Personal Statement on Generative AI

by Kartik Kini (co-founder)

Preface

I want to make clear that we are not making a moral judgement of anyone else’s use of generative ai. Our policy exists to ensure that we are aligned internally and that externally our values are clear. This rant is my best attempt at organizing messy thoughts on the technology that is going to cover a lot about moral and business frameworks for why our policy is the way that it is, but this does not mean that I think ill of anyone who chooses to use generative AI. I am able to separate my opinion on the technology and the people who use it. My opinion of the companies that develop it, however, should be clear based on the moral arguments I present below. My views do not reflect the views of all members of Finite Reflection Studios, but we have aligned on the policy above. Each member may have their own reasons for doing so.

Also if there are spelling errors or grammar mistakes that would be because I typed this up in a text file without any kind of spell check, and relied on a wide range of friends and colleagues to peer review this.

Generative AI for Creativity

Let's start easy. Using GenAI for art / music / story is a nonstarter. These elements are deeply tied to human expression and we should care that art takes time and effort to make. This is not something that should be optimized in this manner, and generative AI is not equatable to a technological advancement like Photoshop as some have argued. That debate is beyond the scope of what I can argue here. This is so fundamentally obvious to me, and plenty of folks much smarter than me have explained this point far better than I can.

The only way these tools are able to create assets is by stealing them. It is impossible for these tools to generate any of the above without widespread massive theft. That's immoral. I do not want to build games using a tool that is fundamentally immoral.

Lastly, generative AI "art" will never create something truly surprising as it is incapable of generating new ideas. People want to be surprised by art, and seek out new art to understand the perspectives of others. We have already hit a point where genAI is able to produce assets that are indistinguishable from handmade art, so there may be a brief window of time where generative AI products will sell, but it will not last.

Generative AI for Coding

GenAI for coding is tricky. To my knowledge it is impossible for any generative coding tool to be useful enough to generate code that doesn't require immense re-writing and scrutiny without having been trained on an inconceivably large breadth of data. You do not get that much data that quickly without cutting corners morally. Which means the best coding agents out there are built upon theft. If I built a game by digging through the code of two other games, not paying anything to the original programmers, copying and pasting lines into my game, sold it claiming to have written it all myself... I believe any sane person would consider that blatant theft. Diluting that theft across 1000s of games does not make it any less theft. Many of the major gen AI companies claim to source data legally, but that presents two problems. One, most legal systems are perpetually lagging behind the latest technological advancements. That which is legal is not inherently moral. Two, large multi-billion dollar companies are known to lie, so despite lax legal requirements, they are highly likely to be doing things illegally anyway. All this to say, generative AI coding tools are plagiarism machines. They are also fundamentally immoral.

Business Cases

My moral stance aside, generative code also conflicts with my business goals. I’ve organized them into two main business cases that have convinced me that it would not be good for our business to be generating code.

Business Case #1: Avoiding generative code is a safe bet.

We are at a crossroads today. One path leads us down a road of rapidly adopting prompting and agentic coding. The other leads us towards ignoring the technology. The middle path is inefficient as it will anger anti-gen-ai customers AND be ineffective at reaping the alleged benefits of full scale adoption. If my argument is wrong, and the future truly lies in prompting, some day will come when we have to change our methods and adopt a new way to code. This has already happened many times in my career, moving from language to language, engine to engine, etc. Every time it happens, I've adapted just fine because I build upon existing knowledge and a passion to learn. Not to mention the time spent going down the "wrong" road is not wasted, as we're still building our skills and refining our programming instincts that will inevitably help us to be more effective generative ai engineers if I'm wrong about all this and that ends up being some kind of necessity. However, if we adopt today in a rush to compete with others and it turns out this entire industry is a scam and we all must return to the "traditional" programming techniques, now we WILL have wasted time going down that road as we would not have been refining our problem solving skills in traditional programming. Of course these are both the worst case examples of both routes, but looking at all outcomes, its still a safer decision to stick to traditional programming for now. The middle path will widen and bend one way or another as the technology matures, and perhaps one day we will walk somewhere in that direction too. However the risk is too great to do so today.

Business Case #2: We have become reliant on yet another company.

We already pay for a variety of tools. Our margins are small as it is. As of November 2026, using coding agents is not that expensive, but it is almost a guarantee that this is not how it will remain. For these companies to recoup their costs in the sheer number of data centers they built and GPUs purchased and staff and investment etc, they will NEED to scale the profitability of their tools very intensely. What shape that takes doesn't matter, as it is beyond a doubt going to narrow our margins further. In theory the increased output of the studio will offset the cost of the tools we're using, however these tools are available to all. Which means if the tools are as effective as advertised, everyone's outputs are going to increase dramatically. We already live in an age of intense market saturation for video games, and players do not have enough time nor money to play all these games. This problem will only get worse as a result. So now we are not only chasing the horizon in terms of maximizing output to keep our margins from tightening as the tool we've become reliant upon rapidly increases in cost year over year, we are also trying to out perform every other studio also using the technology. I see no sustainable endpoint to this. At some point this market will have to crash. Not enough studios will make the hits they need to make to survive, and will fold as a result. They'll stop paying for these extremely expensive to operate generative models, who will raise costs on remaining customers to make their ends meet. Once that vicious cycle starts, there's no parachute for anyone involved.

Environmental Concerns

None of this addresses the elephant in the room: environmental impact. By financially supporting these companies, we are directly contributing to the construction of more data centers. Simply search the internet for "data center impacts" and ignore whatever generated answers you may get as I do not trust the industry to attempt to smother information regarding this topic. I do not have the time to articulate all the ways this infrastructure is making the lives of human beings worse today, and how badly it will affect the lives of generations to come. This is bad. The hyperscaling of data center construction is BAD. I cannot emphasize this enough, and some days I feel like this point ALONE should be enough to convince anyone not to touch this technology until we figure this out. I cannot build a career using a tool that contributes to the deterioration of the world we live in.

Well, Everyone Else is Doing It

There are many business decisions I could be making that could be making us a lot more money much faster and efficiently. However, I don't even consider debating most of them because our moral compasses are (presumably) aligned on them. This is why we as a studio do not hire desperate college students and pay them minimum wage simply because we could benefit from their labor with little to no ramifications otherwise. Plenty of studios have profited off of intense crunch production practices, leaving developers psychologically struggling for years afterwards. In fact, I'd argue most business strategies that make a quick buck tend to have moral quandaries attached to them. Many profitable businesses are profitable because they have no moral compass. How strict each business chooses to be with their compass can be detrimental to their own bottom line. That being said, just because many other companies are choosing to benefit from letting their compasses sway doesn't mean we should too. I am not interested in competing in moral grayness olympics as company after company nudges their goalposts closer and closer to apathy. I'm not saying our company is inherently better than any other company that chooses to use generative AI in some capacity. As I said, its moral grayness, not black and white. However, the thing about moral grayness is that unless you establish a threshold you are not willing to cross, you'll inevitably slide in the direction of immorality.

Closing Thoughts

I refute any amount of whataboutism to these arguments. Yes, other actions I take in my day to day life are contributing to the deterioration of the Earth. I drive regularly. I shop at companies that pollute our rivers. It pains me to think about, constantly. Therefore I think finding ways to reduce the harm we contribute without making it unreasonably difficult to simply exist in the world is worthwhile and good. Committing five wrongs is still better than committing six wrongs. Maybe our actions will inspire future generations to commit even fewer wrongs, rather than spiraling into further wrongs simply because other wrongs are too difficult to stop.

I like making games. I like the problems that I get to solve every day in my endeavor to make games. I want to enjoy living, and I want our next generation to enjoy living too. If you read all the way to the end, I love you. Thank you for reading, and thank you for caring.