WCNegentropy

I vibecoded an open source MIT CLI tool and desktop app called Retro Vibecoder that allows you or your agent to procedurally generate boilerplate, working, fully scaffolded tech stacks from a single input seed. Then you just implement your custom logic. Saves thousands of tokens and tons of time! It's a Universal Project Generator (UPG) that procedurally generates fully-scaffolded, production-ready boilerplate projects from a single seed number. Think of it like Minecraft for software architecture—deterministic generation with constraint-based assembly instead of copy-paste templating. Why this matters for AI coding workflows: When working with any AI agent, you typically burn hundreds or thousands of tokens just describing your project structure, dependencies, and configuration files. With Retro Vibecoder, you skip all that: * Generate complete project scaffolds instantly (backend APIs, web apps, mobile apps, desktop apps, games, CLIs, libraries) * Focus your context window on actual business logic instead of boilerplate * Consistent, working configurations across 30+ tech stacks * Perfect for agentic workflows—Claude can generate the seed, you generate the project Supported Stacks: ✅ Web: React, Vue, Svelte, Next.js, SvelteKit, Angular, Qwik ✅ Backend: Express, FastAPI, Django, Axum, Actix, Gin, Echo, Spring Boot, .NET, Rails, Laravel ✅ Mobile: React Native (Expo), SwiftUI, Jetpack Compose, Flutter ✅ Desktop: Tauri, Electron, Qt, Flutter Desktop ✅ Game: Phaser, PixiJS, Bevy, Unity, Godot, SDL2, SFML ✅ Systems: Rust, Go, C++, Python, TypeScript/JavaScript How it works: bash # Install CLI (remove the brackets around @) npm install i [@]wcnegentropy/cli # Generate a project from a seed upg seed 42 --archetype backend --language typescript --output ./my-api # Or run a procedural sweep to discover valid configurations upg sweep --count 100 --validate Desktop App: Also available as a standalone desktop application with a retro UI—no CLI needed. Download from the releases page: github.com/WCNegentropy/r… For Agentic Developers: The constraint solver ensures every generated project has compatible dependencies and proper structure. Your agent can: * Request specific archetypes + languages * Get deterministic output from seeds * Generate multiple project variations quickly * Validate configurations before generation Token Savings Example: Instead of prompting Claude with: "Create a NestJS project with TypeORM, PostgreSQL, Docker, GitHub Actions CI, ESLint, Prettier, Jest tests, and proper TypeScript configuration..." You just run: bashupg seed 12345 --archetype backend --language typescript --framework nestjs 100% MIT Licensed & Open Source Built with Claude Code over the past few weeks as a solo dev project. The tool itself and all generated templates are MIT licensed—fork it, extend it, use it commercially, whatever you need. Links: * GitHub: github.com/WCNegentropy/r… * Latest Release (0.2.2): github.com/WCNegentropy/r… * CLI (remove the brackets around @): npm install i [@]wcnegentropy/cli Would love feedback from the community, especially on: * Which additional frameworks/stacks you'd like to see supported * How you're integrating this into your Claude workflows * Any bugs or edge cases you encounter Built this to solve my own token-burning problem when scaffolding projects with Claude. Hope it helps others in the community too!

@goodalexander @DeepDishEnjoyer And then there’s the weirdos like me that just drink it for the taste, any hour of the day or night, yet can go months without having any just fine 🤔🤷🏻‍♂️

@DeepDishEnjoyer cool thing about quitting coffee is you quit and then you stabilize and you're like "hey, life is objectively worse without coffee" and then you start again

love conducting real life case studies on how poorly i function without my daily coffee

@Kisliy_K @C8Luna @samuel_0xK Meanwhile the real ones have moved to Hermes agent running cheap or free OS models either through inference providers or locally 😉 Hermes Agent CLI + DeepSeek-v4-Pro or even Flash is SO much better than my Claude or Codex subscriptions were. That includes Claude Max 5x

@C8Luna @samuel_0xK Well, nowadays you need to be fast adapting to reality. As well to the prices and quality pf product. So actually nothing will be changed

Github copilot isn't usable anymore :( just one single small codereview costs 61 credits???? do you still use github copilot or did you switch after June 1?

@appealstoheaven @cheesealloverme @michael_waples @0xdippo @thsottiaux Oh okay then it’s just back to you misunderstanding the whole situation and complaining about getting the tokens you already used refilled and still getting your other four windows this month, got it! 😉

@WCNegentropy @cheesealloverme @michael_waples @0xdippo @thsottiaux Yeah, none of that was ever a part of this discussion haha. I mean good try though I guess

Hi. Over the last 24 hours we had three separate small incidents that affected Codex reliability. Those are three too many and we are taking active steps for them to not reproduce. I have reset usage limits for Codex across all paid plans. May the tokens flow again.

@marclou They’re good at backend if you don’t look at it too closely (This is sarcasm, always review AI code lol)

AI is so good at backend, but so bad at UI/UX. Any recent models one-shot my new features, but I'd have to spend another 10+ prompts to get the design right.

Meta's AI support bot could reset passwords and add recovery emails for any account. Attack requirement: a username and a fresh email. That's it. Worst part: OWASP published guidance on this exact vulnerability (LLM08: Excessive Agency) in 2023. Three years of warnings, ignored.

@kimmonismus Preview for red teamers, I should add, not GA or gated rollout yet

@kimmonismus It was already released in preview and then paused due to someone in China reselling API access

Get ready, friends. Anthropic appears to be preparing the release of its Mythos-level model. Pricing: $16 per 1M input tokens / $80 per 1M output tokens. The release is likely very close, possibly even in the same week as GPT-5.6. Competition is heating up again. Gemini 3.5 Pro is about to face serious pressure. It better be a banger.

@birdabo @synthwavedd Why does it say openai there tho? 🤔 Definitely not an OpenAI model, or related in any way.

‼️it seems Anthropic is ready to publicly launch a new version of Mythos, something better than Mythos Preview. a codenamed model “Oceanus” was given access to some red teamers yesterday according to @synthwavedd. it’s apparently been paused already, due to someone reselling access through a Chinese API proxy lmao 💀 Mythos pricing might also end up at with $16 Input, $80 Output according to @scaling01

@appealstoheaven @cheesealloverme @michael_waples @0xdippo @thsottiaux Reading comprehension doesn’t appear to be your strong suit. This actually helps your point, take the w lol What doesn’t help your point is that in the actual terms pages that you never looked at it has legalese that disclaims all of this.

@cheesealloverme @WCNegentropy @michael_waples @0xdippo @thsottiaux Never was unlimited usage even a part of this discussion. you're crazy lol. It says unlimited picture/video gen. Cool stuff man, we're talking about codex

When X starts asking if the AI bubble is bursting The AI bubble might be bursting

@AnthropicAI You guys are VERY late to the party lol Your models have directly been used at the other frontier AI companies to help them create their own flagship models. Remember the whole thing about xAI using Claude?

@AnthropicAI You do realize that people have been building their own AI models from scratch with your agents for well over a year now, right? And with GPT before that. I’m not talking “help me post train this model”, I’m talking “let’s design a completely novel AI architecture & full train”

Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor. It’s happening faster than we thought, and the implications deserve greater attention. anthropic.com/institute/recu…

Mark my words Theo is going to end up directly working at one of the big AI companies. And that’s not a compliment lol

@AdamRackis @thekitze Nah the closed-source for-profit industry is at real existential risk if they don’t perfectly navigate the rising headwinds. Open Source won’t go away, but the closed AI industry? We’re about to see a lot of bankruptcies and acquisitions very soon, mark my words.

@thekitze jesus christ man agents are not going anywhere - nobody is saying otherwise. But ofc there’s a real chance the subsidized prices end and maybe we these tools a bit less for the easy shit we never needed them for to begin with

guys can u imagine how sad it's gonna be in 5 yrs that ai and llms and agents won't be a thing because of the token economy and the bubble that's gonna burst etc it's gonna be so so sad when humans will do everything manually again :( but at least we'll all be overpaid again 🙂

@thekitze Hey at least I finally began to learn coding because of AI. Not joking, it may be a slop factory for straight product but it’s an amazing way to learn if you use it right. Vibecode projects until you reach a stopping point, then learn how they work and improve them yourself!

The central problem is that "my" AI output is the result of a carefully crafted prompt pipeline that I honed over multiple decades, including all of my hard earned, tacit knowledge baked into the soul of every letter in the prompt, and "your" AI output is just pure fucking slop.

goosebumps rose on my arms. you want me to do this, without ai? is that right? "haha yes, jimmy. just an LRU cache. you can use an editor but no ai or autocomplete." there's no fucking way. there's only 30 minutes left in this onsite. my heart starts racing. cold sweat hits hard. im this close to an offer and now they want to see me fucking code? without ai? its over. no its not. i can still do this. ok, the editor is open. i just need to make a class to start. i cant remember anything. not even what its supposed to look like. what does python even look like? "how about you just get something down first and we'll go from there" how about you go fuck yourself. asking me to write code in 2026. the indignity. well here's how i would prompt an ai - "no i need to see code. please write the code." i can feel my heartbeat in my eardrums. my toes curl in common project sneakers. my whoop buzzes. strain goal hit. can you excuse me a second? im need to use the restroom quick. i take my phone out and call a waymo.

@abhi_agarwal4 I agree but it’s also like walking into a circus and asking people not to be clowns, so, eh 🤷🏻‍♂️ Canned LLM responses just tell me who not to deal with. But then again, everyone you need to deal with in the business seems to be going for AI for everything, so it’s a catch 22 🥲

Can’t believe I have to request people to cut AI slop from actual conversations now Suddenly felt like Claude found a way to reach me on WhatsApp

Blaming individuals or groups of users for the platforms own choices, WHETHER SUSTAINABLE OR NOT, is not the way. Especially not blaming “broke” users when you’re an unapologetic tokenmaxxer who literally did a $40k Copilot tokenmaxxing stream, fking hypocrite (yes it’s Theo)

I’ve never seen more AI bros crashing out before than I have over the last couple days. Like c’mon, I use AI too and think it’s a genuinely amazing tool. But that doesn’t mean I can’t call out the terrible corporate culture that’s taken it over and caused the issues.