We lean into the weird, the bold, and the unproven. We want to make experiences that resonate deeply with the players who crave something different. Whether it's a cerebral sim, a quiet narrative, or a systems-heavy sandbox, we aim to build games that stick with you long after the credits roll.
Under the hood, we specialize in data-driven design and algorithmic systems that empower player agency and support extensive modding. Our games are built to be explored, extended, and reshaped by the community.
We're not here to chase trends. We're here to make the kinds of games we want to play, and hopefully, the kind you’ve been waiting for.
Tactical space sims, cerebral puzzle worlds, and simulation-heavy experiences. Explore our catalog, and play select games right in your browser!
View PortfolioDevelopers have a responsibility to respect the time and money our customers put into our products.
We fully support the spirit and letter of the Stop Killing Games EU Citizen's Initiative, and we encourage
everyone to learn more about how publishers unfairly restrict your freedom to use the products you fairly purchased.
Learn More
Games should be for the players.
The gaming industry is buckling under its own weight. Years of squeezing players for every dime has finally started to show the cracks in the big-budget, shareholder-first business model. The giants have used nostalgia, brand recognition, star power, and psychological tricks to manipulate the market into a hellscape of endless monetization models.
Gamers are tired of it. Content in a full-price game (a price that seems to rise every few years) cut off and sold back as "DLC". Battle passes designed to exploit FOMO. Jarring, unimmersive cosmetics meant to manipulate attention-seeking behavior. Entire sections of gameplay locked behind paywalls. Loot boxes that prey on gambling impulses. Live-service games that demand your constant attention, or your progress means nothing.
The "games as a service" model has turned what should be art into a treadmill. Players have been transformed from human beings into revenue streams to be optimized, retention metrics to be maximized, whales to be identified and exploited.
It needs to stop.
We didn't get into game development to nickel-and-dime players or to build Skinner boxes disguised as entertainment. We got into it because we love games - real games. The kind you buy once, own forever, and play because they're actually fun, not because you're afraid of missing out on some limited-time event.
Games should earn your trust the old-fashioned way: by being worth playing, and worth paying for, because they stand on their own merit. No psychological manipulation. No predatory monetization. No cutting content to sell it back to you. Just a complete experience, made with passion, that respects your time and your wallet.
That's the standard we hold ourselves to. That's the promise we make to you.
We recognize that modern AI systems carry real environmental and geopolitical costs. Training and running large, closed models at scale consumes enormous amounts of electricity, water, and critical infrastructure, while concentrating computing power and economic benefit in a handful of corporate data centers. We refuse to participate in that model. Instead, we choose a smaller, more accountable path.
Our hardware reflects that choice. We run inference on older, repurposed GPUs: retired or second-hand NVIDIA RTX and GTX cards, rather than renting time from hyperscale cloud providers whose energy grids are dominated by fossil fuels and whose business model depends on artificial scarcity. We prioritize energy efficiency over raw throughput and accept slower iteration times in exchange for a lower carbon impact.
Our software does the same. We prefer open-weight models over proprietary alternatives because open weights let us inspect what is actually running, audit its behavior, and modify it to better align with our values. We avoid closed or black-box APIs that offer no transparency and lock us into infrastructure we do not control. Where feasible, we use smaller models (7B, 13B parameters) fine-tuned for our specific tasks rather than relying on monolithic general-purpose systems.
We treat user data as ours and theirs, not as a commodity. Data processed by our models stays within our own infrastructure and is never used to train third-party systems. We avoid scraping or ingesting data from sources that did not consent to being part of a training set.
We share our configurations and tooling so others can replicate our approach. We acknowledge that our current setup has limitations: slower inference, smaller context windows, and less capability. We accept those trade-offs knowingly.
Small is accountable: smaller systems are easier to audit, fix, and shut down. Open is verifiable, we can only trust what we can inspect. Older hardware is sufficient, most AI tasks do not require state-of-the-art GPUs. Local is sovereign: keeping compute under our own control resists monopolies. Slower is sustainable: accepting slower iteration is an environmental choice, not a bug.