Development
A one-person production, measured as it goes
Parazit is built by a solo developer. The biggest risk in a project like this isn't writing code — it's losing track of what you've done: which asset is actually wired into the game, which fix quietly broke something else, whether a problem solved last month has crept back in.
So the project has a measurement pipeline alongside the production one. None of the figures below are maintained by hand; each is the output of a command and can be reproduced on demand.
Production figures
As of 18 September 2026
76 days
Development time
First commit 5 July 2026. From a grey-box prototype to an Unreal Engine 5.8 production project.
7 levels
Playable maps
Laboratory, lower floor, security line, disposal shaft, smuggler corridor, waste valley, mining site.
1,072
Voice files
529 Turkish, 543 English. Dialogue is recorded in both languages.
11,500+
Game assets
364 skeletal meshes, 2,969 props, 1,709 animation assets, plus materials.
143,209
Lines of gameplay code
C++ across 273 files. Possession, shell collapse, AI, alarm, elevators, interface.
2,536
Production tools
Permanent Python tooling for asset import, placement, auditing and measurement.
Verification
Measured before it counts as done
This project gave its most expensive class of bug a name: produced but never wired up. An asset gets created, saved to the right folder, ticked off on a list — and the game never reads it. No error, no warning, nothing visible anywhere.
- Completion gate — before any task counts as finished, a checklist for that kind of work runs end to end; the automated half is measured by a single command that prints passed or failed.
- Regression tests — the game's rules are tested independently of the game, which catches silent drift between sessions.
- Automated player — a bot walks the map and logs behaviour; results are compared numerically rather than by eye.
- Performance measurement — frame time is measured in packaged builds, never in the editor. Verdicts use the worst 1%, not the average.
- Engineering memory — every solved class of trap is recorded with its case and its evidence; there are 1,451 entries today. The point is to never solve the same problem twice.
What's next
Roadmap
- Demo polish — visual and audio acceptance across the seven levels, playtest rounds, balance work.
- Steam page — store page and trailer, ahead of the Next Fest registration deadline.
- Steam Next Fest — 22 February 2027, public playable demo.
- Full release — targeting June 2027.
These are targets, not commitments. The game won't enter the festival before the demo is ready — a title only gets one shot at Next Fest.
Questions?
Happy to talk about the process, the technical approach, or working together.