PC, internet and mobile are 3 tech platform shifts I’ve lived through in my lifetime. It’s clearer than ever that AI is now the 4th.
A few years back during the initial mass market craze over chatGPT, my sentiment was “this is novel and seems promising (and inexplicable), but how will it actually matter?” But even then it was clear that this was something entirely different from VR, metaverse or crypto, 3 recent hypes that reflected a lot of wishful thinking from builders and investors – the winners and losers in mobile were largely set, and so the motivated parties collectively willed these new gold rushes into reality.
AI is different. A couple of months ago I started vibe coding at work in earnest, and my personal journey was illuminating. It has also been echoed many times by others (of varying levels of programming expertise) who’ve gone through a similar experience.
As background, I’ve never been a professional programmer, though I guess I had been well-positioned at several points in my life. My father was an early adopter of PCs in the early 90s, so I had the childhood memory of copying simple games code written in Basic from magazines (borrowed from the local library). In the late 90s, I learnt some rudimentary HTML to create a static webpage as part of the submission material to apply to an event for middle schoolers. And in college in the early 2000s I formally learnt C. I think I enjoyed learning programming – but I never ventured far in building things personally. (It’s one of the things I’ve puzzled over the most about myself – what were the missing ingredient(s) that stopped me from being more of a creator than a consumer? Complacency, broad yet shallow interests and the lack of persistence to stick through the “boring” bits are perhaps some of the culprits.)
All that’s to say – my perspective is clearly that of a novice coder. And this is one demographic that is primed to be bullish on where things are headed, because the ability to create working software (without relying on others) is immensely liberating. Anyone that collaborates with (and rely on) coders and are technically literate enough (or simply brave enough) to start vibe coding will quickly be drunk on this sensation – say, designers, project managers and product managers.
The biggest contributor to this sensation is the dramatically shortened feedback loop. If you have an idea, it’s now possible to experience the first approximation of that idea in a few hours, as opposed to days or weeks. And the next iteration might be mere minutes away. Paraphrasing a manager/mentor I had, “the best predictor of quality is the number of iterations the software has gone through” – and vibe coding changes the scale on this.
But what about code quality? Or tech debt? Or security risks? I’m not qualified to hand-wave these concerns away – but I will predict that these are stumbling blocks that will be worked out. We are still very early in this cycle – it’s not clear which IDE or LLM will be the consensus “best” a month from now, patterns are emerging but paradigms are not settled.
It’s also clear that the nature of coding work will change – as Ben Evans likes to say, “new tools start out being made to fit the existing workflows, but over time the workflows change to fit the tools.” If the economics make sense, we will adapt to the shortcomings of AI code while maximizing their strengths. This means what good code looks like – and what good coders look like – will shift too.
A version of this shift is happening in other parts of game dev – most notably, art, where it is coinciding with a brewing consumer backlash against genAI art. (For the purposes of this post, I can’t help but look past that debate and think about the possibilities for creators who are now granted the wings for art.) And zooming out to when the dust settles – it’s hard to imagine that the day-to-day of game-making will not change with AI-native workflows as core components. I’m fascinated by the thought of building a game team for that future, but I also have anxiety over the challenge of navigating current teams through a to-be-defined transition.
(Featured image: created via chatGPT.)