Unfinished
October 22, 2025
Tim Stevens
I watched DLC and patches ravage the gaming industry, now OTA is killing cars
By Tim Stevens
Once upon a time, software could be finished. Developers could reach a point where everything was compiled, tested, and handed off for writing to floppies, burning to discs, or flashing onto memory. Whole teams of engineers were drafted to facilitate this process, enforcing rigid quality checks at every step, monitoring not only what the software was supposed to do, but how well it did it.
And then ubiquitous internet connectivity came along and ruined everything. Now, software is never truly done, even long after it’s shipped. The temptation to fiddle and tweak after the deadline has passed is not only acknowledged but is, in many places, standard operating procedure.
I’ve been in this journalism game for a long time now, and I did my time inside the software world as well. I watched video gaming turn from a joyous hobby of release-day excitement to a buggy wasteland so flagrant that being a day-one gamer today often feels like an act of masochism.
I see it happening again. This time it’s the automotive scene that’s in free-fall, and the risks of failure are much, much higher.
Gold Master
I did my first product review in 1997. It was a pretty inauspicious start, a review of a game called Spot Goes to Hollywood. I got the assignment not because I had some unique set of skills or perspectives. It was because I could write a little bit and, more importantly, I owned a Sega Saturn, a console that most American gamers completely ignored.
I reviewed hundreds of games over the years that followed. Most came to me on discs created in low-volumes on desktop CD-R machines. Those discs generally had a distinctive yellow hue, which gave rise to the term “gold master.”
These gold discs typically carried the same software that was going to the replicator for mass production, but sometimes I’d get something in a not-quite-finished state. The disc would come with a piece of paper, or I’d get a note via email explaining what wasn’t quite done in that version. Maybe one of the character’s animated hair clipped weirdly in certain situations, or maybe the sound effects weren’t finished.
I remember testing one racing game and crashing a car into a trackside lake just to see what would happen. Instead of a splashing sound, I got a recording of some dude, probably one of the developers, whispering the word “sploosh.”
Those sorts of occasionally hilarious glitches were rare at first, but as the industry progressed, releasing games with greater and greater levels of connectivity, the quality took a nose-dive. It got to the point where we game reviewers banded together on nascent forums and mailing lists just to help each other through the wholly broken parts of games they’d been assigned to evaluate.
The discs started to show up more frequently. Where they used to have words like “FINAL” or “MASTER” written on them, now it was dates or version numbers scribbled in black Sharpie. Then the downloadable versions appeared, often with weekly or even daily updates. Increasingly, I never got to play the “final” version of the game until it was in the eager hands of gamers, who themselves had to download day-one updates. Video game reviews themselves started requiring updates to match the ever-shifting software being evaluated.
I’m frightened to see that same slippery trend hitting the automotive scene, and the ramifications this time are far more serious.
A Rapidly Decreasing Standard
I should clarify that buggy software isn’t exactly a new trend in the automotive world. Tesla was way ahead of the curve here, shoveling out buggy and incomplete software in its vehicles for well over a decade now, but it increasingly feels like the bulk of the industry is doing its damndest to make up for lost time.
When evaluating a new car at a vehicle launch, I often find myself counting software bugs and glitches. Sometimes they’re minor, like a touchscreen so sluggish that swipes looked more like slideshows, or a navigation system that lost its way whenever I changed drive modes.
Increasingly, though, these problems are getting more severe. I remember testing a certain sports car on track, which threw an error my very first time down the start/finish straight. The dashboard lit up and the car disabled all its stability and safety systems. Every other example of the same car did the same thing in the same spot, which brought the track time to a somewhat abrupt end.
I’ve had SUVs lurch to a halt to avoid phantom cars, sedans scream about engine faults whenever driven up mountain passes, and crossovers refuse to unlock the special blessings of a developer with a laptop. None of this was under extreme testing: All these incidents (and many more) happened on drive routes provided by the manufacturers of the vehicles in question.
I’m not alone. Where once journalists like myself would circle up over dinner to talk about the nuances of ride quality and performance, now we compare notes on glitches and bugs. All the while, we’re serenaded by assurances that fixes and updates are coming, sometimes so quickly that journalists driving on subsequent days have a wholly different experience.
And yet software glitches are already a leading cause of headaches for new vehicles in the United States, and the rate of software-related recalls nearly doubled in the last year alone.
None of this is acceptable.
What’s the Problem?
Issues like these are often labeled as “bad software.” And, since developers are the ones writing that software, they typically take the blame. As a former software developer, designer, and architect (yes, I’ve worn a fair few hats over the years), I feel uniquely positioned to tell you the truth is not so simple. Yes, there certainly are bad developers out there, but any time bad software is released to consumers it’s not a failure of an individual or even a team. It is a failure of the entire organization, all the way up to the very top.
When I was in the software industry, I worked on lots of different types of applications. I deployed internal systems where deploying a bug fix was as easy as running a single command on a terminal. I also worked on commercial software products that shipped on CD. The overall methodologies and procedures for writing software weren’t radically different across these organizations. However, the frameworks for testing and verification were radically more rigid when fixing a problem meant the cost and headache of sending users new discs. It should be no surprise, then, that the quality of those releases was consistently higher.
Those verification frameworks have taken a beating across the industry over the years. Quality assurance and control teams, those noble souls who test the software, used to be just as large and powerful as the software development teams they partnered with. Those teams have been decimated in many organizations, numbers replaced by automated testing suites with as much situational awareness as a self-driving car lacking both radar and LIDAR
In the pursuit of lower cost and shorter timelines, quality has been left behind.
What’s the Fix?
So, what now? Do we ban OTA updates? Do we rip the modems out of our cars and force the companies to get it right the first time? Honestly that sounds pretty nice, but I can’t see the industry shoving the software-defined vehicle concept back in the box. Consumers increasingly expect cars that can evolve and thrive in a connected world. While some smaller brands might do well to pointedly ignore this trend, mainstream machines will need to keep up.
What we need to do is start treating software-related issues far more seriously. I see it again and again, fans of a certain brand saying that recalls that can be fixed via OTA update aren’t “real” recalls and so shouldn’t count.
These recalls should not only count, they should be treated with far greater scrutiny. Plenty of buyers won’t consider a new car for their families if it doesn’t earn an IIHS Top Safety Pick+ or five stars from NHTSA. We need an equivalent rating for software reliability, a framework for testing and tracking both initial and ongoing quality, with an easy-to-understand score. Empower buyers with the knowledge to avoid cars with bad software and, in time, the economics should take care of themselves.
I don’t always believe in the power of capitalism, but if manufacturers start to see the direct impact on their sales, I think we’d see some real progress. We may never again see a time when software is truly finished, but with the right bottom-line pressure, these organizations might at least figure out a way to deliver stuff that is fixed.
3 responses to “Unfinished”
-
Hear hear. The Volvo software update v3.5 which caused total brake failure after 100 seconds of downhill driving, leading to a scary crash for an XC90, is the worst example I’m aware of. That a car and a brand with a top reputation for safety could be abruptly turned into an unstoppable bobsleigh by a software update really shows that no modern car is really ‘safe’.
As well as better testing, I’d like to see architectural changes which mitigate the impact of single failures, since these systems are so complicated it’s hard to prove the absence of flaws even with extensive testing.
-
I just had to buy a new phone because the last OTA update of my previous phone made the dialer not work. My old phone is now a very small tablet. My old phone is also close to end of support, so it’s possible this issue will never be fixed.
That’s the future of used cars. I am not looking forward to it.
-
As someone who works in software to this day, and remembers the days we used to Go Gold, which begat having release managers negotiate the Gold Master* as requiring a Day-One Update, to real-time “We can roll back” releases… my view is that its all money driven, with quality and customer satisfaction taking a fold down, third-row seat to the Shareholders with the massaging seats up front.
It’s disappointing, and I hate seeing 60,000 dollar cars go the way of 60 dollar entertainment products, with zero reflection on the fact that the use cases are completely different. As cars start to go to FSD and Brake-by-wire, I find myself wholly unwilling to trust my life to the Agile Manifesto and “fist-to-five”, feelings-based releases based on financial pressure and not quality/safety.
Recent Posts
All Posts
December 16, 2025
December 16, 2025
December 15, 2025
Leave a Reply