I've been dreaming of this C++ profiler for my game for over a year. It's now here. It shows me the slowest code logic that runs over my game components in real time. Each one is a lambda that I can find at the file name and line number. An hour after I made this, I already optimized my game significantly by actually seeing what was slow and then optimizing it. For example, I optimized a C++ lambda from 750us down to 50us (cost per frame). This level of introspection into my game is god-like vision. Not seeing any perf impact of having this on either, so it's on by default.

To all whiny game devs who are crying about AI slop in Steam Next Fest, go read the data on the previous Fest for this year, before collecting tears, group hugs and demanding that "something must be done!" howtomarketagame.com/2026/07…

In this messy discourse of ai-slop this, slop that, what's real gamedev, ceos etc.. I want to remind all developers of this. Do not buy into the fear mongering that #indiedev is dead. The amount of games you compete with is _the same_, literally hasn't changed in years.

Control versus Speed. Whenever I use AI, I feel this tension of a trade off between the two. I can create faster with AI but give up control over the final product. Without AI the progress is slower but I maintain control over the final product because I understand how it is built. This graphic was made with AI because I did not need a high level of control and precision to express my idea. For any work that requires little control and precision, such as a Blender script to do some minor tasks, AI is amazing. If you are making a game and want to express an idea with a high level of precision and control over how the game feels, plays and looks, then AI as a code is a wrong trade off. AI has to be used more as a knowledge assistant and code reviewer while you build. Whenever a game dev says that AI coding is great for their game, that means he accepted less control over the final product. Is a product better with "more stuff" of lower quality? Depends on the product. Is it easier to polish in the long run with AI? Is it faster to polish a game that is 200,000 lines of C++ with AI? I doubt. My is 200k of C++ and I still only accept tiny bug fixes from AI. But if someone has a game like that, I would love to read their blog on the cost of polishing a game of a significant scope and size. The cost of tokens and context must be high.

Game devs are so whiny that they invent ghosts out of thin air and then tremble in fear over it. Does anyone know of a Roblox copy that made a ton of money over a viral game idea from X? Grok sees no examples.

Order of the Sinking Star has been out for a week now, so I decided to run some stats on Steam where it's easy. The Witness was our previous game that was very successful when launched 10 years ago. Order has sold 45% more units on Steam in its first week than Witness did. Order is 25% cheaper (Witness was $40, Order is $30 with some bundle discounts, and there's been some inflation), so you would expect more units at the lower price. But regardless, it looks at least as financially successful as the Witness was, maybe a little more eventually, depending on how the tail goes. Thanks everyone for supporting the game, and I'm glad you are enjoying it!

Valve was told that Steam will never happen. 99% of the companies they spoke to said: "Users are not going to want this. It will never ever happen." Valve went against everyone. And now we turn round and call them a monopoly for the work they did in spite of everyone! So much for not listening to studies and statistics, huh? > Gabe “It was a very weird time. Like, I don't think people understand how many times we would go to people and say, ‘No, you will be able to distribute software over the internet,’ and have people just say, ‘No, it will never happen.’ I'm not talking about one or two people. I mean, like 99% of the companies we talked to said, ‘It will never happen. Your retail sales force will never let it happen.’ But also people would say, ‘Users aren't going to want this.’”

Nevermind, my hope was wrong. It is still WebView! 🤡

An actual first real AI side product from an AI company. Nice.