Alex So

@0PercentSteam @bunjavascript @PixiJS @boshen_c I've tried itch couple of times but it didn't work well for me (~200 or so users in total).

@afrokick @bunjavascript @PixiJS @boshen_c Consider using itch to gain traffic but you could also host it on your own platform. Link from itch to your game. For desktop you wanna try to get it on steam, feel free to hit me up if you need suggestions. Hope to see you out there at some point :) !! keep it up

Making MMO (already published) with @bunjavascript and @PixiJS - Bun as a game server runtime and tooling(RIP bash) - Pixi as the lightweight rendering library(it supports KTX as well!) - SQLite for player's data - in-memory Redis for leaderboards - oxlint/oxfmt by @boshen_c The game includes PvE, global events, Enemies AI etc. And you know what? - The server handles 250+ players(ws connections) on a single thread without any troubles(10ms per server frame). - The client draws thousands of objects(including particles and texts) in a few draw calls - 0 deps on the server(only bun) - It costs me $10/month because of cheap german vps + free cloudflare cdn. I definitely recommend this setup for game development!

@0PercentSteam @bunjavascript @PixiJS @boshen_c Thank you! This is a WEB game, you are right. Using Electron/Tauri/etc it could be wrapped into native game. The plan was to test the game on web first and then make mobile/desktop releases.

@afrokick @bunjavascript @PixiJS @boshen_c Impressive stuff, you make it sound simple but this takes skill. Also the game itself is pretty catchy looking, any plans to publish on steam? It does feel like a browser first game if im honest

DAU for the last 90d. 3k DAU give me ~50CCU.

@AGulev @karpathy @defold Could you share the ideas here? Really interesting to see what was on the list

Yesterday, I decided to try the auto-research concept by @karpathy and look for ways to reduce the web build size for the @defold Game Engine. For this, I wrote my own looper script. The idea is a loop that runs an agent on a specific task, while an external script runs a benchmark to evaluate whether the solution improves things. The results are then fed back to the agent, which is asked to try another solution. Essentially, it’s a loop focused on improving a single parameter. --- How it works: 1. Run the benchmark once to capture the baseline. 2. Ask the selected agent to improve the project. 3. Run the benchmark again and compare the result. 4. If the result is better, keep it and commit eligible changes. 5. If the result is worse or unchanged, revert non-protected agent changes. 6. If the benchmark returns an error text, ask the agent to fix the error first. 7. Repeat until -iter-count iterations are completed. --- I use @OpenAIDevs ChatGPT 5.4 thanks to their open source support program! I primarily used it with Codex CLI, but also tested @opencode, which works well too. Before the main run, I executed a few iterations and noticed that it tended to take a "lazy path", relying on compiler optimization flag `-Oz` to reduce build size, which wasn’t what I wanted. So I asked Codex the same question: find the top 30 optimizations for web build size. As expected, it returned 30 fairly uninteresting suggestions, such as compiler flags and minor tweaks, along with repeated advice to use functionality we already provide for engine size reduction. Cheater! Ooook, I took these top 30 optimizations and explicitly excluded them in my prompt for the looper, instructing it not to use any of them and to find alternative approaches. I ran the loop overnight again for 20 iterations. This morning, I reviewed the results and found that it produced a couple of interesting ideas. I don’t like the implementation, but I do like the direction. My main goal was to get new ideas, not to fully automate the process. Now I'm implementing the best 3 out of the 17 ideas it found. It looks promising. I'm able to reduce the release binary size by about 5%, which is a solid result considering how well optimized our build size already is. I think 2 more ideas from this list are worth attention as well. So, I expect to reduce the build size by around 6% using this approach.

@YoavCodes @ivanpopelyshev Looks great! Any chance to support mobiles in near future?

Electrobun + WGPU is here. Read the full blog post: blackboard.sh/blog/wgpu-in-e…

@jarredsumner Does it isolate your crons from the others?

This uses your operating system’s scheduler so it doesn’t need to keep the process running to schedule a job, but each invocation is run as a separate process This is useful for CLIs or long-running servers that perform regular but infrequent cleanup

In the next version of Bun `Bun.cron(file, schedule, name)` schedules a function to be called on a recurring interval

@d954mas You should try phaser or Pixi. HTML is great, but you still need a lot of stuff like particles, tweens, atlases etc. and AI works great with those libraries

Drill vs Snake 🐍🛠️ A lot of improvements. Interesting project, it is realy fun to vibecode. AI worked good with html and js. It is easy and fast to make cute prototype. But html is really bad in performance. #gamedev #vibecoding #aidev #Claude #screenshotsaturday

@anulagarwal Without marketing it doesn’t have any sense. You’re trying to compete with millions of apps

2010: solo devs built websites 2015: solo devs built SaaS 2020: solo devs built no-code tools 2026? native mobile apps are the new passive income play i build iOS apps that make ~$450/mo without any maintenance heres what my stack costs: • xcode: $0 • claude code for coding: $100/mo (used for other purposes too) • app store connect: $99/year ($8/mo) • no backend servers: $0 • no infrastructure headaches: $0 total monthly cost: $108 revenue potential: subscriptions while you sleep saas needs constant babysitting mobile apps just run recently shipped 15 apps in 10 weeks with ai • you can start today • total cost: under ~$100/mo • no server management, no scaling issues, no 3am alerts • a decent app can hit $2k-5k/month recurring the travel laptop lifestyle actually works now apps earn while youre offline no infrastructure to maintain just ship and collect subscriptions anyone else building mobile instead of saas these days?

I made another javascript car 🚗 #threejs @JamangoGame

navcat, meet world. world, meet navcat. navcat is a new javascript pathfinding library for games and simulations, and it's fully open source ❤️ it does navmesh generation, pathfinding, crowd simulation, navmesh constrained movement, raycasting, more... and it's out on npm now

My teammate just finished a new map… So talented person! Unmute the video for a moment to get the best vibe. Next WoW soon? Powered by jamango with typescript, @threejs and @bunjavascript (yes, that’s an online multiplayer experience!)

@d954mas @Poki Congrats man! Thank you for sharing your dev journey! And the game is awesome!

🎉 Drills Merge Master is global on @Poki - Built in just 4 months ⏱️ - My first game with pets 🐾🐉 Happy with result, fun idle game i love to play:) poki.com/en/g/drills-me… #madewithdefold #webgame #idle

@bunjavascript @webstandards_ru worth to mention in the next episode 😉

Bun is now on MDN!

@isaac_mason_ @robotiguy @babylonjs might be a good addition for bjs

My javascript pathfinding library is now largely in the "document and stabilize" stage ahead of a v1 release. Here's a new example "custom areas" - I'm marking areas during generation as either "ground" or "water" areas, and writing a query filter to limit the search.

Special Announcement! Babylon contributor Jas has added Clustered Lights support to Babylon! Now you can add a TON of lights to your scene while maintaining performance! doc.babylonjs.com/features/featu… #3D #WebDev #gamedev #indiedev #WebDevelopment #webgl #gamedevelopment #IndieDevs

@jarredsumner Any logger library like winston with built in adapters for fs/aws/logstash

What else should we add to this list?

When I was a kid I didn't have a trampoline... So... #webgamedev #mangolango

Sneak peek at "nav3d", my upcoming javascript navmesh generation and querying library. If you used recast-navigation-js and enjoyed it, but felt sads about the big wasm bundle and wasm quirks (manual memory cleanup, js->wasm call overheads), then watch this space :) #threejs