@Lalith

I built a free, privacy-first MVP called StanceLoop: a live, browser-based AI coach that watches you train via webcam and gives real-time form feedback, rep counts, and spoken cues—100% locally on your device.The Motivation
Most fitness tech either requires expensive hardware or uploads your private camera feed to external servers. I wanted to solve a simple problem: Could a browser and an on-device pose model act as an honest, real-time coach without sending a single frame to a server? StanceLoop was built to prove that private, accurate form feedback can run frame-by-frame on any standard browser.What It Does: Zero Backend, 100% Private
StanceLoop turns your webcam into an active coaching tool with no accounts, no server uploads, and no tracking:Core Drills: Detects and counts push-up reps, times handstand holds, and tracks extension-and-return for jab-cross shadowboxing combinations.Real-Time Guidance: Provides spoken cues, skeleton overlays, and haptic feedback during your sets.Honest Scoring: If your full body isn’t in frame or camera alignment is lost, the app explicitly tells you instead of guessing or faking reps.Async Ghost Duels: Share a simple URL containing drill rules to challenge friends—no central server or database required.Local AI Coach: Optionally connects to a local LLM server (like Ollama running on localhost) for personalized coaching advice.How We Built ItFrontend: React 19 and TypeScript, bundled with Vite 6 and deployed as a static site to GitHub Pages.Computer Vision: MediaPipe Pose Landmarker (@mediapipe/tasks-vision) with custom landmark smoothing, confidence gating, and rule-based state machines.Testing: 340+ unit tests with Vitest and 11 end-to-end tests with Playwright running in CI.Challenges & LearningsThe Subpath Blank Screen: The initial deploy showed a blank page due to absolute asset paths resolving incorrectly on GitHub Pages. Fixed by configuring relative base path handling in Vite.Jitter vs. Latency: Raw camera pose points jitter constantly. Balancing instant audio feedback with noisy pose data required landmark smoothing and cue cooldowns so feedback stays useful instead of chaotic.Honesty Over Everything: CV coaching state machines are hard because pose models love to guess. Engineering the app to recognize when it can’t see you properly was key to making the feedback feel real.What We’re Proud OfA complete, real-time computer vision app running offline-capable in the browser with zero backend costs.A robust testing suite (350+ combined tests) ensuring the state machine rules stay reliable across updates.Built-in versioned consent and privacy disclosures directly in the UI.Try it out:Live App: https://hustlenix.github.io/stance-loop/Source Code: https://github.com/Hustlenix/stance-loopTech Stack: React 19, TypeScript, MediaPipe Pose, Vite 6, Vitest, Playwright, GitHub Pages.

I built a free MVP called Level Up: a 28-chapter self-development book built from a 20-hour video transcript, running entirely in your browser.The Motivation
It started with a simple question: What if you could turn a 20-hour self-dev video into something you could actually read, search, and verify? Most online self-development content exists to sell courses, not to present evidence. I wanted something honest. So, I built a real book where every notable claim is graded against cited research.What It Is: Zero Tracking, 100% Local
The result is roughly 32k words across four pillars, a 30-claim evidence audit (graded A–D/U), 13 testable protocols, and full-text search.The best part? Zero tracking. No accounts, no APIs, no cookies. Your reading progress, streaks, highlights, themes, and quiz scores live entirely in your browser’s localStorage. Nothing leaves your machine.How We Built It
The original video transcript was ~20k lines. We tag-teamed Claude, Gemini, and Opencode to compress it. From there, we built the site in layers:The Content Pipeline: scripts/build-data.mjs compiles Markdown chapters into SQLite using node:sqlite, then emits JSON data and a 30KB MiniSearch index.The Frontend: Next.js 16 static export and Tailwind v4. We used useSyncExternalStore for reactive local state via pure TypeScript modules.The Deploy: Automated via GitHub Actions straight to GitHub Pages.Into the Trenches (Challenges & Learnings)The CI Spiral: We added a React 19 QuoteCard. It built flawlessly locally (22/22 tests green), but failed eight consecutive GitHub Actions runs with a phantom TS error. It wasn’t a code bug; it was an environment disagreement. Lesson: Sometimes a feature just isn’t ready. We deleted it rather than fighting the environment or shipping it broken.Turbopack’s Quiet Types: The dev server uses Turbopack, which skips full type-checking for speed. We shipped a PR that built locally but failed in CI. Lesson: Always run tsc locally, or CI will aggressively remind you of the difference between “builds” and “types.“Case Sensitivity: GitHub Pages serves paths case-sensitively. NEXT_PUBLIC_BASE_PATH=/LevelUp is fundamentally different from /levelup. We learned this the hard way.What I’m Proud OfHonest Grading: No invented citations. If a famous “5x flow” stat is folklore, the site tells you so.Backend-less Magic: Search, quizzes, highlights, streaks, and schema-validated JSON backup/restore—all completely offline-capable.The Architecture: The data layer isn’t just a pile of JSON; it’s a statically served, compiled artifact from Markdown.Try it out:Live site: https://hustlenix.github.io/LevelUp/Source code: https://github.com/Hustlenix/LevelUpTech Stack: Next.js 16, Tailwind v4, node:sqlite, MiniSearch, GitHub Pages

Guys, I’ve been researching free resources to build an AI form coach for calisthenics and combat sports. Here is the ultimate breakdown for the computer vision pipeline:MediaPipe vs. YOLOv8-PoseGoogle MediaPipe (Free, Apache 2.0): Best for Calisthenics (Push-ups, Planches). Excellent on-device mobile speed. Outputs 33 3D body landmarks. Struggles slightly with heavy motion blur.Ultralytics YOLOv8-Pose (Free, AGPL-3.0): Best for Combat Sports (Striking, fast sparring). Needs more compute (better on server/GPU). Highly robust against fast, blurry movements. Outputs 17 2D keypoints.Top Open-Source Repositories to Fork
Instead of starting from scratch, use these as your foundation:yakupzengin/fitness-trainer-pose-estimation (MIT License): One of the best starting points. Uses MediaPipe for 18 exercises with a YAML-based system. You don’t hardcode math—just define ideal joint angles in YAML, and it scores the user 0-100 based on angle and tempo.The FormCoach Pipeline (arXiv:2508.07501): Uses Vision-Language Models (VLMs) to understand form and give contextual feedback (e.g., “Keep your back straight”) instead of just flashing red when an angle is wrong.Movelytics/posetracker-llm-prompts: Perfect for boxing or Muay Thai. Extracts keypoints and compares the user’s camera feed against a reference video of a pro fighter to generate a similarity score.Muqaram0/Excercise_Pose_Correction: A complete pipeline combining YOLOv8 and MediaPipe. Has pre-built logic for analyzing back alignment and posture during push-ups and squats, plus a free Streamlit web UI.imanoop7/AI-Agents-as-Personal-Trainers: A multi-agent AI system (LangChain) that generates tailored routines. Run it locally via Ollama to keep the AI coaching completely free on your own hardware.How to Stack Them for DevelopmentFrontend/UI: Fork Muqaram0 and run the Streamlit app for an instant, free user interface.Vision Backend: Use MediaPipe scripts to track calisthenics angles via webcam. Implement the YAML config system from yakupzengin so you aren’t hardcoding angles.Coaching Backend: When the vision script detects bad form (e.g., “hips dropping”), pass that text flag into the local Ollama LLM setup from the imanoop7 repo to generate human-sounding advice.The Problem:
I’ve stacked all this, but it is still kinda mid and is not working properly! How do I fix it? And everything I try to test it out, it is like a whole workout just debugging it! 😭😭😭😭😭😭

POV: Me trying out it and it absolutely wont work but i finish my workout😉😉.

I am on a mission to build a Workout app with CV that it can recognise what we are doing, and will guide us in live time

Upgraded this opensource AI coding Agent (Opencode) for free, that even Iron man is jealous of me now🤣. Devlog one. First i thoght that changing opencode is immposible but after few days of tweaking, learning, and coding i somewhat made it good looking i mean integrated it wil 4 mcp’s where one edits my videos, one researches and i am looking forward in adding more. and added like 200 skills from nvidia, and added a agent with the help of opencode it self, and added some @’s to the game, more like a role for it. and i am looking froward to make it even better than any coding agent out there and give it to you guys. STAY HARD GNG!!!!!!!!!!!!!!!!!!!!!

So called Devlog, 3

The agenda today is to simplify the home page. I paid attention to it for 8 hours and ended up with two fewer things on screen than when I started lol😅✌🏻.

After removing the large hero box and changing the chapter list to an accordion, the homepage appeared empty. I created a big wishlists of items to add: heatmaps, quote-generators, streak-trackers, quiz-cards, XP-counters, etc. Everything was built local-first on Next.js 16 (static export on GitHub Pages) using localStorage keys like levelup-streak-v1 no tracking or backend APIs.

I constructed a QuoteCard component to retrieve data from /data/quotes.json. All 22 tests passed on my machine, but for some reason, GitHub Actions fails a phantom TS7006 type error 8+ times in a row because of that one CI error. I kept spamming commit messages like “force CI re-run”, “force CI re-re-re-run, and this time ts-ignore…”. It turned out to be a weird CI environment mismatch that ate half my day.

I discarded the quote card to avoid clutter. In what was shipped was a simple PrivacyBadge (“100% Private · On-Device Storage Only”) along with a ThemeTimeShift script that alters the –hue CSS variable based off your clock.

Next thing we must do is transition smoothly into the chapter! Stay techulated or smtg. STAY HARD GNG!

I am on my way to build a AI agent by myself from scratech, its been 5 hours since i’ve starteed. wish me luck

This is the starting of my new project. Everything is a copy of a copy of a copy of a copy of a copy.

Upgraded this opensource AI coding Agent (Opencode) for free, that even Iron man is jealous of me now🤣. Devlog one. First i thoght that changing opencode is immposible but after few days of tweaking, learning, and coding i somewhat made it good looking i mean integrated it wil 4 mcp’s where one edits my videos, one researches and i am looking forward in adding more. and added like 200 skills from nvidia, and added a agent with the help of opencode it self, and added some @’s to the game, more like a role for it. and i am looking froward to make it even better than any coding agent out there and give it to you guys. STAY HARD GNG!!!!!!!!!!!!!!!!!!!!!

I am currently on a mission to make my website https://hustlenix.github.io/The-Rime-Of-The-Ancient-Mariner/ more fun, interactive, and non boring website both in mobile and desktop. So tely’ll be more interested in studying and learning. through my website.

Here is a ultra-crisp, casual version designed to fit right into your Hack Club devlog style:Devlog #5 — Teacher Tools & Parchment Restyle (~7h logged)Turned the portal into a full teacher workspace—built a question bank, a test paper generator, and a print library, then polished accessibility, bundle size, tests, and CSS!What got built:Teacher Backend (db.js, teacher.js): Added schemas for questions and paper templates. Generating a paper snapshots the questions so editing the bank later won’t ruin past test history.Smart Autofill (autofill.js): Builds papers based on difficulty (1–5), question types, and chapters while tracking usage frequency so questions don’t over-repeat.Paper Builder & CSV Import: 3-step wizard with preset CBSE formats (like 29-mark setups) and bulk CSV question uploading.A11y & Performance: Lazy-loaded all 14 routes to cut bundle size from 221 KB to 186 KB. Fixed screen-reader toggle text and added offline PWA caching.Parchment Theme: Redid CSS with warm ivory (#f0ece2), clean cards, serif fonts (Source Serif 4), and dark mode tokens.Autofill Logic Snippet:JavaScriptconst candidates = pool
.filter((q) => !excluded.has(Number(q.id)) && !picked.has(Number(q.id)))
.filter((q) => matchesSection(q, spec, user))
.map((q, i) => ({ q, i, tiebreak: rng() }))
.sort((a, b) => a.q.usage_count - b.q.usage_count || a.tiebreak - b.tiebreak)
.map((x) => x.q);
What I Learned:A11y live regions: Always announce the result of a toggle to screen readers, while setting the button label to the next action.GitHub Pages hosting: Used static fallback data for student views so everything works smoothly without a live Express server.Next up: Direct PDF exports for the print queue! I am gonna finish this in no time and get my money, right.

Finished writing the git hub repo, and used AI to organise the words cause it was full of gramatical mistakes and stuff. I am ready to ship this project GNG, what did I miss, please check this out and give some valid review please.

This is my trust worthy, non tracked website for leveling up in your life but somehow works without any login or stuff, the tracking just works, inspired by this 20 hour video: https://www.youtube.com/watch?v=wvaY5bG5p7A&t=65s which is totally worth it. To build this I started by getting the transcript for making this website, and with that 20k lines transcript I pasted it into claude to compact this, and it ran out of tokens, and i tried gemini too sadly it too ran out of tokens, and lastly I tried Opencode few time and it finally gave me the summary for this and after few hours i finished the website and pushed the code to git hub, branched it real good, and finally build and in actions started a build and deploy workflow and got this running. STAY HARD GNG!!!!!!!!

Ig that my free study website for stressed 10th graders from India to Indians is coming to an end, now that I have entered all of the data for the website like all poems, lessone and stuff I think with some touch ups and some polish I am gonna ship this in no time. W GNG

Finally, IG that this project is ready to ship as it is working just fine. But, the problem is that Nest is still pending so i can’t get it to 24/7, and IDK why but the app isnt working so the slash commands are still broken on Slack

I tried a Neo-Brutalism style thing for the got, but it is not going well😭✌🏻. Someone pleaase help me make a study website from scratch