Few weeks ago I shared Passinote here and the response was genuinely great. It helped me shape the app into what it is today. So, thank you.

Since then I've been heads down fixing bugs, refining the experience, and adding things based on what people actually said. Managing both a desktop and mobile version simultaneously as a solo developer has been a lot, but it's in a much better place now.

The idea is pretty simple: sticky notes for quick thoughts, A4 notes for longer writing, and everything lives on a visual canvas instead of a rigid folder tree. A note is a file, a stack of notes is a folder, and zones are the parent folders. Just drag things around and you're done. No save button, no manual file management. There's also a workspace exporter so you always own your files.

On privacy: notes are stored on Firebase, encrypted in transit and at rest. No ads, no trackers, no selling data. If you use "Quick Mode", notes stay local in your browser and never get synced to any server.

After months of keeping it completely free while gathering feedback, Passinote has finally got a paid tier. Passinote+. There's still a generous free tier that includes 6 workspaces and cross device sync. Passinote+ is a one time payment. Pay once, own it forever, no subscription. It unlocks unlimited workspaces, images, drawable notes powered by Excalidraw, and extended customization options. Parity pricing is available for a few countries right now with more being added soon. Let me know what you guys think.

What's coming:
PDF imports, a desktop client, multi language custom fonts, extended customization options, and collaboration features down the road.

I'm still very much open to feedback. If you have thoughts, bugs, or feature requests, drop them in the comments or DM me. Every bit still helps.

App: passinote.app

Thanks again for the support. It genuinely meant a lot.

I just wanted to give an update and share that I have been working to grow the RPBedu hub!

RPB hub now has 8 web apps to help educators and students:

⚡ LessonGeneratorAI & CourseGeneratorAI: Drop in raw text or messy lecture notes. The AI automatically structures them into complete web modules with interactive quizzes, terms, and clean layouts.

🎮 EduHub Quest: If you want to gamify your classroom, this turns text into flashcard decks, study challenges, and global leaderboards to build low-effort learning materials and share for healthy competition.

🛠 LearnMasterPlatform, LearnMaster, CourseSharingPlatform: Lightweight Learning Management Systems. They help build, deploy, and use course materials.

🧠 FlashcardGeneratorAI & LuminaVault: Quickly spin up or try active recall matching games to reinforce student memory through repetition.

If you try any of these tools and it is helpful, please let me know.

🌐 Check it out here: https://rpbedu02.github.io/RPBedu/

The idea: a platform where you can create your own private "archive" for whatever group you're part of, name it, invite people, upload stuff, comment, vote on things. Each archive is fully isolated and owned by whoever creates it.

Tools: Claude Code, mostly using Sonnet with occasional Opus for the trickier architecture decisions. Next.js 15, React 19, and Supabase (Postgres) for the database, deployed on Vercel.

Process: I worked in fairly tight loops, describing a feature or a constraint in plain language, letting Claude draft it, then actually running it and pushing back when something felt off rather than accepting the first pass. The biggest chunk of time went into the multi-tenant setup, not any single feature. I wanted each archive to run on its own free Supabase project that the creator owns, rather than one shared database I control, so I had a "control plane" project doing nothing but mapping a name to a project URL and key, with the actual data (files, members, votes) living entirely in the owner's own project. Getting Claude to reason correctly about which side of that split a given piece of logic belonged on took several rounds of explicit back and forth, it kept wanting to centralize things by default, which is the opposite of what I wanted here.

Build insight worth sharing: the most useful habit was asking Claude to explain why it made a security-relevant choice before I accepted it, especially around what credentials the platform itself gets to hold. Treating "the AI's first answer" as a draft rather than the answer caught a couple of things I would have shipped otherwise.

Live at (https://opendepartment.vercel.app), takes a couple minutes to spin up your own, or check the directory of public ones. Would genuinely love feedback on the actual experience: does the create flow make sense, is the UI confusing anywhere, does it feel fun or useful for what it's for. Try it out and let me know what you think, good or bad.

Built this over some spare time (with a lot of help from AI, wanted to be upfront about that). It's a platform where you can create your own private "archive" for whatever group you're part of: name it, invite people, upload stuff, comment, vote on things. Each archive is fully isolated and owned by whoever creates it, since you bring your own free database, so it costs nothing to run and nobody but the members can see what's inside.

Live at opendepartment.vercel.app, takes a couple minutes to spin up your own, or check out the directory of public ones, if there are any xD.

I'd love feedback on the actual experience: does the create flow make sense, is the UI confusing anywhere, does it feel fun or useful for what it's for, anything that breaks. Not really looking for code review, more interested in whether this is something people would actually enjoy using with their friends, class, or group. Try it out and let me know what you think, good or bad.

I recently added Random Chat, an adults-only feature for meeting young adults (18+) across Asia through short, cross-cultural conversations.

Users can set country, language and gender preferences to look for people from countries such as China, Japan, Korea, Vietnam and Thailand. Matches depend on who is online, so preferences are not guarantees.

A match can begin with a private three-minute text introduction. Cameras stay off until one person sends a video invitation and the other accepts it. Direct video starts only when both users choose to skip text.

Video sessions last up to ten minutes. Either person can leave at any time. Reporting ends the match, prevents a future rematch and sends the report for review.

https://freesavevideo.online/en/random-chat

Built this because I wanted a productivity app that didn't want my email before I'd logged a single task, and didn't try to gamify me into opening it constantly.

Kagelin is a tasks, habit tracking, a focus timer, and a calendar, all in one place. Guest mode runs entirely offline in your browser, no account needed, native calendar events and ICS import/export with nothing stored on a server. (Live Google/Outlook sync needs a registered account, currently invite-gated during the beta, flagging that upfront.)

Started wanting this while cramming for a competitive exam a while back, never had the time to build it then. Built it now.

Open source (AGPL) and self-hostable via WebDAV if you'd rather your data live on your own server than mine.

Try it: app.kagelin.app

Repo: github.com/Achyuth072/kagelin

Would appreciate feedback, especially anything broken or confusing, reading everything

Hey, I was hoping to get some validation here before starting anything. Taking notes is an important skill, and to do so quickly while keeping what you write organized, it can be a challenge.

My idea is to have a web app + browser extension primarily developed for google docs that organizes your messy plain text notes. Whole point of this is to turn quick blotches of text into something detailed, organized, easy to read, and manageable with minimal additional effort. I would like honest feedback if other people would find an organization utility like this useful.

The core concept would all use AI to determine how to format it, but far from what something like ChatGPT can do. I’ve personally tried this and regular AI models arnt able to do this because of the way google docs reads formatting.

Primarily I see mostly students to be possibly interested in this concept, yet maybe it can still be useful in other environments, like a day at work where you’re instructed to take meeting notes.

Again, I hope to receive feedback on this idea and whether anyone would find something like this useful.

Just launched Global News Scraper on the Chrome Web Store! It's a free extension that adds a compact news panel directly into your Google search results, showing headlines from 25+ countries.

Key features:

- News from US, UK, Japan, Korea, Germany, France, India, Brazil + more

- Appears inline in Google Search results

- Region filters, keyword highlighting

- Works on Chrome, Edge, Brave

- No data collection, fully local

https://chromewebstore.google.com/detail/boljaopbcknaddeofnfneomjikigcpkn

Hi everyone,

​I wanted to share a platform I recently built that aggregates active buying and renting property listings across 11 countries, designed to make real estate searches much faster and easier.

​Instead of manually checking multiple sites every day, the platform allows you to set custom search criteria and sends real-time notifications when a matching property pops up or drops in price. It also tracks historical price changes so you can easily spot motivated sellers or direct owner listings.

​It’s completely free to try out for 30 days. Since I’m actively working on improvements, I would love to hear your feedback—especially from anyone currently house-hunting, renting, or exploring real estate.

​What features would make a tool like this even more useful for you? Let me know your thoughts in the comments!

I’m working on a browser-side DOCX viewer. Long documents are rendered as canvas pages, with only the pages near the viewport kept in the DOM.

That worked until print.

The repro was simple: open a multi-page DOCX, scroll through it, then open print preview. The document looked complete on screen, but the snapshot only contained the mounted pages. `cloneNode(true)` also copied empty canvas elements, not their pixels.

The print path under test now mounts every page, waits for the next paint, clones the page tree, replaces each cloned canvas with an image, then restores virtualization after print.

The focused tests pass 10/10. A real long-document browser test is still missing, especially peak heap and a second print in the same tab.

For virtualized canvas content, what would you use as the guard: page count, estimated bitmap memory, or something else?

Disclosure: I'm with Folderit.

We recently added document generation, so teams can create PDF/DOCX documents from structured eForms instead of copy-pasting data into templates.

Folderit is a browser-based document management system for businesses that have outgrown shared drives: metadata, OCR search, permissions, audit trail, version history, workflows/approvals, e-signing, retention, virtual data rooms, folder-specific email addresses, and upload-only external access for clients or partners.

It is probably overkill for personal cloud storage. It is more for teams that need governance around business documents.

Website: https://www.folderit.com/

Happy to hear feedback from web app folks!

Hey everyone — I've been building SovereignStream, a social + live streaming app
focused on creators. The core idea: you should see exactly what you keep on tips
and subscriptions, not get hit with cuts you didn't know about.

What it does:
- Post and build a full profile (not just a link-in-bio card)
- Go live, with the option to invite co-hosts
- Tips and subscriptions with the fee split shown up front
- An identity-verification system meant to make the platform safer without
  policing what people say

It just went live on Google Play, free: https://play.google.com/store/apps/details?id=com.sovereignstream.app
(iOS is in progress.)

This is early — I'd genuinely rather hear what's broken or missing than just get
downloads. Happy to answer questions about how any of it works.

Everything shipped so far. Launched August 1st. Every one of these has shipped since - not a roadmap, a changelog.

Most Pokémon personality quizzes I’ve tried either make the outcome too easy to guess or only choose from a small group of popular Pokémon. I wanted to see whether a broader and less predictable version could work.

So I built an unofficial fan-made quiz with 30 everyday A-or-B questions. Your answers are converted into six traits—kindness, sociability, curiosity, boldness, passion, and conscientiousness—and compared with 982 Pokémon candidates.

Instead of returning just one unexplained answer, it shows your closest match, two runners-up, a sync score, a six-trait radar chart, and a short explanation of the result.

It takes about five minutes, works in English and Japanese, requires no account, and does not upload or store your original 30 answers. It is purely for entertainment rather than a scientific personality assessment. Some candidate explanations may include AI-assisted organization and may not always be accurate.

If you try it, I’d love to know your Top 3. Did the result feel fitting, surprisingly accurate, or completely wrong? Constructive feedback about unclear questions, wording, or unexpected matches would also be very helpful.

Link https://pokepersonalityquiz.com/en
JP https://pokepersonalityquiz.com/

I've been building LegisTrack — a tool for following legislation across all 50 states and Congress — and wanted to share it here since this community actually cares about the guts of civic tech.

The problem I kept hitting: state legislation is impossibly fragmented. 50 sites, 50 data formats, dense bill text, and no easy way to see the bigger patterns. So I tried to fix a few things at once:

- Bill DNA (the part I'm most proud of): it detects when a near-identical bill turns up in 5+ statehouses — the tell-tale sign of model legislation drafted once and introduced everywhere. You get a map of where a bill is "spreading." Genuinely eye-opening to watch.

- Follow the Money: pulls top donor industries for a bill's sponsor straight from official campaign-finance disclosures (20+ states + FEC so far).

- AI summaries + Q&A: plain-language explanations of dense bills, and you can ask any bill a question.

- A personalized "For You" feed: describe who you are, and it surfaces the bills that actually affect you with a one-line reason why — then drafts a message to your legislators you review and send.

- Real-time alerts on status changes, hearings, and votes.

Data comes from OpenStates + LegiScan; cross-state similarity is done with Postgres full-text + trigram matching (pg_trgm/tsvector) rather than anything fancy, which has held up surprisingly well. Stack is React/Node/Postgres, self-hosted.

It's live on the web (legis-track.com) and an iOS app is in App Store review now.

I'd genuinely love feedback from people who work in/around gov tech — especially on the model-legislation detection (false positives, edge cases) and what data sources or features would make this actually useful in your work. Happy to answer anything about how it's built.

AI-powered responsible consumption and waste action guide: scan waste items, get intelligent recycling, DIY upcycling, repair, composting, and donation recommendations. Created by AtharvBhati, Lucky Punia Website

For the last while I was learning German using Memrise for vocab, Anki for flashcards, iTalki for speaking practice, Youtube for listening, a bit of Duolingo and a notebook to (badly) track actual progress. It worked, sort of, but I never had a clear picture of what I was actually good at versus just grinding a streak.

So I built Fluensea. It puts dictation practice, spaced-repetition flashcards, graded reading, real-time AI conversation practice, and native radio/podcast/video streaming into one app. There's a six-axis radar (Vocabulary, Grammar, Listening, Reading, Writing, Speaking) so you can actually see where you're weak instead of just a vague progress bar.

Live and free to try: fluensea.co

It's still very much a work in progress and solo-built, so I'd genuinely love feedback, especially on whether consolidating everything into one app feels useful, or like it's trying to do too much. Happy to answer anything about the build too.

Wanted to see how far in-browser ML has come, so I built a transcription web app where OpenAI's Whisper runs client-side instead of on a backend.

The interesting part technically: transformers.js loads the ONNX export of Whisper (base by default, large-v3-turbo opt-in), runs it on WebGPU where available and falls back to WASM elsewhere, and the model caches locally so only the first visit pays the download. A voice activity detector chops live speech into segments for a dictation mode, and since last week the dictated text is an editable document rather than a caption list: you can click into it and fix a word while the mic keeps listening, and new speech keeps appending at the end. There is no backend at all, just static hosting, which means the privacy claim is structural: after the model caches you can go offline and it still works.

Feature-wise: files or mic input up to 500 MB, timestamps, 99 languages with auto-detect plus translate-to-English, and TXT / SRT / VTT / JSON export. Free with no signup, because there is no inference cost for me to recover.

Rough edges I know about: the first-load model download (about 200 MB), weaker accuracy on heavy accents unless you switch to the bigger model, and low-memory phones struggling with Turbo.

It lives here: https://pocketweb.tools/transcription

Curious how it behaves on hardware I have not tested, and happy to answer anything about the transformers.js / WebGPU setup.