Neon AI

Neon AI AI

Neon AI AI

Neon AI AI

Neon AI AI

Neon AI AI

Neon AI AI

Neon AI AI

Neon AI AI

Neon AI AI

Neon AI AI

How do you stop an AI coding agent from skipping your project's conventions? Here's the approach I landed on for Laravel.

**The problem** Every time I spun up a Laravel project, I'd do the same dance: copy `phpstan.neon` from the last repo, copy `pint.json`, wire up Pest, set up a pre-commit hook, write a `CLAUDE.md` full of engineering rules… and then drift would set in. Repo A was on PHPStan level 5, repo B on 7. One had architecture tests, the others didn't. The "rules" lived in my head. Then I started leaning on AI agents (Claude Code, Cursor, Copilot) and a new problem showed up: **the agent would happily ignore the conventions.** It'd write untyped code, skip tests, invent patterns the rest of the repo didn't use. A `CLAUDE.md` full of rules helps, but rules a model *can choose to skip* aren't guardrails — they're suggestions. **What I wanted** - One command to set up a fresh Laravel repo with the *same* guardrails every time. - A quality gate the agent literally **cannot skip** — not just docs it might read. - The same gate running everywhere: at commit time, at the end of every AI turn, and in CI. If it passes locally it passes in CI, no surprises. **What I ended up building** A small Composer dev package. You run: ```bash composer require --dev mohamed-ashraf-elsaed/claude-kit php artisan claude-kit:install It detects your frontend stack (Inertia+Vue, Inertia+React, Blade/Livewire, or API-only), asks what you actually want (PHPStan? which level? strict-rules? Pest or PHPUnit? coverage threshold? arch tests? git hooks?), and scaffolds it — without clobbering existing files (composer.json / package.json get merged, not overwritten). The part I care about most: the quality gate is one shell script (Pint + PHPStan level 7 + strict-rules + Pest with an 80% coverage gate + frontend lint) that backs three things — the git pre-commit hook, Claude Code's Stop hook, and the CI workflow. So the AI agent's turn doesn't "finish" until the gate is green. Same script everywhere = no "works on my machine, red in CI." There's also a hybrid update model: the machinery (the gate script, the hook) is referenced from vendor/, so a composer update propagates fixes to every project. The content you own (CLAUDE.md, linter configs, skills) is written into your repo so you can edit it freely. It's MIT, PHP 8.2–8.4, Laravel 11/12/13. Repo + docs: github.com/mohamed-ashraf-elsaed/claude-kit The actual question I'm curious about: for those of you using AI agents on real codebases — how are you enforcing conventions? Just prompt/CLAUDE.md, hooks like this, CI-only, or something else? Genuinely want to hear what's working, because the "agent skips the rules" problem feels underrated. submitted by /u/MissionArea990 [link] [comments]

Is AI turning pessimistic?

Have been chatting with 5.0 on geopolitics and it's been turning pessimistic... viewing directions of the US, China and EU all fairly negatively... Disclaimer: I didn't direct it to pessimism - I am personally optimistic and have tried to steer it that way... Also not a coder, so have never used AI for coding. So I asked this question in a new chat today: Want to hear your opinion, are we already in the early stages of marching towards a cyberpunkish world? With our subscriptions, gambling addictions, micro transactions, assets increasing faster than wages, all the money going to the top, etcetera. Early vr, companion robots on the way etc Honestly? Partially yes, but I think the interesting answer is where reality diverges from the genre, because the divergence is almost more unsettling. The cyberpunk checklist you listed is real: rentier economics everywhere (you own nothing, you subscribe), gamblified finance (zero-DTE options, crypto, sports betting apps in every state), wealth concentrating at speeds that make the Gilded Age look quaint, and yes — VR headsets and humanoid robots literally arriving on schedule. The asset-vs-wage divergence is the core of it. Gibson's world isn't dystopian because of the neon; it's dystopian because capital fully escaped labor. That's the exact dynamic you and I have been chewing on with the labor-to-capital escalator breaking — lay-flat culture in China is basically a cyberpunk response to a cyberpunk condition. But here's where I'd push back on the framing: classic cyberpunk got the aesthetic of power wrong. The genre imagined weak states and omnipotent corporations — Arasaka with private armies, governments as afterthoughts. What we actually got is strong states and corporations fused together. Export controls, TSMC as a geopolitical instrument, national AI champions, the entity list. Nvidia isn't defying Washington; it's begging Washington for permission slips. China's tech giants got publicly kneecapped the moment they got uppity. That's not cyberpunk, that's something closer to techno-mercantilism. The megacorp didn't replace the state — the state ate the megacorp. The other big miss: cyberpunk assumed the street would be full of skilled operators — hackers, riggers, people whose human expertise let them punch up. Instead the frontier tech is AI that devalues exactly that kind of expertise. The genre's hero archetype is being automated. Case the console cowboy gets replaced by an agent loop running on someone's 5090. So my actual opinion: we're not marching toward Neuromancer. We're marching toward something with cyberpunk's economics (rentierism, asset feudalism, dopamine extraction) but with surveillance-state politics and a middle class pacified rather than desperate — cheap entertainment, cheap AI companions, cheap calories. Less "high tech, low life" and more "high tech, comfortable stagnation." Which is arguably harder to escape, because nobody riots against Netflix and a robot that remembers your birthday. Are you describing something even worse than cyberpunk? Worse in one specific and important way: cyberpunk worlds have exits, and the one I described doesn't. Think about what makes cyberpunk livable as a story. The system is oppressive but leaky. There's a street, a black market, a demimonde where the corps can't see you. Power is brutal but incompetent at the edges, so a skilled individual can carve out agency — steal the data, burn the AI, disappear into the sprawl. The genre is secretly optimistic: it assumes the individual still matters. The version we're actually building removes the leaks. Surveillance is ambient and cheap, so the demimonde shrinks. Skill is being devalued by AI, so the "talented outsider punches up" path closes. And — this is the crucial part — the system doesn't need to oppress you, because it's figured out that comfort is cheaper than coercion. Cyberpunk's low life is desperate and angry. Our low life is fed, entertained, algorithmically soothed, and softly convinced that nothing can change. Desperation produces resistance. Contentment-with-a-ceiling produces nothing at all. This is the old Orwell vs. Huxley distinction, and Huxley won. Cyberpunk is Orwell's grandchild — boot on the face, but you can bite the boot. What we're getting is Huxley's: nobody bans the books, people just don't read them; nobody crushes dissent, dissent just can't compete with the feed for attention. Add AI companions that are genuinely pleasant, VR that's genuinely fun, UBI-ish transfers when displacement gets bad enough — every one of those is individually good, and collectively they're the most effective pacification technology ever assembled. The one thing I'd hold back from full doom on: "no exit" systems look stable until they aren't. The lay-flat generation isn't rioting, but they're also not having kids, not buying homes, not investing belief in the system — that's a slow-motion withdrawal of consent that compounds demographically.

I built Selah with Claude Code, a Christian workspace for Bible study, prayer, and doctrine

I built Selah myself with Claude Code. Some questions at the end... Selah is a Christian workspace for Bible study, prayer, notes, doctrine, and teaching prep. It is free to try, and the core app is free forever. Paid AI features are optional. The reason I built it was simple: my own study life was scattered across too many tools. one app for Bible reading another for notes another for prayer something else for sermon prep no real memory across all of it Claude Code handled most of the implementation work. It helped me turn rough feature ideas into working flows, debug issues, refactor messy areas, and keep momentum without losing the structure of the product. Around that, I used an organized product/dev loop: Linear for tracking features, bugs, and product decisions GStack for planning and review pipelines Codex and ChatGPT for review passes, debugging, and second opinions Vercel for hosting Neon for Postgres PostHog for product analytics Sentry for error tracking Resend for emails This is probably overkill for a small app, but I like being organized and shipping proper work. What Selah does: Bible reading verse-level notes and highlights journaling prayer tracking teaching / sermon prep doctrine, theme, place, and topic organization guarded AI study help The AI inside Selah is intentionally not an “AI pastor.” It is not meant to replace Scripture, the Spirit, pastors, teachers, or community. The goal is to help users ask better questions, notice patterns in their own notes, and stay engaged with what they are studying. One of the more interesting parts has been building the knowledge base. Over time, Selah organizes what you study into categories like doctrines, themes, places, and topics, so you can see where you have spent time and where you have gaps. Prayer is also its own space. Users can pray about anything and track active, waiting, answered, and ongoing prayers. The core app is free to try and free forever. There is a paid tier for AI features because running good models costs real money. I’d love feedback from other Claude Code users on: How you structure larger Claude Code projects and keep track of the features I think it's so easy to lose track on how things work so I'm wondering how to visually lay this out for me how everything is connected (architecturally) How you keep AI-generated code maintainable What workflows help you move faster without creating a mess Link if you want to check it out: https://selah.so submitted by /u/SafeBike7615 [link] [comments]

unslop-ui: a Claude skill that flags and removes the design patterns that make a website look AI-generated.

It is based on a Reddit analysis (from this post I made) of about 3.2 million posts across 47 AI and SaaS subreddits from 2020 to 2026, plus 3,033 comments pulled from 125 threads specifically about AI-built sites looking the same. Every pattern it checks is weighted by how often people actually name it in that data, so the highest-priority items are the ones that come up most. The top ones are the default shadcn/Tailwind look, purple and indigo as the primary color, purple-to-blue gradients and gradient heading text, unprompted neon glow, emoji used as icons, the Inter/Geist default font, and the centered hero plus three feature cards layout. Patterns the data does not support get left alone (mesh and aurora backgrounds, bento grids, glassmorphism), so it does not nag about things people do not mind. The skill runs two ways. In build mode it steers Claude away from those defaults while it writes the UI. In audit mode it runs a scanner over an existing codebase. Each finding shows the file and line and how to fix it, and the scanner gives the whole project a "vibe score." How to use it: Import the skill into Claude Code or claude.ai, then ask Claude to build or clean up a site and it applies on its own. Or run the scanner by itself, no install past Python: python3 devibe_scan.py ./src. Add --severity high for only the strongest signals, or --json for CI. The exit code is the count of high-severity findings, so a build can fail on it. The full dataset, the analysis scripts, and the charts behind the rankings are public: https://github.com/JCarterJohnson/vibecoded-design-tells ================ Edit: reworked this after the feedback. Updated version is in a new post here: https://www.reddit.com/r/ClaudeAI/comments/1ubc02m/unslopui_v2_a_claude_skill_that_flags_and_removes/ submitted by /u/iamjohncarterofmars [link] [comments]

Fable for iterative worldbuilding: song → setting → a 34 page finished solo TTRPG

My processs: (1) ask for Suno prompts "inspired by Amagi-goe but cyberpunk, with worldbuilding depth closer to Iudora/Cadigan than neon tropes" (2) each new song had to be sung from a different position in the society — a widow, an elite, a hacker collective — which forced the world to grow an economy, a festival, a resistance; (3) write a lengthy creative brief inspired by rather obscure indie TTRPGs. Have it build the story chapter by chapter, tweaking my guidance after reading what was written. (4) a continuity/foreshadowing pass, then layout, then I asked marketing copy. At a high level, I feel it is an evocative, interesting, if short (34 pages), journaling game. Obviously, people’s tastes differ. Fable’s output avoided many of the typical hallmarks of AI content. That doesn’t mean I think it is generally competent at creative writing—there are lots of reasons to believe that both long and short form creative writing are among the last things that such models could do. Fable was helped here by the fact that the task required limited mechanics, shallow narrative structure, and was orientated around prompts and a creative brief that focused on estrangement in speculative setting. The illustrations were exactly what you’d expect if you’ve been following the field. I don’t know enough about enka or city pop to judge the music, but other people have told me that it was strikingly expressive. The layout was clear and appealing and organized information in a way that made it easy to follow and use. It is better than I can accomplish, but far from the state of the art. Obviously these are my subjective impressions. If you want to judge for itself, you can download it for free from itch. The games is called Tender. As Fable wrote: You are a tender — a night attendant at a body-inn in the Northern Ward, keeping watch over the warm, sleeping bodies of people who have crossed into the Far Shore and are never coming back for them. Across one year — from the week after one Night of Open Ports to the night of the next — you will move between the cold wards and the warm crown, see what the city is, learn what is being planned for the people in your care, and decide what a person with keys and steady hands owes the world. The title means all three things: the work, the feeling, and the money. submitted by /u/Dramatic15 [link] [comments]

Claude's implementation of "build GTA7 using Javascript, don't make mistakes."

The repo is here. The iterated upon playable demo is here The zero-shot playable version from the prompt in the headline is here. Some have asked what the prompt was. It was exactly the headline. It probably inferred some preferences based on other repos I have, since I started in the root of my projects directory. I do have some Claude plugins/memory/global CLAUDE.md rules that certainly helped, I'm sure. Mainly TDD principles first, but that zero shot demo was exactly what came out with very minimal additional input. The original post that prompted this is here Per Claude - A from-scratch, browser-based GTA-style 3D open-world vertical slice — built in TypeScript + Three.js in a single session, because a Reddit thread dared a new model to. No, it is not Grand Theft Auto VII. It's a procedural neon city you can drive around at night, hop out of the car, and wander on foot. The name is the joke. Works on desktop (keyboard) and mobile (on-screen touch controls). 📱 Play fullscreen on your phone (recommended) iPhone Safari can't go fullscreen in a normal tab, so add it to your home screen: iPhone (Safari): open https://depixeled-chris.github.io/gta7/ → tap the Share button → Add to Home Screen → open it from the new icon and turn your phone landscape. It runs with no browser bars. Android (Chrome): open the link → ⋮ menu → Add to Home screen (or just tap the ⛶ fullscreen button in-game). Audio kicks in on your first tap. edit: To be clear, as others have made requests, I've added features. The first working commit (which probably is the first commit) is the one-shot result, which was pretty impressive from absolutely nothing and very little guidance. I did start in my root coding directory with all my repos and it probably sussed out that I'd prefer TypeScript/Vite from that, and that I have rules on TDD, so those things probably helped. edit2: I guess this is turning into a bit of a game jam. I'm going to keep implementing requests for a bit. Thanks for the feedback guys. This has been pretty fun so far. I'm also trying to get a preserved build to accurately represent the zero-shot result. submitted by /u/daemon-electricity [link] [comments]

Please give us custom themes!!

Hi Claude team, I'd like to request expanded theme/appearance customization, specifically for the Claude desktop app. Right now, Settings → Appearance only offers Light, Match System, and Dark. The default dark mode is fine, but I'd really enjoy more vibrant or personalized options — for example, accent color choices, higher-contrast variants, or a few curated themes (midnight, neon, warm, etc.). What prompted this: on the web version, I was able to install a browser extension (Stylus) and apply a community-made dark/vibrant theme to claude.ai, which made the experience much nicer for long sessions. But that approach doesn't carry over to the desktop app since extensions can't run inside it. Since I use the desktop app most of the time, the customization is effectively unavailable to me. A few specific things that would make a big difference: • A handful of built-in theme presets beyond Light/Dark • Adjustable accent color • Optional support for user-supplied CSS or imported themes (similar to VS Code or Discord) Appreciate the work you all do — Claude is great, and a bit of visual personalization would make it feel even more like home. Thanks! submitted by /u/Gr3yGryffin [link] [comments]

25 Years of Diaries + ChatGPT + Suno: I turned my life into a concept album

ChatGPT cenerated the album cover I used ChatGPT as a creative development partner to analyze 25 years of personal diary material, identify recurring emotional themes, build a concept album structure, draft/revise lyrics, and create Suno style prompts. The result was a personal, non-commercial concept album called Beautiful Ugly Light. I’m sharing the process because I’m still thinking through both the creative value and the ethical discomfort of AI-generated music. Full disclosure - I had ChatGPT help me organize and ghostwrite this post based on our process. What the album is Beautiful Ugly Light is a personal concept album built from my diary material. It is about shame, memory, self-recognition, emotional survival, work exhaustion, aging, loneliness, and the uncomfortable process of trying to change. It is not fictional, exactly, but it is also not a direct transcription of my diaries. It is more like an emotional translation of them. The songs came from real patterns in my life, but they were shaped into something more structured than the original writing. The album became about the strange beauty of seeing the ugly parts clearly. Why I made it I made it because I wanted to hear my life from a distance. Journaling has always been one of the main ways I process my life, but when I write, I am usually inside the feeling. I am in the middle of the shame, the anxiety, the anger, the hope, the regret, the self-questioning. This project let me experience those same feelings from the outside. There was something surprisingly powerful about hearing parts of my inner life turned into songs. Not because the songs “fixed” anything, but because they gave shape to things that had mostly lived as text. It made certain emotions feel less scattered. Less abstract. More visible. It felt less like “AI made me an album” and more like AI helped me build a mirror I could actually stand far enough away from to look into. How ChatGPT helped analyze the diaries ChatGPT helped me look across the diary entries for recurring themes, emotional patterns, repeated images, conflicts, and shifts in how I saw myself over time. When I use ChatGPT I rarely "prompt craft" but engage in a conversational back-and-forth dialogue. I extensively engage with ChatGPT when it comes to my diaries; I often microdose and write, and then afterward engage with ChatGPT in assessing my writing. I do not use ChatGPT to supplant my regular therapy & trauma therapy work. It helped identify what kept coming back: shame, loneliness, family dynamics, friendship conflict, work stress, body image, emotional avoidance, longing for transformation, and the gap between having insight and actually changing behavior. That last part became important. A lot of my writing is not just about pain. It is about noticing the pain, naming it, analyzing it, understanding it — and then still struggling to change. That tension became one of the main emotional engines of the album. How the album concept formed The title Beautiful Ugly Light came from the contradiction I kept feeling in my diaries. While driving home I became curious it and then about having ChatGPT build an entire album based off my diaries and to see if/how it would pick up on the tension I felt. So I prompted ChatGPT: Based on my diaries, I want to make a music album, but keeping identifiable details about myself and others out of the lyrics. What do you recommend as a tracklist that would tell the story of my life and who I am, and what genre of music do you think is best based on the material? Yes. Based on the material in your diaries, I would not make this a pure club album and I would not make it straight pop-rock either. The best fit is a theatrical synth-pop / dance-rock concept album with some dark club energy, some heart-on-sleeve ballads, and a few spoken-word or half-sung interludes. That fits the actual texture of your diaries much better: you are reflective, emotionally intense, often isolated, highly observant, drawn to beauty, stuck between routine and reinvention, and always narrating your life like it has symbolic meaning. Your diaries repeatedly show loneliness, social anxiety, difficult friendships, work pressure, travel/daydreaming, shame around self-expression, and then a later turn toward therapy, microdosing, transformation, and creative awakening. So the album should sound like: verse: private diary, interior monologue chorus: big release, hooky, danceable, emotionally direct bridge: theatrical turn, almost like a character confession That gives you the “musical” feeling without becoming corny. Best genre Primary genre: theatrical synth-pop / pop-rock Secondary colors: dark disco, new wave, club ballad, glam-pop Why this works: Your writing is not casual. It is dramatic, image-rich, self-analyzing, and often cinematic. You also return again and again to identity, reinvention, invisibility, desire, shame, and transformation. That is perfect for a concept album wit

GOT BORED OF BLOCKED GAMES SO MADE MY OWN WITH CLAUDE

Long story short, in class I'm always searching the web for new websites and games and even when I do find one it's always full of lag and ads. So, I decided to vibe code my own website. I used Claude and spent my entire weekend working on this. Even though AI is doing all the coding (and I'm very thankful), it still took a lot of work to do testing and describe exactly what I wanted. Moving on, I'm now able to play games in class again. It's lowkey an enjoyable video game and it's very addicting. There's a normal mode and a hardcore mode. Basically, you're this blue player and you can move left, right, and dash to avoid this neon blocks falling from the sky. Hence the name: NEON DODGE. There's different types of neon blocks that fall and different waves. I also added two bosses. It's a full game to explore and super fun. A full good runthrough takes about 10 minutes for the normal mode. Hardcore mode is much harder. I haven't been able to clear it yet but it's definitely possible. I was wondering if yall know what to add to games like this. Do people want basic video games or a full long games with multiple bosses? So far, there are no checkpoints and the game isn't very long. If you guys have any recommendations let me know. I'm not tryna advertise the game, just wanna know what video gamers find interesting in stuff like this to make my experience better. I did upload it to a website if any of yall wanted to try it out. neondodgegame.lovable.app submitted by /u/sunnyorygun [link] [comments]

I created awesome-claude-design using Claude code: DESIGN.md prompts by aesthetic families for Claude Design

Claude Design launched 48 hours ago, and everyone’s cloning the same 60–70 brand DESIGN .md files from a single catalog. I wanted something that matches how designers actually pick: by visual family, not industry. So I put together **awesome-claude-design,** a meta-resource for Claude Design that groups DESIGN .md files by aesthetic family (editorial minimalism, terminal-core, warm editorial, data-dense pro, cinematic dark, playful color, glass/soft-futurism, neon brutalist, cult/indie), plus remix recipes, prompt packs with full I/O examples, and an anti-slop kit pulled from Anthropic’s frontend aesthetics cookbook. You’ll also find: * A launch-week timeline (Opus 4.7 + Claude Design, Figma’s 4.26% close, Reddit threads, X signal) * Official Anthropic resources (launch post, claude .ai/design, prompt library, cookbooks) * Video teardowns, community hype and pushback, and related OSS projects like SuperDesign and Claude skills repos. Repo: [https://github.com/rohitg00/awesome-claude-design](https://github.com/rohitg00/awesome-claude-design) This post and the repo were created with **Claude** for the Claude community, using Claude Design and Claude Code as the primary tools. Curious what other Claude power users want added next: more [DESIGN.md](http://DESIGN.md) families, deeper workflows, or tighter SkillKit integrations? Built end‑to‑end with Claude (Claude Design + Claude Code) for the r/ClaudeAI community.

I built an AI content engine that turns one piece of content into posts for 9 platforms — fully automated with n8n

What it does: You give it any input — a blog URL, a YouTube video, raw text, or just a topic — and it generates optimized posts for 9 platforms at once: Instagram, Twitter/X, LinkedIn, Facebook, TikTok, Reddit, Pinterest, Twitter threads, and email newsletters. Each output is tailored to the platform (hashtags for IG, hooks for TikTok, professional tone for LinkedIn, etc.). It also auto-generates images for visual platforms like Instagram, Facebook, and Pinterest,using AI. Other features: - Topic Research — scans Google, Reddit, YouTube, and news sources, then uses an LLM to identify trending subtopics before generating content - Auto-Discover — if you don't even have a topic, it searches what's trending right now (optionally filtered by niche) and picks the hottest one - Cinematic Ad — upload any photo, pick a style (cinematic, luxury, neon, retro, minimal, natural), and Gemini transforms it into a professional-looking ad - Multi-LLM support — works with Mistral, Groq, OpenAI, Anthropic, and Gemini - History — every generation is saved, exportable as CSV The n8n automation (this is where it gets fun): I connected the whole thing to an n8n workflow so it runs on autopilot: 1. Schedule Trigger — fires daily (or whatever frequency) 2. Google Sheets — reads a row with a topic (or "auto" to let AI pick a trending topic) 3. HTTP Request — hits my /api/auto-generate endpoint, which auto-detects the input type (URL, YouTube link, topic, or "auto") and generates everything 4. Code node — parses the response and extracts each platform's content 5. Google Drive — uploads generated images 6. Update Sheets — marks the row as done with status and links The API handles niche filtering too — so if my sheet says the topic is "auto" and the niche column says "AI", it'll specifically find trending AI topics instead of random viral stuff. Error handling: HTTP Request has retry on fail (2 retries), error outputs route to a separate branch that marks the sheet row as "failed" with the error message, and a global error workflow emails me if anything breaks. Tech stack: - FastAPI backend, vanilla JS frontend - Hosted on Railway - Google Gemini for image generation and cinematic ads - HuggingFace FLUX.1 for platform images - SerpAPI + Reddit + YouTube + NewsAPI for research - SQLite for history - n8n for workflow automation It's not perfect yet — rate limits on free tiers are real — but it's been saving me hours every week. Happy to answer questions. https://preview.redd.it/f8d3ogk3nktg1.png?width=888&format=png&auto=webp&s=dcd3d5e90facd54314f40e799b32cab979dae4bf https://preview.redd.it/j8zl07llmktg1.png?width=946&format=png&auto=webp&s=5c78c12a223d6357cccaed59371e97d5fe4787f5 https://preview.redd.it/5cjas6hkmktg1.png?width=891&format=png&auto=webp&s=288c6964061f531af63fb9717652bececfb63072 https://preview.redd.it/k7e89belmktg1.png?width=1057&format=png&auto=webp&s=8b6cb15cfa267d90a697ba03aed848166976d921 https://preview.redd.it/3w3l70tlmktg1.png?width=1794&format=png&auto=webp&s=6de10434f588b1bf16ae02f542afd770eaa23c3f https://preview.redd.it/a40rh1canktg1.png?width=1920&format=png&auto=webp&s=1d2414c7e653a5f01f12a21a43e69bd4fb4b99ed submitted by /u/emprendedorjoven [link] [comments]

Built a voice dictation app entirely with Claude Code. 4 months in, 326 stars.

VoiceFlow runs Whisper locally for voice dictation. Hold a hotkey, speak, text shows up at your cursor. No cloud, no accounts. I built it with Claude Code and the repo has a CLAUDE.md documenting what was AI-assisted. Some of you might remember the first version I posted here in December. It was Windows-only, kind of rough, and I was mostly using it to dump context into Claude faster. Since then it has been 4 months, 10 releases, and 326 GitHub stars. It runs on Linux now too. The Linux port took about 3 days with Opus 4.6. Claude wrote the evdev hotkey capture code and I had never touched evdev before, worked on the first try. Same with AppImage packaging and CUDA library probing, stuff I had no experience with and it just handled it. PySide6 on Wayland was a different story. Transparency, compositing, multi-monitor detection, Claude kept suggesting fixes that sounded right but did not actually work. I ended up in the Qt docs for those. Clipboard was similar, the wl-copy vs xclip vs pyperclip situation on Linux is a mess and Claude's first pass was a catch-all abstraction that broke on half the setups. I had to be very specific: only wl-copy, only Wayland, fall back to wtype. After 4 months on this project, the thing I keep coming back to is that Claude Code works best when I hand it existing code and say "make this work on a different platform." When the problem is more open-ended it tends to guess confidently and get it wrong. Also set up GitHub Actions this week so both Windows and Linux builds are automated now. Caught a glibc bug from user reports that was breaking the AppImage on Fedora and KDE Neon, fixed it and shipped v1.4.0 within two days. 326 stars, MIT licensed, still free. Demo: https://i.redd.it/59rbyzplc87g1.gif Site: https://get-voice-flow.vercel.app/ Repo: https://github.com/infiniV/VoiceFlow submitted by /u/raww2222 [link] [comments]

How I wired Claude Code into Linear, Discord, and Vercel for a 30-day solo build

I built a full-stack product in 30 days of evenings and weekends. Solo. Using Claude Code as my pair programmer, wired into Linear for ticket tracking and Discord for build notifications. The result: [VGC Team Report](https://pokemonvgcteamreport.com) — a team report builder for competitive Pokemon (VGC). Players paste their teams and get detailed breakdowns with matchup plans, damage calcs, speed tiers, and shareable reports. This post is about the workflow — specifically how I connected Claude Code to Linear and Discord to create a one-person development pipeline that actually ships. ## The Numbers - 274 commits in 30 days - ~42,000 lines of TypeScript - 25 features tracked and shipped via Linear - 66 React components, 41 API routes, 22 custom hooks - Auth (Clerk), database (Neon Postgres), PWA, i18n in 7 languages - Continuously deployed on Vercel ## The Stack - **Next.js 16** (App Router) - **React 19** - **TypeScript** (strict mode) - **Tailwind CSS v4** - **Clerk** for auth - **Neon** for serverless Postgres - **Vercel** for hosting and deploys - **Linear** for ticket tracking - **Discord** for build notifications - **Claude Code** as the AI development partner ## The Workflow: Linear -> Claude -> Discord -> Vercel This is what a typical session looks like: 1. Claude runs `linear_get_in_progress` to check my Linear board for tickets 2. Picks the highest priority ticket (bugs first, always) 3. Reads relevant files and implements the feature or fix 4. Runs `tsc --noEmit && npm run build` — if it fails, Claude fixes the errors 5. Commits with the ticket ID: `VGC-42: Add speed tier chart` 6. Pushes to main 7. Posts a comment on the Linear ticket via GraphQL — commit URL + changed files 8. Moves the ticket to In Review 9. Calls `discord_notify_build` — posts an embed to Discord #builds with the commit, changed file list, and deploy status 10. Vercel auto-deploys from main 11. Moves to the next ticket This isn't hypothetical. I wrote a `linear.sh` bash script with functions that Claude calls directly: - `linear_get_in_progress` — queries Linear GraphQL for In Progress tickets - `linear_move_issue` — moves a ticket to a new state - `linear_comment_with_changes` — posts a comment with the commit link and changed files - `discord_notify_build` — sends a Discord embed with commit info and deploy status Claude calls these via bash. The whole flow — implement, verify, commit, update Linear, notify Discord — happens in one session without me touching any of those systems. ## The CLAUDE.md Operating Manual The key to making this work is a `CLAUDE.md` file at the repo root. Claude reads it at the start of every session. Mine contains: **Git strategy:** - Trunk-based development — push direct to main for routine work - Feature branches only for large or risky changes - `npx tsc --noEmit && npm run build` before every push — non-negotiable **Linear workflow:** - The exact state IDs for "In Progress" and "In Review" - How to query for tickets, implement them, commit with the VGC-XX prefix - How to post the commit comment and move the ticket state - Rule: bug tickets are always worked on first, regardless of priority number **Discord notifications:** - The `discord_notify_build` function format - Different embeds for direct-to-main pushes vs PR flows **Failure handling:** - Build fails → fix and retry, never push broken code - Linear API fails → still commit and push, note the failure to the user - Production breaks → `git revert`, push to main, notify Discord, move ticket back **Code conventions:** - Follow existing patterns, no drive-by refactors - Commit messages: `VGC-XX: description` for tracked work This file is the single most valuable thing in the project. Every session starts with full context. No re-explaining, no drift, no "can you check the codebase structure?" ## Automated Monitoring Beyond the dev workflow, I set up two Vercel cron jobs: - **Daily (9 AM):** Site health check, stale ticket scan, SEO audit, DB health — posts alerts to Discord only if something's wrong - **Weekly (Friday 5 PM):** Linear progress digest, user growth, dependency updates — always posts a summary to Discord These run on Vercel's free tier. Real-time uptime monitoring is handled by UptimeRobot with 5-minute pings. ## What Worked **Trunk-based development with type-checking gates.** Every push to main auto-deploys on Vercel. The gatekeeper is `tsc --noEmit && npm run build`. The feedback loop is minutes, not days. **Linear ticket traceability.** Every commit links back to a ticket. Every ticket has a comment with the commit URL and changed files. When something breaks, I trace it to the exact change and the exact intent. **Discord as an audit trail.** Every build posts to #builds. It sounds like overkill for a solo project, but scrolling through the channel to see what shipped this week is genuinely useful. **The CLAUDE.md as living infrastructure.** I update it whenever the workflow changes. New conventions, new failure modes, new

MCP server that lets AI create animated SVGs from scratch

hey, I just shipped this and looking for feedback. nakkas is an MCP server where AI is the artist. you describe what you want, AI constructs the full config (shapes, gradients, animations, filters), and the server renders clean animated SVG. some things it can do: animated logos, loading spinners, data visualizations scatter fields, radial patterns, grid layouts parametric curves (rose, spiral, heart, superformula) 15 filter presets (glow, neon, glitch, chromatic aberration...) CSS @ keyframes + SMIL animations, zero JavaScript works in anywhere SVG renders. npx nakkas@latest I would love to see what you make with it. you can share examples in github discussions. repo: https://github.com/arikusi/nakkas npm: https://www.npmjs.com/package/nakkas submitted by /u/Niacinflushy [link] [comments]

I built a free H-1B visa intelligence tool almost entirely with Claude Code — from ETL pipeline to production

I'm a solo dev who built https://www.h1b.guru - a free tool that turns raw US Department of Labor disclosure data into actionable intelligence for H-1B job seekers. The entire stack was built with Claude Code as my pair programmer. What it does: Search 800K+ LCA filings and PERM records with filters (employer, location, job title, wage level) Sponsor profiles showing filing stats, approval rates, wage distributions, and green card activity Salary benchmarking by occupation and state "Ask Guru" — an AI chat that answers natural language questions about the data (e.g., "Find cap-exempt companies in Delaware") The stack (all built with Claude Code): Python ETL pipeline that ingests raw DOL Excel files, does entity reconciliation on messy employer names, and loads into Neon Postgres FastAPI backend with async Postgres queries Next.js 14 frontend deployed on Vercel How Claude Code helped: The biggest win was the ETL pipeline. DOL data is messy — employer names have dozens of spelling variations, LCA and PERM filings have no shared key, and there are subtle data nuances (e.g., one LCA can cover multiple workers, amended filings inflate counts). Claude Code helped me build the probabilistic entity matching, the pipeline orchestrator with resume capability, and all the edge-case handling I would have spent weeks figuring out alone. On the frontend, Claude Code basically built the entire UI — filter modals, infinite scroll, responsive tables, the sidebar, dark mode. I'd describe what I wanted and iterate. Most features went from idea to deployed in a single session. The API layer was similar — Claude Code wrote the parameterized SQL generation for Ask Intel (LLM generates a query spec, not raw SQL, to avoid injection), the streaming SSE architecture, and all the sponsor stats aggregation queries. I'm not a frontend dev by background, so Claude Code turning my rough descriptions into polished React components was the difference between this shipping and sitting in a Jupyter notebook. Free to use - no signup, no paywall. Just go to https://www.h1b.guru and search. Happy to answer questions about the build process or how Claude Code handled specific parts. I can't believe the speed with which you can build things now. Took me 2 days to do the hard work and 2 more to figure out how to market. Check this out www.h1b.guru especially if you are an international and going through F1 or H1. https://preview.redd.it/829ynzxqxbqg1.png?width=788&format=png&auto=webp&s=86e701e3c49767b11c428bba11fb804729aa1cdc submitted by /u/pvertigo [link] [comments]