Kenshi should not exist. A solo security guard picking a bare-bones open-source renderer in 2006, then spending 12 years building an 870-square-kilometer squad RPG with no quests, no heroes, and no level scaling — that is not a pitch, that is a warning label. Yet Lo-Fi Games did exactly that, and Kenshi sold over 1 million copies by September 2020 and passed 3 million copies by May 2026.
The short answer is this: Kenshi bet on OGRE (Object-Oriented Graphics Rendering Engine) plus C++ because OGRE is not a full game engine. It is a scene-oriented, real-time 3D rendering library. That poverty was the point. In 2006–2008, when designer Chris Hunt started work largely alone, OGRE gave him free MIT-licensed C++ rendering with a modular plugin architecture and nothing else — no imposed character controller, no quest system, no store, no forced upgrades. Hunt then spent a decade building everything Kenshi actually is on top: squad RTS control, limb-level medical simulation, world-state faction simulation, and hand-placed open-world streaming.
That choice explains both why Kenshi feels miraculous and why it feels janky. The same decision produced the depth and the awkwardness.
The facts: who built what, on what, for how long
Kenshi was developed and published by Lo-Fi Games, designed by Chris Hunt, for Windows, and released on December 6, 2018 after entering Steam Early Access in March 2013 as one of the platform's first Early Access titles. Its infobox engine credit is OGRE.
Hunt began around 2006–2008 in Ogre Engine, worked part-time as a security guard for the first five years, went full-time after early sales traction, and worked alone until about 2013 before hiring a team that never exceeded six people during the original game's development. He describes the setting as sword-punk, inspired by wandering ronin and wasteland survival, and his design mantra bluntly: he considers himself the player's enemy.
The Steam page sells the fantasy without heroics: a free-roaming squad-based RPG with no linear story, where you can be a trader, thief, rebel, warlord, farmer, slave, or just food for the cannibals. Lo-Fi's own about page repeats the core rule: you are not the chosen one, you have no extra hitpoints, and the world does not level up with you. Skills start near zero and improve by doing — steal to raise thievery, fight to raise combat — and you can recruit named and generic characters, grow to multiple squads, research technology, and build an outpost or city.
Combat damage is per body part, with crippling penalties, severed limbs, bleeding, starvation pass-outs, and robotic prosthetic replacements. The world reacts through factions and power vacuums: kill a notable leader and towns can change hands or new locations can spawn. The playable area is 870 square kilometres, deliberately mixing dangerous gaps between rewards.
Critical reception was generally favorable. PC Gamer's review praised frightening depth behind presentational ugliness: scavenging battlefields early, hiring mercenaries to raid a slave camp, then hiding in a swamp while a rescued companion's bounty expires because his shaved head marks him as an escapee. The same review warned about grind, cumbersome squad UI, fiddly inventory across 10+ characters, and automated RTS combat with only light tweaks to posture and formation.
Sales vindicated the grind. Lo-Fi's community update, reported by PC Gamer in October 2020, confirmed 1 million copies sold, with proceeds funding Kenshi 2 and an expanded team of 15 in Bristol.
What OGRE actually is — and is not
OGRE is a C++ class library for 3D rendering, first registered on SourceForge in February 2000 by Steve Streeting, with 1.0 released in February 2005. It is cross-platform, open-source, and since version 1.7 in 2010 released under the permissive MIT License (previously LGPL). The official OGRE site still presents it as an open-source 3D graphics engine in the narrow sense: rendering, not gameplay.
That distinction matters. OGRE handles scenes, materials, skeletal animation, and render-system plugins. It does not ship Kenshi's squad AI, pathfinding, economy, dialogue, save system, or UI. Those had to be written. Even physics came from elsewhere: both Lo-Fi's site and Steam credit NVIDIA PhysX under license. In other words, Kenshi's stack was OGRE for pixels plus PhysX for rigid bodies plus a decade of custom C++ for everything players remember.
OGRE's history also explains its appeal to a soloist. It powered titles including Torchlight, Hob, Rebel Galaxy, Running with Rifles, and Roblox from 2009–2014, plus robotics and simulation tools. It was stable, documented, forum-supported, and free of royalties — invaluable for a small team on a tight budget, as contemporary testimonials on OGRE's site stress. It was never fashionable the way Unity and Unreal became. That unfashionability was protective.
Why OGRE made sense in 2006
In 2006, there was no free Unity with an asset store waiting to rescue you. Unity 1–2 existed but the indie-friendly free era had not arrived; Unreal required serious licensing money. For a C++-comfortable solo dev who wanted a huge 3D open world, the realistic options were: license a commercial engine you cannot afford, write your own software renderer, or take a free rendering library and own the rest.
OGRE offered three practical wins:
Zero cost, zero permission. MIT (and LGPL before it) meant Hunt could freeze a version, hack internals, and ship commercially without negotiating. When your plan is 12 years, you cannot afford a middleware owner changing pricing or deprecating your render pipeline in year four.
C++ control over memory and streaming. An 870 km² seamless world with squads scattered across the map lives or dies on streaming, LOD discipline, and custom background simulation. C++ plus source access lets you build the exact nine-zone loading scheme Hunt later described — only simulating zones around the camera, loading three ahead and unloading three behind — rather than fighting a black-box world partitioner.
No gameplay opinions to fight. A full engine would have pushed a first-person controller, a quest graph, or level-scaled encounters. OGRE pushes nothing. If you want shops that stock what locals would stock instead of what matches player level, or a tower full of cannibals instead of loot as a deliberate trap, you just code it. Hunt's team hand-placed towns, loot, weather, vegetation, mining spots, NPCs, and shop inventories across roughly 60 biomes by staring at the map for hours and solving it like an interlocking puzzle.
What the stack gained — and what it cost
Gained: a simulation-first RPG that a commercial engine might have talked Hunt out of. Because there was no quest system, Kenshi has no quests — only self-set goals as trader, farmer, raider, thief, or settlement leader. Because there was no level-scaler, shops show fancy gear you cannot afford, which creates the exact longing Hunt wanted. Because AI had to navigate a world that often is not in memory, Hunt built abstract AI motorways — global road graphs that let off-screen characters pathfind without terrain loaded. Because the world is hand-placed but simulated, a hashish run from the lawless Swamp to the Great Desert can pay a 2,000% markup if you survive hunger, infertile land, and guard shakedowns. Those are not OGRE features. They are the features OGRE's absence made room for.
Cost: everything players joke about. PC Gamer's arcane MMO UI complaint, fiddly base automation, camera-plus-click movement, and grindy early hours are classic symptoms of a custom sim with a thin presentation layer. Pathfinding through hand-edited mountain ranges required artist Oli Hatton to break up impassable ridges so AI could pass. Loading only nearby zones creates the exact pop-in and AI-teleport compromises you would expect. Hiring was harder too: OGRE C++ experience is rarer than Unity C# experience, and every system is bespoke, so onboarding means learning Kenshi, not learning an engine.
The sequel arc proves the trade. In March 2019 Lo-Fi said Kenshi 2 would keep the same engine, upgraded for performance, explicitly to prioritize gameplay over graphics and avoid another million years. Later summaries list Kenshi 2 as an Unreal Engine prequel set 1,000 years earlier. The shift from upgrade OGRE to Unreal is not hypocrisy; it is the bill coming due. Once you have 3 million customers, a 15-person team, and a prequel to ship, the math that favored a freezable renderer for one person flips toward hired velocity and modern tooling for a team.
My take: three contrarian lessons other postmortems miss
Most retrospectives frame OGRE as an outdated burden Kenshi succeeded despite. I think that gets causality backwards. Here is where I diverge, offered as interpretation, not verified fact:
1. Renderer-only poverty was a design firewall.
Full engines smuggle values: heroes, quest arrows, scaled loot, fair fights. Hunt's stated dislike was hand-holding at RPG starts. If he had started in a modern engine, he would have spent years disabling its helpfulness. OGRE had no helpfulness to disable. The lack of a quest editor is why Kenshi's best story — an escaped slave with a shaved head hunted as contraband — can emerge without a quest flag. Constraint did not limit vision; it protected vision from defaults.
2. Boring and freezable beats powerful and moving.
Twelve-year projects die from churn, not from missing features. Unity and Unreal of 2006–2018 changed renderers, scripting runtimes, pricing, and UI frameworks repeatedly. A solo dev cannot re-platform every two years. OGRE 1.x was boring in the best way: stable C++, source-available, community-answered, royalty-free. Hunt could pick a snapshot and keep building simulation while the industry rewrote itself around him. That is antifragility through unfashionability — my term for why the dead choice survived and the live choices would have killed him with maintenance.
3. Jank is legibility.
Kenshi's awkwardness is information. When combat is automated, limbs sever visibly, characters limp or crawl, blood attracts predators, and shopkeepers in the Shek kingdom only sell large weapons because that is what Shek would stock, the system is readable. You can reason about it, exploit it, and tell stories about it. Polished engines hide simulation behind animation blending and smart assists; OGRE left Hunt's simulation naked. Players forgive ugly when ugly is honest and exploitable. I would rather teach from readable jank than from pretty opacity.
None of this means you should start your 2026 open-world RPG on OGRE 1.x. It means you should copy the property, not the product: pick the thinnest stable layer you can freeze, own your differentiating simulation, and let unfashionable tech shield a long project from platform weather.
What indie devs can steal this week
- Write your anti-hand-holding rule first. Hunt's I am the player's enemy is testable: no scaled shops, no scaled enemies, no quest arrow. If you cannot state your hostility in one sentence, a full engine will state its friendliness for you.
- Hand-place unevenness on purpose. Do not evenly distribute loot. Put treasure far away, danger in between, and emptiness as a feature. One desolate zone with only wind and rumor is worth five balanced zones.
- Build AI travel before pretty travel. If squads cannot cross your map off-screen via abstract roads, your map is decoration. Prototype the motorway graph before the vista.
- Make shops cultural, not leveled. Decide what each faction would stock, then let the player suffer the trip. Longing is a quest system that needs no UI.
- Freeze your renderer. Pick a version, vendor it, and ban upgrades unless they fix a ship-blocker. Your 12-year self will thank you.
Train the instincts in a browser before committing to 100 hours
Kenshi demands 30 hours before it clicks — PC Gamer still felt unfinished after that. If you want to rehearse the mental muscles without installing 14 GB, these free browser approximations each isolate one Kenshi instinct. They are simplifications, not replacements, but the transfer is real.
For open-ended roaming without a mission arrow, spend an hour in Sandbox City – Cars, Zombies, Ragdolls!, an open-world playground where you can drive freely, roam the city, and decide whether to hunt zombies or test stunts. Its no strict mission path freedom is the five-minute version of Kenshi's lay siege as a warlord or get rich as a trader ethos: practice setting your own goal and living with the consequences.
For the settlement half of Kenshi — research tech, raise walls, automate mining and farming — warm up in Mine Clone, where you start with only surroundings, break and reuse every block, and choose Creative Mode for layout sketching or Survival Mode for gather-craft-defend pressure. Starting small with a hut before a fortress is exactly how Kenshi veterans advise placing your first outpost outside taxation range.
For the emotional core — starting at zero, getting beaten early, and slowly becoming someone — try Horse Simulator 3D, where you roam an island, avoid predators early, raise strength and endurance first, and only expand your herd after you can defend it. Its strengthen before expanding tip mirrors Kenshi's brutal opening: do not recruit a second mouth until the first can survive a beating.
Play all three and you will know within an evening whether you enjoy self-directed survival. If you do, Kenshi's 870 km² will feel like home. If you do not, you saved yourself a grindy week.
Bottom line
Kenshi bet on OGRE because one person building an uncaring world needed rendering without governance. OGRE gave pixels and plugins; Hunt supplied cruelty, consistency, and time. The cost was jank and a sequel that eventually outgrew the stack. The reward is the rare RPG where shops do not love you, the map does not scale to you, and a tower of treasure can be a hundred cannibals — and that is entirely the point.
Top comments (0)