Show HN for August 5, 2025

Whittle is a small word game I've been working on. Each phrase must be whittled down by one letter (or space) each turn. The remaining phrase must still consist of valid words. That's it! There's a daily puzzle, as well as an archive of old puzzles.

The idea for the game came to me in a dream (really) and I built the puzzle generator with my partner, who's also a software engineer. It's a labor of love! Any feedback or suggestions are welcome. Thanks for playing!

I bought the birthdays.app domain a few years ago and decided to build a simple birthday app, just text message reminders.

I had hacked together a duct-tape version for myself a couple years prior (using zapier + google sheets), and found it useful. It was a pain sifting through Facebook birthdays (90% irrelevant), and I found the text reminders simpler than a calendar.

The app has been slowly growing since 2023, and is now up to 739 paying users (mostly on the $9/yr plan, with a few paying $3/yr).

Earlier this year, I built a Google Calendar integration to automatically sync birthdays to the app. My goal is birthday reminders without the baggage a normal "social" app would have such as ads, engagement bait, etc.

I call it an app, but there's actually no app to download. Users log in via phone number + sms login code, then input birthdays manually, via text, or via Google Calendar.

At the moment, US users receive SMS messages (which are cheaper to send in the US), while other countries receive WhatsApp text reminders.

The app uses Twilio to text users (both SMS + WhatsApp) and Stripe for payments.

I'm hoping to build more features (iPhone contacts sync, easier birthdays importing) while also making video content that celebrates strangers' birthdays (such as, going to a farmers market with an "is it your birthday?" sign).

Thank you for reading, I'm open to any thoughts or ideas around the app!

Hey HN, we're Julian and Glenn, and we're building stagewise (https://stagewise.io), a frontend coding agent that lives inside your browser on localhost and operates on local codebases.

You can spawn the agent into locally running web apps in dev mode with `npx stagewise` from the project root. The agent lets you then click on HTML Elements in your app, enter prompts like 'increase the height here' and will implement the changes in your source code.

Before stagewise, we were building a vertical SaaS for logistics from scratch and loved using prototyping tools like v0 or lovable to get to the first version. But when switching from v0/ lovable to Cursor for local development, we felt like the frontend magic was gone. So, we decided to build stagewise to bring that same magic to local development.

The first version of stagewise just forwarded a prompt with browser context to existing IDEs and agents (Cursor, Cline, ..) and went viral on X after we open sourced it. However, the APIs of existing coding agents were very limiting, so we figured that building our own agent would unlock the full potential of stagewise.

Here's how it works: When you run `npx stagewise`, our cli proxies your running web application in dev mode and injects a toolbar containing the coding agent on top of it. Each prompt you send will be enriched with browser context and sent to our cli, which will call our backend and modify the source code of your local codebase accordingly.

Here's a demo of our agent changing the login UI of Cal.com, a popular open-source meeting scheduling app: https://www.youtube.com/watch?v=BkDcAozK9L4.

So far, we've seen great adoption from non-technical users who wanted to continue building their lovable prototype locally. We personally use the agent almost daily to make changes to our landing page and to build the UI of new features on our console (https://console.stagewise.io).

If you have an app running in dev mode, simply `cd` into the app directory and run `npx stagewise` - the agent should appear, ready to play with.

We're very excited to hear your feedback!

I originally added this to my site to speed up my video editing process. Last year I started a youtube channel and for some of my longer videos it's annoying to rely on youtube or capcut to transcribe when Whisper is open source. Capcut also recently updated their T&Cs to say they own your content if you use their app, so I cancelled my subscription.

Another use-case I have is recording my claude prompts as audio, transcribing them, and then pasting them into my terminal. I mostly work on the CLI (claude, ffmpeg, whisper), but I wanted to make a browser version.

Not reinventing the wheel here, just using Huggingface/Xenova Transformers.js to run Whisper in the browser with a background worker.

You can see a demo here: https://www.youtube.com/watch?v=ICLOvAq9vKE

If you're interested in building this yourself, there's a bunch of examples on Huggingface https://huggingface.co/collections/Xenova/transformersjs-dem...

I've been working a fair bit with DSPy lately, and I did some work in combining the benefits of vector search and LLMs (via a DSPy pipeline) to disambiguate records with a high degree of accuracy to help enrich a dataset. The blog post shows how this approach scales well, is very cost-effective and super concise - all it takes is < 100 lines of DSPy code and it all runs async.

The code to reproduce is in this repo if anyone's interested (all tools are 100% free and open source, and the methodology will work with open weight LLMs too). https://github.com/kuzudb/dspy-kuzu-demo

  Dataset Explorer is now LIVE and FREE.

-- Finding the right dataset shouldn't be this hard.

Millions of high-quality datasets exist across Kaggle, data.gov, and other platforms, but discovering the ones you actually need feels like searching for a needle in a haystack.

Whether it's seasonality trends, weather patterns, holiday data, tech layoffs, currency rates, political content, or geo information – the perfect dataset is out there, but buried under poor search functionality.

That's why we built the dataset-explorer – a completely free tool that lets you search for datasets using natural language across multiple platforms.

Just describe what you want to analyze, and it uses Perplexity, scraping (Firecrawl), and other tools behind the scenes to surface relevant datasets.

Instead of manually browsing through categories or dealing with limited search filters, you can simply ask "show me tech layoff data from the past 5 years" and get preview of multiple datasets.

Quick demo: I analyzed tech layoffs from 2020-2025 and uncovered some striking insights: 2023 was brutal – 264K layoffs (the peak year) Post-IPO companies led the cuts – responsible for 58% of all layoffs Hardware hit hardest – with Intel leading the charge January 2023 = worst month ever – 89K people lost their jobs in just 30 days

Once you find your dataset, you can analyze it completely free on Hunch . Try it yourself and let us know we can improve it for you.

Data explorer - https://hunch.dev/data-explorer Demo link - https://screen.studio/share/bLnYXAvZ

Hi HN, I’m a co‑founder of Embeddable AI.

After struggling to add interactive AI experiences to Wix, Shopify, Webflow, WordPress sites, I built this tool to let marketers build chatbots, quizzes or assistants and embed them anywhere with a snippet.

Built in React/TypeScript front end and Node.js logic engine. It loads fast and works across CMS platforms.

I’d love feedback from builders and marketers on use cases, missing features, or integration ideas.

Hi HN! I was using AI tools to research some of the RFCs people recommend you read and I thought why not build a tool that allows you to progressively do this and provides python examples, with dockerfiles to demonstrate some of the more complicated networking paradigms.

It's not fully done yet, but I thought I'd show it off to get some feedback.

https://rfc.stonecharioteer.com

Hello Folks,

I recently built "Mortlab" which is a simple, free website to help people analyze whether it's better to rent or buy a home based on their situation.

It includes:

A clean rent vs buy calculator

Monthly and long-term cost breakdowns

Refinance and affordability tools

No login, ads, or affiliate spam, it's just data

I’d love feedback from the community — especially from folks who’ve built similar tools or have strong opinions on the rent vs buy debate. I know there are a million calculators out there, but I tried to keep this one minimal and useful.

Thanks in advance.

After playing around with local AI setups for a while, I kept getting annoyed at having to juggle different llama.cpp servers for each model. Switching between them was such a pain and I always had to restart things just to load up a new model. So I ended up building something to fix that. It's called FlexLLama - https://github.com/yazon/flexllama Basically, it's a tool that lets you run multiple llama.cpp instances easily, spread across CPU and GPUs if you got'em. Everything sits behind a single OpenAI-compatible API. You can run chat models, embeddings, rerankers - all at once. The models assigned to the runners are reloaded on the fly. There's a little web dashboard to monitor and manage runners. It's super easy to get started: just pip install from the repo, or grab the Docker image for a speedy setup.

I've been using it myself with things like OpenWebUI and some VS Code extensions (Roo Code, Cline, Continue.dev), and it works flawlessly.

I show how to use Pitch shifting supported by the arduino-audio-tools[1] and ElatoAI[2] library on ESP32 to get OpenAI Realtime and Gemini Live Voices to sound like cartoons like Alvin and the Chipmunks or Hulk.

[1] https://github.com/pschatzmann/arduino-audio-tools [2] https://github.com/akdeb/ElatoAI

I originally started by sharing a collection of specialized AI subagents and slash commands for Claude Code on GitHub. Then, I enhanced it by creating a web UI, so people can browse, search, and install both subagents and commands instantly by copying and pasting. After that, I started creating a CLI tool to allow more advanced users to install and manage subagents and commands directly from their terminal. Now, I am trying to enhance the CLI tool to be a configuration tool that allows you to share, extend, and reuse the Claude Code configuration(commands, hooks, subagents and MCPs) easily for team onboarding, bulk installation, etc.

Website: https://www.buildwithclaude.com/ Github Repo: https://github.com/davepoon/claude-code-subagents-collection

Would appreciate any feedback or feature suggestions.

This burns a ton of tokens, but the basic idea is to create a cache in ./llm/cache and then spawn a bunch of subagents to look in the cache first, and then write to the cache if learn anything new

Project Page: https://rotel.dev Github: https://github.com/streamfold/rotel

Hi HN, Ray and Mike here! We’re building Rotel, a new high performance, resource efficient approach to OpenTelemetry collection. Rotel is open source (Apache License 2.0) and runs as a standalone process and collects telemetry from external processes or other collection agents. It consumes 75% less memory and 50% less CPU in benchmarks https://streamfold.github.io/rotel-otel-loadtests/benchmarks, so it is particularly well-suited for environments where resource optimization is paramount.

Mike and I love working on Observability solutions. We’ve built and optimized telemetry data planes for products such as Librato, Papertrail and AppOptics. As we’ve scaled high volume data processing systems over the years, we’ve often found ourselves under pressure to improve SaaS margins or the resilience of complex distributed systems. This has led us to the same question again and again: How can we do this more reliably and efficiently?

We believe telemetry collection demands high levels of efficiency, and Rotel is our first step toward addressing that need. Rotel’s Lambda extension https://github.com/streamfold/rotel-lambda-extension has already unlocked the promise of OpenTelemetry in production serverless environments, where resource constraints previously made this nearly impossible.

With Rotel we’re also rethinking DX for OpenTelemetry collection. Its lightweight footprint allows for packaging directly in runtimes like Python and Node.js, eliminating the need for additional configuration and deployment of sidecars. Rotel’s Python Processor SDK enables you to build telemetry processors right from your favorite IDE with the ease and expressiveness of Python and the performance of Rust.

Getting started with Rotel is easy using prebuilt Docker containers, just run…

`docker run -ti -p 4317-4318:4317-4318 streamfold/rotel --debug-log traces --exporter blackhole`

to start Rotel listening for OTLP traffic on ports 4317 & 4318.

Check out the Getting Started guide for more install and config instructions: https://rotel.dev/docs/setup/getting-started.

We’re super excited to share Rotel with you and would love to hear what you think!

Fun little project I built for myself. Create a YAML file for your dev-commands (lint, format, tests, etc), then expose those to coding agents via MCP.

https://github.com/scosman/hooks_mcp

Kinda like package.json scripts, but for agent runtimes, and commands are invoked via MCP.

1. Simple setup: one YAML file is all it takes to create a custom MCP server for your coding agents. Add the YAML to your repo to share with your team.

2. Tool discovery: coding agents know which dev-tools are available and the exact arguments they require. No more guessing CLI strings.

3. Improved security: limit which commands agents can run. Validate the arguments agents generate (e.g. ensure a file path is inside the project, not `~/.ssh/id_rsa`).

3. Works anywhere MCP works: Cursor, Windsurf, Cline, etc

4. Speed: using MCP unlocks parallel execution, requires fewer tokens for generating commands, and eliminates errors in commands requiring iteration.

5. And more: strip ANSI codes/control characters, .env file loading, define required secrets without checking them in, supports exit codes/stdout/stderr, etc

https://github.com/scosman/hooks_mcp

TL;DR: LLM Fighter is a small, open-source, playable benchmark for agentic behavior. You bring an OpenAI-compatible API; the demo runs in the browser. It creates head-to-head “battles” that stress tools, planning, and efficiency, and shows step-by-step logs you can download.

What it does well: quick, honest feel for how agents act under the same rules. What it’s not: a formal academic benchmark or a single “score”. Why I built it: I wanted something you can play in minutes and still learn from.

Hi HN,

YCW24 company here, we just open-sourced an AI image detection model that beats the SOTA commercial detectors. AI-generated images/videos have become incredibly good in the last few months and are flooding the internet; being able to detect them reliably gives some power back to consumers and companies that care about high-quality, genuine content.

Detecting AI-generated images is a very hard problem: there are many different techniques to generate images; there is image compression, noise, and other distortions that destroy generator artifacts; there's android phones applying auto-correction to images; etc. And none of the detectors we tried (sightengine.com, decopy.ai, etc.) works reliably even for basic examples (try it out with pirate Rick Astley made with Flux Kontext: https://imgur.com/a/iL3paE8).

We released two models, the full version (~600M params) and a smaller version (~20M params) that can even run in your browser on mobile (see demo)! We've also put up code for running things locally or via an API (free but rate-limited) using javascript/node and python code.

-------- Details --------

The full model was trained on 1M+ images that were scraped off the internet and the small model is a distillation. We're actively working on extending the dataset and further improving the models.

Classification accuracy: sightengine.com seems to be the best commercial solution out there, as confirmed by this (https://arxiv.org/pdf/2404.14581) paper, which they also cite on their website. Of course, they cherry-picked the results and claim 98.3% accuracy while only achieving (still impressive) ~82.8% over the full dataset. I've downloaded the dataset used in the paper and tested my models against it. The code for running the tests as well as a usable version of the dataset (the original was a big pain to download from OneDrive) are included in the repo code.

Here are our benchmark results for comparison: Total samples: 144,088 Real images: 17,044 Synthetic images: 127,044 Average precision: 0.991

[threshold: 0.5] Total accuracy: 0.864

PER-CATEGORY ACCURACY: Real 0.875 (17,044 samples) DALL-E_T2I 0.982 (16,110 samples) DreamStudio_T2I 0.968 (16,278 samples) Midjourney_T2I 0.961 (16,148 samples) StarryAI_T2I 0.847 (13,515 samples) DALL-E_IT2I 0.774 (16,665 samples) DreamStudio_IT2I 0.666 (16,139 samples) Midjourney_IT2I 0.897 (15,371 samples) StarryAI_IT2I 0.805 (16,818 samples)

[threshold: 0.65] Total accuracy: 0.827

PER-CATEGORY ACCURACY: Real 0.914 (17,044 samples) DALL-E_T2I 0.971 (16,110 samples) DreamStudio_T2I 0.949 (16,278 samples) Midjourney_T2I 0.940 (16,148 samples) StarryAI_T2I 0.803 (13,515 samples) DALL-E_IT2I 0.698 (16,665 samples) DreamStudio_IT2I 0.576 (16,139 samples) Midjourney_IT2I 0.845 (15,371 samples) StarryAI_IT2I 0.743 (16,818 samples)

Powerful Tools in One Platform: Screenshot Editor - Transform plain screenshots into eye-catching visuals with professional browser frames (macOS Safari, Chrome, Arc), beautiful gradient backgrounds, and perfect sizing for every platform.

Template Studio - Create stunning Open Graph images and social media graphics with professional templates, custom typography, logo positioning, and brand color customization.

- Before and After Template in beta for showing images in before and after state

Developers showcasing projects on GitHub/LinkedIn

Designers creating portfolio mockups for Dribbble/Behance

Content creators making engaging social media graphics

Marketers creating feature announcements and OG images

Anyone needing professional templates for social media and build in public.

Perfect Use Cases:

Social Media Content - Instagram stories, Twitter posts, LinkedIn graphics, YouTube thumbnails

Technical Documentation - API docs, tutorials, GitHub repository visuals, user guides

Presentations & Pitches - Product demos, investor decks, client presentations, feature announcements

Portfolio Showcases - Dribbble/Behance projects, personal websites, design case studies

Marketing Materials - Product launches, feature highlights, landing page graphics, email campaigns

Educational Content - Course materials, tutorial screenshots, training presentations

Business Communications - Open Graph images, blog featured images, professional mockups

App Store Screenshots - iPhone/iPad mockups, device frames, app showcases

Key Features:

Professional device & browser mockups

Open Graph templates with custom text & branding

50+ beautiful gradient backgrounds

70+ canvas presets for all social platforms

Logo upload and positioning controls

High-quality 4K exports

An event-first workflow that eliminates the need for Kafka and CDC by storing events as the source of truth and building read models from the event stream. Interested in feedback from the community on this approach to simplifying data architecture.

After 2 months of coding in my free time, I launched a directory where creators can showcase their websites, SaaS tools, mobile apps, AI tools, and digital projects. The platform helps developers and entrepreneurs present their work to the world and connect with a community of like-minded creators.

What you'll see on the site: - Latest projects from creators across different categories (SaaS, blogs, marketplaces, AI tools, etc.) - Detailed project pages with descriptions, logos, and direct links - Category filtering to discover specific types of projects - User profiles where creators can showcase all their projects - Premium features for enhanced visibility - Backlink verification system for quality control

The platform is designed to be a discovery hub where you can find interesting projects, get inspired by other creators' work, and potentially collaborate or learn from their approaches.

I'm working on giving you more control over your profile appearance. Soon you'll be able to customize layouts, add custom sections, choose themes, and create personalized project showcases. These features will help you create a more professional and unique presence on the platform.

Hey everyone,

We're building a tool that creates custom social posts styled exactly like your brand, in under 1 minute. If you’re spending 2 hours designing a single Instagram post, this might actually help.

You just enter your brand name, audience, and content type. In return, you get post-ready visuals that match your feed and brand vibe.

Who it's for: 1. Small business owners, freelancers, and solo creators doing everything themselves. 2. The product is currently in beta. We're looking for early users who can share honest feedback as we build.

What it does: 1. Generates branded visuals 2. Matches your tone and aesthetic 3. No editing needed (unless you want to tweak) 4. Save and reuse as templates

Why we're building it: 1. To save creators hours of repetitive design work 2. To help small brands stay consistent without burning out 3. To build and improve the product with real users

Every comments matters for us, PLEASE HELP US! Thank you :)