Show HN за 11 августа 2025 г.

It requires no registration, no installation, and no configuration. Just run it on any computer and start drawing freely.

Lots of fun to do. I would have not taken the time without the speedup provided by Claude.

https://andyrosa.github.io/Sinclaude/simulator.html

I built a live video privacy filter that helps smart glasses app developers handle privacy automatically.

How it works: You can replace a raw camera feed with the filtered stream in your app. The filter processes a live video stream, applies privacy protections, and outputs a privacy-compliant stream in real time. You can use this processed stream for AI apps, social apps, or anything else.

Features: Currently, the filter blurs all faces except those who have given consent. Consent can be granted verbally by saying something like "I consent to be captured" to the camera. I'll be adding more features, such as detecting and redacting other private information, speech anonymization, and automatic video shut-off in certain locations or situations.

Why I built it: While developing an always-on AI assistant/memory for glasses, I realized privacy concerns would be a critical problem, for both bystanders and the wearer. Addressing this involves complex issues like GDPR, CCPA, data deletion requests, and consent management, so I built this privacy layer first for myself and other developers.

Reference app: There's a sample app (./examples/rewind/) that uses the filter. The demo video is in the README, please check it out! The app shows the current camera stream and past recordings, both privacy-protected, and will include AI features using the recordings.

Tech: Runs offline on a laptop. Built with FFmpeg (stream decode/encode), OpenCV (face recognition/blurring), Faster Whisper (voice transcription), and Phi-3.1 Mini (LLM for transcription analysis).

I'd love feedback and ideas for tackling the privacy challenges in wearable camera apps!

Hi HN,

I've built an app for macOS that allows performing common SSH operations on Linux servers using a native GUI.

The problem:

Managing multiple Linux servers usually means juggling terminal windows and copy-pasting snippets/scripts. After dealing with tens of production/staging VPSes at previous jobs, I realized there had to be a better way for common operations I did on a daily basis than my collection of bash snippets.

Features:

- Quickly switch between different servers. Tag servers with arbitrary key values for easy search.

- Real-time dashboard with CPU/memory graphs, disk usage, and uptime.

- Table based interface for processes (sortable/filterable), Docker containers, systemd services, network ports, and system logs etc.

- Built-in file browser.

- Full-featured terminal when you need to drop to the command line.

You can check out the screenshots at https://serverbuddy.app/screenshots for a quick overview of the features supported.

All the above are done through SSH, there are no agents/scripts to install on your servers.

From using the app for a few weeks(admittedly a short duration), I can say I much prefer the ServerBuddy based workflow to my previous workflows.

Pricing:

Free forever for one server, $59 one-time for unlimited servers (includes 1 year of updates).

If you're a developer or sysadmin managing Linux servers from Mac, please do try out the app. I'd love your feedback regarding additional features/workflows etc.

Thank you!

Hi everyone — I’m Clark, creator of Keeps, a way to send physical postcards that play your voice with a quick QR code scan.

WHAT IT IS Upload a photo (or pick from Unsplash), write a message, optionally add a 60 second voice note, and send a 4×6 postcard. QR code printed on the postcard links to a “digital card” that plays the audio. $5 flat, printing + postage included. No account required.

WHY I BUILT IT I love mailing postcards to my family but hate it's a hassle (finding card, buying stamp, handwriting, praying it arrives on time). I wanted to merge the charm of physical mail, the warmth of voice messages, and the efficiency of software.

HOW IT WORKS Built with Next.js, Supabase, Stripe, Resend, and Lob. QR codes link to hosted audio. Stripe handles payment. Lob prints and mails the cards. Resend emails status updates.

WHAT'S DIFFERENT - Voice note + physical card combo - Minimal editor, no sign-up friction - Transparent $5 price (everything included)

TRY IT FREE 1. Go to https://sendkeeps.com 2. Create your card 3. Use promo code SENDKEEPSFREE for a free send (first 50 people)

FEEDBACK WANTED - Editor UX - QR/digital card flow - Address autocomplete UX - Critical missing features - Interesting use cases (wedding thank yous, airbnb hosts, etc)

Happy to share details or answer questions!

I often found myself doing this tedious routine:

1. Download a file 2. Open Finder and locate it 3. Drag it into the browser/app to upload

For things like tax paperwork or testing uploads, I was repeating this dozens of times a day.

QuickShelf is a small macOS menu bar app that lets you instantly open a specific folder in a compact window and drag files directly — without opening Finder.

It’s lightweight, quick, and stays out of the way.

- https://github.com/Slowhand0309/QuickShelf - https://quickshelf-app.slowlab.dev

Right now it's a Nicotine patch for brain rot/doom scrolling. You still get the scroll, but with more meaningful content.

WHERE I WANT IT TO GO: Users mark what they are interested in (e.g. AI, or Space or Whatever), and it should crawl the latest scientific articles, high quality YouTube videos, pods, present it in a format that fits you, in a nice feed.

NOW: You can search for anything (e.g. "Sea Explorers" --> The app crawls Wikipedia or YouTube and pulls together a short feed (e.g. Columbus, Vasco da Gama etc.,). I'm using RAG so that I always can reference the original data source (feel that is key)

Regardless of source, I'm pulling images from Wikipedia. Right now, this is the biggest struggle, as the quality and number of images per article is limited --> my articles doesn't come out as nice.

Would really appreciate some feedback, especially if you tend to browse Wikipedia a lot

I built an Excel add-in that connects to Ollama, so you can run local LLMs like Llama3 directly inside Excel. I call it XLlama.

You can use it like a regular formula: =XLlamaPrompt("Is Excel a database") or run it on an entire range.

No API calls. No cloud. No subscriptions. Everything runs locally.

It’s useful for quick tasks like extracting names, emails, or phone numbers from text, or for doing light data analysis without leaving Excel.

Would love feedback, especially from people who use Excel daily.

Hey HN,

I’d like to share an open-source project I’ve created. It is an email archiving tool that can back up your whole email system with full-text search capabilities.

The project is called Open Archiver, and it is able to archive and index emails from cloud-based email inboxes, including Google Workspace, Microsoft 365, and any IMAP-enabled email inbox. You can connect it to your email provider, and it copies every single incoming and outgoing email into a secure archive that you control (Your local storage or S3-compatible storage).

Some features:

- Archive and index all emails and attachments

- Import and sync the whole organization's emails: For Google Workspace and MS 365, Open Archiver can ingest all individual inboxes' emails

- PST and .eml files import

- Role-based access

- Full-text search: All archived emails and attachments are indexed, so you can search the content of all emails and attachments

- You can choose to store your files either on your local machine or on any S3-compatible storage provider

- API access

It's AGPL-3.0 licensed and free to use for personal and business purposes. I'd be happy if you could give it a try and give me some feedback.

You can find the project on GitHub (with Demo site): https://github.com/LogicLabs-OU/OpenArchiver

I recently built an iOS app called Settld: Group Restaurant Finder that helps friends decide where to meet by finding restaurants that are roughly equally far from everyone’s location, and displaying information about them. We’ve all been in chaotic group chats where no one can agree on where to eat — this app cuts through that by calculating a “sweet spot” for the group. For 2 people, it’s the midpoint. For 3 people, it’s the circumcenter. For 4–6 people, it uses a minimum enclosing circle approach (Welzl’s algorithm).

It then shows the top 15 nearby options so there’s no more “where do we meet?” chaos — or $50 dinners after a gruelling 2-hour trip just because no one planned. If anyone’s wondering why I capped it at 15 options, it’s to cut down on decision paralysis.

Would love to get your thoughts: https://settld.space/

I am happy to announce the next major version of pywebview, a lightweight Python framework for building modern desktop applications with web technologies. The new version introduces powerful state management, network event handling, and significant improvements to Android support. See https://pywebview.flowrl.com/blog/pywebview6.html for details.

KARMA-OpenMedEvalKit is an expandable toolkit for assessing AI models in medical applications, featuring multiple healthcare-focused datasets with particular emphasis on the Indian healthcare environment.

KARMA can evaluate text, image, and audio-based medical AI models using 21+ healthcare datasets We support popular models (Qwen, MedGemma, IndicConformer, OpenAI, Anthropic models - via AWS Bedrock, and practically any HuggingFace models) out-of-the-box KARMA also handles medical-specific evaluation needs like ASR models that need language-aware post-processing, or having LLM as a judge on rubric based evaluations. KARMA caches model outputs so you can iterate on metrics without re-running expensive inference.

Medical AI evaluation is currently fragmented – researchers often build custom evaluation scripts for each project. KARMA provides standardized metrics and a registry system where you can easily plug in your own models and datasets.

KARMA has extensible registry system with decorators for easy model/dataset integration. It supports custom metrics with dataset-specific post-processing. The model's output are cached based on the datapoint and the model configuration to speed up evaluation iterations.

The Indian healthcare focus came from our work focused on building AI systems for India. Most medical AI benchmarks are heavily skewed toward Western contexts, missing important regional variations in medical terminology, disease prevalence, and clinical practices.

To aid in this, we are also releasing 4 datasets - Medical ASR Evaluation Dataset, Medical Records Parsing Evaluation Dataset, Structured Clinical Note Generation Dataset, Eka Medical Summarisation Dataset. Find the collection here - https://huggingface.co/collections/ekacare/ekacare-medical-p...

Along with our datasets, we are also releasing 2 models from our Parrotlet series in the public domain licensed under MIT. Parrotlet-a-en-5b: A purpose-built model for automatic speech recognition for medical context for English and Parrotlet-v-lite-4b: A purpose-built model for medical report understanding. Link - https://huggingface.co/collections/ekacare/ekacare-public-he...

We've been using KARMA internally and thought the community might find it useful. Happy to answer questions about the architecture or specific use cases!

GitHub: https://github.com/eka-care/KARMA-OpenMedEvalKit

Docs: https://karma.eka.care

Release blog: https://info.eka.care/services/introducing-karma-openmedeval...

I've been experimenting with LLMs + Python/Jupyter + Agent Frameworks to see if an AI "team" can handle the kind of exploratory business data analysis you would normally give to a junior analyst.

It's not for "give me the top 10 campaigns by CTR."

It's for: - Predict next quarter’s sales from 2 years of data - Identify profit drivers in Amazon sales and model how to increase net profit by 10% - Build custom customer cohorts from sales + activity data

How it works (high-level / happy to share technical learnings/notes based on comments): - Planner agent interprets the business question, breaks it into steps, interpret python execution results - Coding agent runs pandas / matplotlib / other libraries inside an isolated Docker container running Jupyter - Report agent creates formatted PDF reports/narratives

Early learnings - With a good prompt, it can do much of what a 1–2 year analyst could; - Industry-specific system prompts (context) improves outcome

Here is a screencast of how an exploration looks like: https://youtu.be/oMkgj1e1cxM

This is the read-only version of the exploration: https://app.askprisma.ai/share/public_0kjM3bN3_UAVyfG_3s3kaQ

I would love feedback on: - Which industries would find daily, sticky use? - Whether you would prefer cloud, self-hosted, or dedicated instance? - How you should price something like this? - How soon should I focus on a specific vertical or ecosystem (eg: shopify, or a specific industry)

P.S: I’m open to collaborations, especially with folks who have deep industry expertise and believe a system like AskPrisma will serve the industry well. Email: akshay [at] askprisma.ai

We run multiple voice AI agents on LiveKit and kept running into the same headaches: * No way to see TTFT (time to first token) per call * Hard to measure cost per call * No visibility into latency across STT → LLM → TTS * Debugging was basically guesswork

So we built Whispey to solve it.

What it does: * Cost analytics * Latency breakdowns (VAD → STT → LLM → TTS) * Agent evaluation metrics * Campaign tracking * Downloadable recordings + transcripts

Runs on Vercel + Supabase today, AWS-native version is coming. It’s tightly integrated with LiveKit.

https://www.whispey.xyz/ https://github.com/PYPE-AI-MAIN/whispey

If you’ve built voice agents on LiveKit, we’d love for you to try it out and share what’s been working for you. Always looking for best practices from other builders.

Plain text storage for easy syncing and versioning. Integrates with your existing workflow, not the other way around

I’ve been working on creating diagrams from JSON, YAML and similar formats for about three years. Over time it has grown into a general-purpose diagramming tool. With the recent addition of the MCP Server and ToDiagram Chat, I’m optimistic about where it’s headed.

You can use your own OpenAI key, stored locally, without needing to sign up and generate diagrams by using natural language.

With QuotationGenie you can: Create customized quotations in minutes Generate invoices and track payment status (paid, unpaid, overdue) Draft, send, and sign contracts digitally

I built this after getting frustrated with juggling multiple tools for quotes, invoicing, and contracts — now it’s all streamlined in one dashboard.

Would love your feedback on usability, features, and what you’d want to see next.

Thanks for checking it out!

I’m Victor Kuznetsov. For the last ~6 weeks I’ve been building ViktorADAM Core (VAC) — an AI system with a human-like memory stack (working, episodic, semantic, procedural, meta) used directly for reasoning. Unlike typical RAG, VAC runs entirely offline and adapts over time to the user.

Why it’s interesting

Human-like memory: consolidates context across sessions; periodic cleanup keeps long-term state sane.

Self-learning (Q-utility): prioritizes what to remember/retrieve based on usefulness.

Self-reflection: validates and corrects its own reasoning traces.

Languages: English + Russian today; CN/ES planned.

Real-world test Ran in an air-gapped setup processing 15–20 MB PDFs (policies/protocols) and answering multi-hour dialogue queries without any external APIs.

Tech notes

FAISS for ms-level semantic search

SQLite for offline storage

Async pipeline; streaming responses

Rule/logic parser for “if A and B then C” style queries

Web UI + REST API (/api/chat, /api/upload); modular LLM backend (currently Ollama)

Repo (public marketing repo; core code abstracted for privacy): https://github.com/vac-architect/vac-agi

Clip is a chrome extension that I built recently as add on to Copi.

Copi allows users to easily share secure content to their customers and track their engagement. With Clip, users no longer need to log into our platform to get the links.

They will be able to access any of their Starred Copi Links via the chrome extension, and share the content with their customers on the go~

Feel free to give it a go, and let me know your feedback. Thanks!

Ever had a startup idea stuck in your head because building landing pages, validating ideas, and creating a brand felt overwhelming or expensive? I built StarterPilot to solve that. It’s a web app that helps you take any idea and turn it into a real, visible project—today.

StarterPilot gives you instant market analysis, generates competitor breakdowns, creates unique names (with domain checks!), and even designs custom logo icons. The AI-powered landing page builder means your idea can have a home online in minutes, complete with a waitlist and contact form.

You can publish your idea to our global Idea Map, see how many people support it, and connect with others—all without spending a dime.

If you’ve got a backlog of ideas, or just want to see how fast you can go from concept to launch, give it a try: use promo code "FREESP25"

Happy to answer any questions here!

Hey folks! A few months back I built a GNU/Linux post‑installation setup program to make getting started way smoother. Since then, I’ve given it a boost with new features and refinements — and it’s now live on GitHub!

Built in Rust — fast, safe, and reliable

Works out‑of‑the‑box for: Arch, Debian, Fedora

Automates essential configs:

- iptables firewall rules

- vim setup

- zsh + Oh My Zsh customization

- zram for improved performance

- Installs either default software packages or your chosen ones

Designed to save you time and take the pain out of post‑install tweaks.

If that sounds useful, check it out and drop a star on GitHub — every bit of support keeps the project moving forward!

https://github.com/glcomtech/glwiz

I built JustGrind as a personal execution partner — an AI chatbot that adapts to your goals, keeps you accountable, and nudges you back on track when you lose momentum.

Unlike generic habit trackers, JustGrind uses context from your onboarding to personalize prompts, track your consistency, and help you push through plateaus. It’s designed for entrepreneurs, builders, and anyone chasing ambitious targets who can’t afford wasted days.

Right now it’s in early release, and I’m looking for feedback from people who want to test it and tell me what’s missing.

https://justgrinds.vercel.app/

Hey HN!

We just launched HERO 2.0 - a collaborative workspace combining the flexibility of text, the structure of databases and the power of integrations.

It's built for teams who write formal & structured documents like legal contracts, technical specs, SOPs, and policies.

This project is the result of many years acting as corporate lawyer and mega-project specialist, navigating through the maze of complex, high-stakes, structure-heavy and interconnected documents - overlapping the legal, financial & technical.

It's all about bridging the disconnect between transactional teams and operational ones.

You can try it now at [myhero.so](http://myhero.so/) and let us know what you think!

Best,

The HERO Team