As I'm finishing Lich's Egg, it's going to be the first game I finish that's a complete AI infused artifact from start to end, even though it's rather small. Most people now have converged on the solution of increasing scope due to game copying possibilities, as you can see in the replies to this post (I already answered one of them in #626):

I think that's a bit of a mistake. It's better to get used to getting your pipeline for game production working now while things are moving fast, and start increasing scope once things have stabilized. Opus 5.5 just released and it's a significant increase in game-making capabilities, there's no way to know where Anthropic will move next, and you kind of don't want to spend time doing things that take a lot of effort now only for them to be trivial in a few months.

So it seems like it's safer to make small games still, although, what "small" means now can indeed be bigger than before. I think copying concerns are largely irrelevant, 500 or so games release on Steam per week, 50 of those are popular, even for would be copiers it's a lot of surface area to attack, I think it's a concern that's way overblown. And even if it weren't overblown, now that games can be copied and decompiled and so on so easily, I could easily make the argument you're actually punished for spending more time on any individual game since it could theoretically be so trivially copied. In any case, for me it's clear the correct move is going small and increasing my scope slowly rather than trying to dream too big from the get go. Lich's Egg is a perfectly sized game, I liked how it came out, it allowed me to grasp the strengths and weaknesses of the current tools precisely.

The verdict is that they're good at pretty much everything except sounds, some types of juice, spatial awareness/modelling 3D spaces without references, modelling stylized characters still. There's probably more places where they lack but I haven't hit them in this particular game. As I said in #580, it's a truly competent gamedev companion, but taking a game to the finish line still takes quite a bit of work, I'm not sure that's going to change, for the same reason I argued back in #19 that game engines have an incentive problem with favoring the high percentage of people who never finish anything and not fixing hard problems for the low percentage of people who do. Anthropic has an incentive to make flashy one shot demos, they have no incentive to get the details of the last 20% of a game's development (which is 80% of the effort) right.

As for bartwe's post itself, my opinion on this has always been clear I hope, never take the doompill, this goes back to indiepocalypse days (#34). It's very easy to be demotivated, it's very easy to demotivate others, you need an anchor that grounds you in reality and allows you to assess the situation in a clearheaded, calm and collected manner. I will speak more on this in the next post, but AI ultimately forces us to ask the question of "why are we making games?"