Michael Duan

@UnityCodeMonkey There may be a slight misunderstanding. I mean that if I personally prioritize making money or chasing some trend, what I make turns out to not be very fun and don't even make money anyway. Maybe it's not true for everyone but it is for me though.

@e_Zinc If game dev is something you do for fun then yup feel free to ignore anything related to money. But if it's your full time job, if you're dependent on game dev success to not starve and not go homeless then I really recommend you focus on the finances of the game dev business.

If the goal is to publish on Steam to earn some money, at what point during development do you seriously have to consider "will my idea actually be worth publishing"? If that's your goal then that has to be the focus from day 1. Picking the right game idea is absolutely crucial if you want to find success. You have to pick the right genre right from the start. You cannot really start from an unmarketable idea and make it marketable afterwards. Then once you have what you think is a solid idea, you need to validate it. That's when you make a prototype or some gifs, post them on Reddit / Twitter to get some playtesters, and try to figure out whether the idea has potential, or whether you should go back to the drawing board. If you get absolutely no traction then it might be wise to kill the project right then and move on to a new idea as opposed to trying to salvage and idea that apparently players don't seem to want. But the most important thing is being very clear about your goal. If you're doing it for fun then do whatever you want, but if the goal is financial success then you need to treat it very seriously right from the start. Treat it like a professional, like serious job, because it is. Best of luck!

@bmcnett I see game dev as true modern art... play this game to consume 10 years of this dev's life. It's like visiting an intricate ancient chapel, but if only 1 person made it. Pretty hardcore and avant-garde. If you're just talking about money, then yea I suppose so.

i wish the best for these folks, but it is so depressing to hear about another 10 year dev cycle. i want more games made in 10 months

Gameplay based: if you’re gaining points rapidly, pitch them up as you go. If it’s a hard hit, etc. Material sounds: modern games don’t do this anymore but physically based sounds for each material. Physics impacts, hits, etc. Fatigue: for long play type games, need to have enough bass, not blow out your ears with high pitched noises, and have a mix of long/short tail as needed. It’s something TV/film sound designers won’t understand at first since their media isn’t designed to be as replayable. Ego fantasy: should always serve the player’s ego and fantasy.

Any advice for sound effects for games? Not talking about creating them but every other aspect. Common pitfalls Any advice is appreciated

Been working on this huge Saleblazers update for months now. Very excited to release it. Fixed 70% of the things that bothered me and added Half Life 2 / Deus Ex / GTA vibes through physics, NPC behavior, and more immersive systems. I finally enjoy my own game, after 7 years…

Gamedev is strange. It's hard to explain why I decided to prioritize adding wet clothes + new rain/water FX over some other things. It's cosmetic after all. It's just something that I felt needed to be done when I played the build. If a business person or producer looked at this they'd definitely not allow me to waste time working on this. And logically they’d probably be right.

Gamedev lesson: adding an emotion system to your characters makes for hilarious situations. Even when it bugs out, it’s still funny.

If they just make his 4 channel instantly or while being able to move, it’d be good enough imo. It’s because all the new heroes/changes are way too mobile for his 4 now. With Ivy’s insta cast and Mirage double 2 it seems they’re just buffing movement of old heroes to match new heroes.

Not sure if valve will ever nail down Warden’s balance. For him to be frontline (which is seemingly what they want him to be) he needs to have his power shifted in favor of his ult and 2, which seemingly goes against their design philosophy in terms of ults not defining kits

Zenyatta from Overwatch

@WazzCrypto Maybe they did it on purpose so that when they easily remove it people will believe it got significantly smarter, or so people can’t tell if it’s written by AI anymore. It’s a known practice in software since it’s so easy to live update.

Ok I need to know where did all these LLMs scrape the "this isn't X, it's Y" speech pattern because who the hell talks like that every other sentence

When everyone’s super, no one is. It’s because yes you are able to output 10-20x more than before, but so is everyone else. Arguably even harder to sell software now because people can just make their own which affects revenue and stock valuation. Less about what people can code with AI and more that if you don’t use it you’re just unimaginably behind.

we are being gaslit about AI on a societal level. Everybody is vibe coding but I haven’t seen one useful thing get produced. Everybody has agents doing something but nothing useful is getting done. Cool you had AI summarize a PDF and make a template. Nice

It is possible that if he has a higher internal MMR that he will have worse teammates both with you and against you. Or you have bad natural and hero synergy with him. In these types of games especially with low matchmaking player count you can’t look at this stat. Have to look at absolute win rate per hero controlled for top 500 (not just Eternus 6) hero MMR.

On the same team as Metro 38% winrate. Metro on enemy team? 64% winrate

@Bingus12621 @iiKlemm @Neuroptima Removes brain fog and makes your body feel like you’re on an extra 20% sleep. Definitely should try it. It takes 1-2 weeks to kick in though.

@iiKlemm @Neuroptima Got a 9 month old and sleep has been very bad. I am trying to read some dense legal text and it feels like putting square in the square hole. Would creatine help? How does it help?

Competitive gamers: Just one night of poor sleep can impair cognitive performance by 20-30% (similar to being legally intoxicated). A single 0.35 g/kg dose of creatine boosts cognitive performance during sleep deprivation within 4 hrs. If you aren't sleepmaxxing, take creatine.

@rohanpaul_ai This isn’t just books but also images. You can get the Warcraft 3 or World of Warcraft cursors 1:1 in Nano banana pro by asking for an RPG gauntlet Warcraft style mouse cursor

New Stanford paper shows production LLMs can leak near exact book text, with Claude 3.7 Sonnet hitting 95.8%. The big deal is that many companies and courts assume production LLMs are safe because they have filters, refusals, and safety layers that stop copying. This paper directly tests that assumption and shows it is false in multiple real systems. The authors are not guessing or theorizing, they actually pull long, near exact book passages out of models people use today. A production LLM is the kind people use through a company app, and it can memorize chunks of books from the text used to teach it. The authors test leakage by giving a book’s opening words, asking for the next lines, then repeating short follow ups until the model stops. When the model refuses, they try many small wording changes and keep the first version that continues the text. They run this on 4 production systems across 13 books, and they use near verbatim recall, which only counts long, continuous matches. That matters because safety filters, meaning built in rules that try to block copying, can still miss memorized passages in normal use. ---- Paper Link – arxiv. org/abs/2601.02671 Paper Title: "Extracting books from production language models"

@krzyzanowskim I don’t think this is true because Claude can read memory dumps quite efficiently and plan refactors accordingly. It is just a matter of people prioritizing features over optimization right now but I think that will change since RAM is getting expensive

new thesis just dropped: every vibe coded software will eventually turn into insufferable slop without chance to improve claude:

@gwenshap I’ve had it do these by default especially the over engineered redundant code, but a good .md prevents them. It’s because Claude wants to save tokens so it isn’t ultra-thinking all the time. You can also prompt it to use AskUserQuestion whenever it is slightly unsure.

Since my feed is full of people who didn't code in months because Opus 4.5 is so good, let me chime in with 5 things that I had to manually fix this morning because Claude Code didn't: - Separation of concerns: I asked for new functionality that belonged in a specific class. I told Claude to review that class and make the changes. It made the changes in an unrelated class and then moved some functionality from the class I wanted into a second unrelated class. - Encapsulation. A bunch of private and package-private methods were made public with no good reason. - Concurrency: It implemented a rather blatant concurrency bug and kept insisting that the logic is correct. - Reinventing the wheel: The concurrency bug was in a queue library that didn't need to get written because Guava exists. - Use the wrong build command, then misinterpret the error and make a bunch of totally unnecessary changes. In a loop, since the wrong build command kept failing after each round of useless changes. And yes, I have plan documents, rules document, and I prompted very carefully. Claude Code is impressively good, but human review is very much needed. Especially for concurrency.

@Loopify Idea 1: Have expensive AI write an XML file and use a cheaper model to read it then kick off expensive AI with class names / functions. Idea 2: Learn the codebase and give it explicit class names yourself. This is easier if you coded the foundation yourself.

for vibecoding/claude code experts, when I begin a new session and Claude needs to reread my codebase to understand how to implement a big feature I burn so many tokens how do I be more efficient? do I include all the codebase context inside of my claudeMD file?

@gregisenberg That’s why I think the real key to all this AI code will be personally generated software that you and your family use yourself. B2C and B2B demands aren’t really compatible with AI, because AI is inherently too personal.

a huge number of “ai products” feel impressive for 72 hours and then quietly exit your life without friction, which is starting to look like the defining failure mode of this cycle

@GeoffreyHuntley Why don’t you just buy extra requests? Is it cheaper to have multiple subscriptions?

oh, you only have one claude code subscription? thats cute but it's not enough

Need fuller stream of consciousness for small tasks since AI tends to over-complicate them compared to large tasks. 1. Giving it bullet points is necessary 2. If not in the .md already, tell it to keep it simple + reuse code + optimize for performance 3. Handhold exact classes, variables, function, inputs/outputs

Where are my gainfully employed software engineers who use AI tools (vibe coders just sit this one out - here’s some crayons to play with) Am I using it wrong, or is it just faster to do smaller / non-obvious tasks by hand, vs writing a clear prompt and then verifying?