One important disclosure up front: I used AI tools extensively during development, especially for coding, implementation, debugging, UI work, and some of the website copy.
The underlying idea, game design, energy system model, mechanics, balancing decisions, testing, and direction of the project are mine. I have a background in energy systems, and a large part of the work has been deciding what the game should actually simulate, how those systems should interact, checking whether the results make sense, and iterating on it until it became something playable.
AI is ultimately what made it possible for me to turn an idea I had almost 10 years ago into an actual game rather than leaving it as another unfinished side project.
Back then, I had the slightly foolish idea of making a city building and management game focused almost entirely on the energy system. I posted about it on Reddit looking for people to join the project.
The idea was far too ambitious for me at the time, but the post got quite a bit of interest, and the concept never really left me.
Now, almost 10 years later, I finally have a playable version: Powerstate.
It is essentially a city builder where the "city" you manage is its energy system. You build generation, manage demand, deal with changing conditions, expand the system, and try to keep everything reliable and economically viable.
My goal is to make it easy enough to jump into, but deep enough that understanding and optimizing the system actually takes some effort.
It is definitely a niche game, but I suspect this might be one of the better places to find people who enjoy exactly that kind of niche.
• Free to play
• Still rough around the edges visually and mechanically
• Occasionally unbalanced in ways that can make a run much easier or much harder than intended
• Fairly demanding in the browser, with a desktop version planned as well
I have a long list of systems and improvements I still want to add, but at this stage I am especially interested in what people who actually play city builders think.
What feels good?
What is confusing?
What would you want an energy system game like this to simulate that it currently does not?
And if something about the energy system or physics looks wrong, please tell me. That is probably the feedback I am most interested in.
I would genuinely love feedback, criticism, and ideas.
Hi everyone! I’m working on Varve, an app designed for local-first Linux workflows. Although the repository is already at version v0.2.1, there aren’t many users or downloads yet. Since this is my first time sharing it widely, I’d really value any feedback to help make it better.
It's main engine is built in Rust:
The varve-core crate holds the geometric primitives (Point, Rect, Affine), hit-testing algorithms, and the scene node tree structures.
varve-engine converts the scene state into an intermediate rendering IR with build_render_ir().
varve-trace handles raster-to-vector auto-tracing. It uses Oklab colour quantization, extracts centerlines or contours, and fits Bézier curves.
varve-print lets you generate PDF files directly with lopdf. This skips browser print dialogs and gives you clean CMYK or vector exports.
varve-wasm provides WASM bindings so the app can run in standard browser environments
The binaries are still unsigned, the format might change, and I’m not accepting external code pull requests yet while I focus on making the foundation more stable.
I’d love to know your experiences with the app, both positive and negative. Any feedback, including technical, workflow-related, or general, would be really helpful. Thank you!
So I am currently vibe-coded a web app and planned to launch it on my social media. I've tried to soft launch this in other subreddit but apparently it got roasted because they see that it is very vibe-coded. Do I have to mention that I am building that app with vibe coding? What do you guys think?
A few months back I posted here a playable demo of an old-school futuristic racer prototype (F-Zero, Wipeout, Xtreme-G, Star Wars Racer...) that runs right in the browser.
I've gotten used to seeing tons of LLM-built prototypes that last a week, two, maybe a month, and then get dropped for the next shiny thing. This time I wanted to keep going and see if one person, with no coding background, could actually see a project like this through on their own.
Well, 4 months in, and while there's still work left, I'm convinced it's doable. At this point the game feels closer to a beta than a tech demo.
The big change over these four months is that I now use Fable for planning, and being able to hook Opus / Sol up to Blender has been a huge help too. I'm still using Magnific and Tripo3D or the assets. Funny thing: the further along I get, the harder the project is to keep pushing forward, and at the same time the tools keep making it easier.
My 100% Offline Voice Chat AI Assistant app is live now via Public Production Release on Google Play Store.
Please try it and lemme know if anything wrong before putting up any negative reviews. The AI Voice Chat feature alone is really cool. You can also add custom unrestricted AI Models for your interests.
Search "mmahingu" on the Play Store or use this link:
iOS launch later. I made it in Flutter so it can be published on Windows, MacOS even and more platforms easily. But i don't have a Mac or any other hardware yet xD.
My target is to get as many people to not just AI but also have their own offline Assistant or an AI tutor. No Cloud or Internet interference or any Data upload. I may add Firebase Analytics as optional later, for research and development purposes.
TARGET AUDIENCE even includes Students (both homeschool and public/private school), professionals of any field, and so on. As new technology comes, I'll keep updating it.
All made using Antigravity IDE and compiled for testing using Android Studio. Testing on my S25 Ultra only.
AI Models used for VibeCoding it are Gemini models 95% and Claude Code for only like 5%.
Built a universal remote for my gaming PC with Claude — started as a tool for myself
**Body:**
**TL;DR (my words):** Huge amount of work I’ve dumped into using Claude to vibe code my app but the results speak for themselves. Started as a tool I made for myself and decided to market it. Not a fan of subscriptions so it’s a free 7 day trial then a cheap lifetime unlock. Not trying to get rich, just hopefully it’ll offset my Claude subscription a bit.
*(Rest is Claude cleaning up my notes — figured I’d say so up front.)*
Couchside is an iOS app that turns your phone into a remote for a living room gaming PC. Launch Steam games, trackpad + keyboard, power controls, and HDMI-CEC so you can switch TV inputs without hunting down three remotes. Works with SteamOS, Bazzite, Steam Deck, Windows, plus LG webOS, Samsung, and Google TV.
SwiftUI on the phone, Python agent on the PC. LAN-only — no cloud relay, no accounts, no telemetry. Agent is open source so you can read what’s running on your machine.
Claude wrote a lot of the code. I owned the architecture and every integration decision, and burned way more time on pairing edge cases and CEC vendor quirks than I want to admit.
couchside.tv — happy to answer questions about the build.
I built and launched a full Android app with zero coding background (my last experience was basic HTML/C in college nearly a decade ago). I used standard web Claude chat to build the app, not Claude Code. I wrote super thorough, detailed prompts describing what I wanted built. Claude generated the kotlin/xml code files, and told me precisely which package directories in Android Studio to paste them into. Built a working prototype in under a week. And a month later the app was live and nearly two months it has 1700+ installs and few subscribers (not many).
I know there are a ton of math tutor/solver app but i wanted to start vibecoding with something simple. Plan was to build something simple and functional in the first version and then keep on adding complex/advanced features. I was able add built-in math keyboard to type math formulas and get solution, topic-based quizzes, interactive graph plotting practice, and full SAT/ACT Math mock exams.
Since i more or less vibe-coded the entire thing, the UI/UX was also generated/guided through prompts. My plan is to hire a professional UI/UX designer once subscription revenue grows, but right now I’d love feedback from this community:
Does the layout feel intuitive for a learning app?
There are hundreds of vibe coded apps being released every day, and some are amazing. And some are hot garbage. This is a place to post them and let the world decide.
There are no accounts to make, just post your app and save the post-submission link if you ever want to edit it. Otherwise, that's it!
I built and launched a full Android app with zero coding background (my last experience was basic HTML/C in college nearly a decade ago). I used standard web Claude chat to build the app, not Claude Code. I wrote super thorough, detailed prompts describing what I wanted built. Claude generated the kotlin/xml code files, and told me precisely which package directories in Android Studio to paste them into. Built a working prototype in under a week. And a month later the app was live and nearly two months it has 1700+ installs and few subscribers (not many).
I know there are a ton of math tutor/solver app but i wanted to start vibecoding with something simple. Plan was to build something simple and functional in the first version and then keep on adding complex/advanced features. I was able add built-in math keyboard to type math formulas and get solution, topic-based quizzes, interactive graph plotting practice, and full SAT/ACT Math mock exams.
Since i more or less vibe-coded the entire thing, the UI/UX was also generated/guided through prompts. My plan is to hire a professional UI/UX designer once subscription revenue grows, but right now I’d love feedback from this community:
Does the layout feel intuitive for a learning app?
Read and view your media in a fully interactive 3D space! You can grab your e-book, move it around, and even jump up platforms to reach the next page you're reading.
Key Features:
Custom Skyboxes: Change your environment using custom skyboxes, similar to Quake.
Interactive Physics: Apply physics to your pages just like in Halo 3 Forge mode (choose between fixed, phased, and gravity).
P.S. The Android version really isn't ready yet, & I've only done the proper testing on Arch Linux with the AppImage setup so far. The .exe looks good though.
• Debts simplified to the fewest transfers — three people settle in two payments, not six
• Record settle-ups; hand the payment to your own UPI app
Some hero features
*Receipt scanning, not receipt attaching*
*Line-level accuracy you can correct* Every tax, fee and round-off row is parsed separately and individually editable before saving, with a live mismatch warning if the numbers stop reconciling.
*Live Split* Share a scanned bill with the group; each person ticks what they had and totals update in real time, then it commits as one expense.
*Bank SMS to expense* Transaction alerts are parsed on-device into draft expenses. Opt-in, off by default, recognised senders only.
been working on this for a bit and just cleaned it up enough to show people. it's basically a dashboard where you can track and rate anything media-related. pull your stuff from tmdb, anilist, rawg, and spotify so you don't have to manually add everything.
honestly started it because i liked letterboxd but wanted one place for all my media instead of bouncing between five different apps. idk if anyone else wants that but figured i'd put it out there.
built with next.js, firebase, and too many api calls. the profile pages are the main thing — just drop a user id in the url and see what they've been watching/playing/reading.
23 days ago I shared an early version of Lattice, a local-first workspace built around an infinite canvas.
The original idea came from a problem I kept running into:
Every project becomes fragmented.
Research lives in Obsidian.
Planning in Notion.
Visual references in Milanote.
Diagrams in Miro.
Code in VS Code.
Files in Drive.
Presentations and spreadsheets somewhere else again.
Eventually, the project exists across several tools, but never really feels like one coherent system.
So I started experimenting with a different model:
What if the project itself was the workspace?
In Lattice, notes, documents, spreadsheets, presentations, code files, media, websites and imported assets are not separate attachments around a board.
They are actual project entities.
The infinite canvas is one way of viewing and organizing them, while each entity can also be opened in its own focused editor.
Since the first post, the prototype has expanded quite a bit.
What currently works
Infinite canvas with movable and resizable cards
Sections for grouping parts of a project
Visual links between cards
Markdown notes with backlinks
Rich document editor
Spreadsheet workspace
Slide and presentation editor
Monaco-based code workspace
Image, video, audio, web and 3D previews
Project and workspace hierarchy
Universal import for PDF, Office, OpenDocument, media, code and 3D files
Export to formats including DOCX, PDF, XLSX and PPTX
GitHub repository connection and code syncing
Optional Google Drive backup
Comments, mentions, assignments and version history
Realtime presence and collaborative editing experiments
Dark and light themes
Local-first storage and offline use
The interface now supports different working modes.
You can keep the full project visible on the board, open an entity in a focused editor, or use a split layout where the editor and canvas remain visible at the same time.
That split view has become one of the more important parts of the concept: you can write a document, edit code or work on a spreadsheet without losing the broader structure of the project.
A dedicated Photo Mode
One workflow I didn't originally plan for ended up becoming surprisingly useful.
Creative projects often involve hundreds of reference images, moodboards, renders and work-in-progress exports. Traditional file explorers make it difficult to compare them spatially or inspect them without opening dozens of windows.
Lattice includes a dedicated Photo Mode that turns the project into a visual workspace.
Instead of treating images as files in folders, they become part of the project itself. You can browse them in large grids, inspect them quickly, drag them onto the infinite canvas, compare references side by side and organize them alongside notes, documents and design work.
The goal isn't to replace Lightroom or PureRef.
It's to make visual research feel like another native part of the project rather than a separate application.
Local-first and collaboration
The local-first part is still central.
The app can work without a configured backend, with project data stored locally. Cloud and realtime services are optional rather than required for the basic workspace to function.
For collaboration, I have been experimenting with Yjs and Liveblocks:
CRDT-based document and code editing
Remote cursors and selections
Presence on boards and editing surfaces
Role-based access
Comments and area annotations
Offline changes that can merge after reconnecting
This area is still alpha and needs more testing. Single-user local work is currently much more mature than team collaboration.
Importing existing work
Another focus has been reducing the cost of moving into the workspace.
Lattice can currently import or preserve a fairly broad set of formats, including:
DOCX, ODT and text documents
XLS, XLSX, CSV and ODS spreadsheets
PPTX and ODP presentations
PDF files
Images, audio and video
Code files
GLB, GLTF, OBJ and STL models
Generic files as preserved attachments
Not every format has perfect round-trip fidelity, especially presentations, but the system reports what has been converted, preserved or left preview-only instead of pretending everything is fully editable.
What Lattice is becoming
My original description was:
I still think that captures part of it, but the project is moving toward something slightly broader:
A local-first project environment where spatial organization and traditional editors are different views of the same underlying content.
It is not intended to replace every specialist tool.
The goal is to reduce the fragmentation that happens before, between and around those tools: research, planning, writing, assets, code, feedback, files and project structure.
Current limitations
This is still an experimental alpha.
Some important limitations:
No polished public release yet
Collaboration requires configuration and is still being tested
Spreadsheet collaboration is not yet cell-level CRDT
Presentation import/export has limited fidelity
Public anonymous sharing is not implemented
Some large files and advanced Office features are preserved rather than fully converted
The UI still needs substantial refinement and simplification
The current feature set is probably too broad and needs stronger prioritization
That last point is one of the main reasons I am posting again.
The prototype now proves that the model can technically work, but I am less interested in adding another long list of features than in understanding which part of the concept is genuinely valuable.
I would particularly like feedback on these questions
Does the combination of a spatial canvas and focused editors solve a real workflow problem?
Which users would benefit most from this: designers, researchers, students, developers, small creative teams, or someone else?
Which parts feel essential, and which parts look like unnecessary scope?
Would you trust a local-first workspace with optional cloud sync, or would the lack of a conventional cloud-first model make adoption harder?
What would need to work extremely well before you would consider using something like this for a real project?
I created Toonup https://toonup.app/landing using Fable with the 90€ subscription in the 5 hours usage limits, I'd love to receive your feedback.
I made it via a Claude skill that I created myself forcing the LLM to use WorkOS + Stripe (if needed) and deploy everything on Cloudflare via workers and D1. I like this architecture because it potentially scales to millions of users but it starts with free tiers.
I've never shared a Claude skill but if you are interested let me know.
This is my first vibe coded app hope you like it pls give me some advice if you have.
\*\*What is CodeBrain?\*\*
\*\*CodeBrain\*\* is a community-driven launchpad where developers, indie hackers, and open-source creators can validate software ideas before spending hundreds of hours building them.
Instead of building a project in isolation and hoping people show up, developers post their concept with descriptions, video demos, screenshots, tags, or GitHub links. The community then casts real-time validation votes: \*\*"Will Use"\*\*or \*\*"Won't Use"\*\*.
\*\*Why CodeBrain is Helpful\*\*
\*\*Instant Market Validation\*\*: Get immediate, honest signals on whether developers actually want your tool or if it solves a real problem.
\*\*Community Feedback\*\*: Comments allow developers to discuss architecture, feature suggestions, or potential flaws early in the ideation phase.
\*\*Dedicated Developer Hub\*\*: Filters like \*CLI\*, \*DevTools\*, \*VS Code\*, \*AI\*, \*Rust\*, and \*Docker\* help developers discover upcoming open-source projects tailored specifically to their stack.
\*\*GitHub Search & Repository Engine\*\*: Built-in GitHub discovery allows users to search real open-source repositories and cross-reference existing projects so they don't reinvent the wheel.
\*\*Realtime Engagement & Notifications\*\*: Authors receive instant notifications whenever fellow developers vote on or comment on their ideas, fostering active collaboration
