

Our constraint is that difficulty adaptation never touches anything purchasable, and never makes a round harder immediately before a paid item is offered. That constraint is in the design document, not in a marketing page, and it is the reason this is taking longer than it would otherwise.

02

Procedural level generation

In development

Generating levels that are solvable, that are not trivially solvable, and that do not feel like the same level with the furniture moved. A generator plus a solver: nothing reaches a player unless the solver has completed it and the difficulty estimate falls inside the intended band.

The interesting failures are not the unsolvable levels, which the solver catches. They are the levels a solver completes easily and a person finds miserable, which is a different measurement and the one we are still working on.

03

Own titles on top

Not started

The plan is to publish our own mobile titles built on both systems, rather than licensing them to other studios. There is no title in production, no working prototype, and no announced date.

It is listed here because it is the actual business model and leaving it out would make the two systems above look like products for sale, which they are not.

Position

What "AI" means here, and what it does not.

The word is doing a lot of work in this industry right now. Here is the narrow thing we mean by it.

A lattice of fine hairlines with a small number of brighter paths branching through it and glowing where they meet.
FIG. 01 / SOLVER SEARCH OVER A GENERATED BOARD (CONCEPTUAL). Not a capture of a running system.

What we mean

Small models, trained on gameplay telemetry from our own titles, running on the device or on a small server, predicting one thing at a time: is this player about to give up, is this generated level fair. Classical search and constraint solving does most of the work in the generator. None of it is a large language model and none of it generates art, dialogue or music.

What we do not do

  • We do not train on anyone else's game data. The only telemetry that goes into a model is telemetry from our own titles, from players who have the standard disclosures.
  • We do not use generative models to produce art or audio that ships in a title.
  • We do not make an automated decision that affects a player's rights, their money, or their access to something they paid for.
  • We do not build player profiles for advertising. Advertising personalisation is off unless a player turns it on, and the models described here never see advertising data.

Why say it this way

Because "AI powered" currently means anything from a genuine research programme to a single if statement, and a company that will not say which one it is has usually told you the answer.

Disclosure

The register, and the gaps.

Registration

AIDEOLOGY TECHNOLOGIES PTY LTD, an Australian proprietary company.

ACN 698 536 953. ABN 81 698 536 953, active from 28 May 2026. Registered for GST from the same date. Located in New South Wales.

Both numbers are on public registers. The ABN, its status and the GST registration can be checked at abr.business.gov.au without an account.

What we do not hold

No ISO/IEC 27001 certification, no SOC 2 report, no IRAP assessment, and no other independent security accreditation. We will not claim one until one is genuinely held.

No published titles, no user numbers, no revenue figures, no awards, no client list, and no outside investment.

One inbox

Technical, commercial, privacy and legal all reach [email protected]. Five business days for ordinary mail, thirty days for a privacy request under the Privacy Act.

What to include.