John Carmack the Programmer Who Walks Away From Fortunes
Greg Alexander had a problem only a programmer in the underground around Quake, a first-person shooter, could create: his unapproved port to Linux, an open-source operating system, worked too well.
The game’s source code, from id Software, the studio behind Quake, had leaked after a licensee was hacked. Alexander, unaffiliated with the company, used it to make a richer Linux version than the earlier port. The ordinary corporate move was obvious. Call counsel. Close the breach. Protect the asset.
John Carmack, id Software’s lead engine programmer, did something stranger. id Software absorbed Alexander’s patches into an official Linux port, then put it under a new hire who later brought more Quake games to Linux too.
That one decision explains more than it first appears to. Carmack keeps leaving fortunes behind because ownership, status, and even money seem to matter less to him than preserving contact with the next hard problem. The asset was valuable. The motion was more valuable.
His early career at id Software makes the pattern visible. Commander Keen, a side-scrolling computer game, needed smooth movement on personal computers, so Carmack built adaptive tile refresh. Wolfenstein 3D, a fast first-person shooter, needed speed, so ray casting became the way through. Doom used binary space partitioning. Quake used surface caching. Doom 3, a later horror shooter, got stencil shadow volumes, the technique often called Carmack’s Reverse. A later online shooter used a large-texture system built to make huge outdoor worlds possible.
Read that list wrong and it becomes a trophy case. Read it closely and it becomes a map of constraints. Each engine answered a bottleneck that had made the next experience feel impossible. The reward was not simply shipping another game. The reward was watching a wall become machinery.
That is a different relationship to work than ambition alone can explain. Ambition wants the win. Frontier work wants the next unsolved resistance. Once the resistance is solved, the mind trained on it begins to starve. The product may still be earning. The title may still impress. The fortune may still be sitting there. But the live wire has moved.
Open source showed the same architecture with a moral charge. Carmack released the Wolfenstein 3D source code in 1995 and the Doom source code in 1997, later moving Doom to the GNU General Public License, an open-source license that requires derivative work to remain open. He later regretted using that stricter license instead of a more permissive Berkeley Software Distribution style license.
That regret matters. It was not a retreat from openness. It was a refinement of freedom. Carmack had learned that future builders can be blocked not only by secrecy, but by the wrong legal shape around released code. Doom’s source release shipped without music because of complications around a third-party sound library. He later warned developers that middleware dependencies can prevent source code from being released at all.
His opposition to software patents fits the same pattern. He has compared them to robbery. That sounds extreme until seen through the maker’s eye. A patent on software can turn a discovered path into a tollbooth. Middleware can turn a finished work into a locked room. A bad license can preserve one kind of virtue while shrinking another kind of motion.
So the Quake Linux choice was not generosity as decoration. It was stewardship. A builder had found a working path. Carmack recognized the value in keeping that path alive.
Then the pattern became expensive.
Around 2000, Carmack returned to a childhood interest in rockets and began funding Armadillo Aerospace, a small rocketry company, out of his own pocket. The cost was “something north of a million dollars a year.” The company made steady progress toward suborbital space flight and eventual orbital vehicles. It won $350,000 in the first level of the National Aeronautics and Space Administration’s Lunar Lander Challenge in 2008, then $500,000 in the second level in 2009. By 2013, Armadillo went into hibernation mode.
Here the easy interpretation breaks. Carmack was not merely walking away from wealth after success. He was converting wealth into a harder arena where progress came slower and the feedback was less forgiving. Games had made him rich. Rockets consumed that wealth. The move only makes sense if money is not the score. It is propellant.
The deeper twist is that Carmack said aerospace was “simple” compared with video games. Not easy in the casual sense. Rockets can explode. Physics does not negotiate. But that may be exactly why the comparison appealed to him. Physics pushes back cleanly. Software organizations push back sideways.
Game engines sit inside changing hardware, shifting player expectations, licensing traps, teams, schedules, and product goals. A rocket may kill the illusion faster. A large organization can keep an illusion alive for years by funding it.
Meta, the parent company of Oculus virtual reality products, became the breaking point. Carmack joined Oculus VR, the virtual reality company Meta later bought, as chief technology officer in 2013, stepped down to consulting chief technology officer in 2019 to spend more time on artificial general intelligence, then left in 2022 to focus on Keen Technologies, his new artificial intelligence company, after raising $20 million.
His frustration was not that Meta lacked people or resources. It was worse. He wrote that the company had “a ridiculous amount” of both and still squandered effort. He said he had not been able to kill bad initiatives before they caused damage, or set a direction and have a team stick to it.
That line reveals the actual enemy. Carmack trusts small steps using local information, the practical magic he compares to gradient descent, where tiny corrections compound toward better answers. But gradient descent fails when the terrain is fake, when incentives hide error, when committees preserve bad direction because no one wants the cost of naming it.
Leaving, then, was not restlessness. It was alignment. When the institution could no longer support honest contact with reality, he chose a smaller field where the contract could hold: work intensely, test locally, correct fast, keep moving.
Most people meet the same question in less famous rooms. The meeting has not started yet. The document is open. Everyone knows the plan is bloated, but the salary, title, sunk cost, and social ease all lean toward silence. The hard thing is unfinished somewhere outside the room, or maybe buried inside the work itself, asking for a cleaner answer than the organization wants to hear.
The question is not whether the opportunity is valuable. It is whether staying inside it will slowly make a person less true.
Carmack keeps leaving fortunes behind because, for him, the fortune was never the destination. It was fuel. The destination is the next place where reality still pushes back hard enough to tell the truth.
Sources
Liked “John Carmack the Programmer Who Walks Away From Fortunes”?
Get notified when new Mastery articles are ready.