Show HN for August 23, 2025

my phone secretly robbed all my dreams and i didn’t even knew it got so bad. when i saw my screen time being about 11hrs with ~95 phone pickups per day, i realized how bad it got.

my problem is that i could not avoid social media entirely as i post actively because i’m involved in marketing and branding.

I tried the usual, app blockers & time limits, but i always find myself hitting “one more minute” like a junkie.

i’m not looking for a complete social media absence or deleting everything as these radical solutions don’t fit my lifestyle and needs and i tend to come back needing even more.

the ideal solution for me would be that when i decided to focus on something for specified time i SHOULDN’T reach out to my phone instinctively and needed a wrist slap every time i did so.

three weeks later i put out the first version of the @hardestfocusapp - it’s built on a simple, physical truth: to focus, you must disengage.

the core mechanic of the app is a commitment device.

focus starts not with a button tap, but with a physical act of disengagement with the phone itself that adapts to any situation:

- keep it face down on a desk for deep work.

- put in the pocket while walking or commuting.

- lock the phone for long and uninterrupted focus.

this makes the act of disengagement itself as a trigger for start of focus.

the consequence is just as real.

picking up the phone triggers a 5-second warning siren duel.

failure to put it back will erase all the progress made, using the concept of Loss Aversion - a powerful motivator that makes the fear of losing progress more compelling than any virtual reward.

it’s not just an app blocker, it became my phone blocker.

now when i’m focusing, i’m not on my phone every other minute trying to escape from the thought I could not confront, the warning siren is so loud, it feels like a slap on the wrist and i immediately put the phone away.

now i’m not just running away from the thoughts, I’m confronting them and planning ahead

OctaneDB is an open-source vector database for Python that focuses on ultra-fast similarity search for high-dimensional data—perfect for AI/ML, semantic search, and large-scale document or embedding retrieval.

What does it do?

Store, index, and search millions of embeddings (text, images, etc.) with sub-millisecond query time.

Supports in-memory and efficient HDF5 persistent storage.

Integrates seamlessly with sentence-transformers for automatic text embedding.

Key Features:

10x faster than Pinecone or ChromaDB for vector search and batch insertions.

Advanced indexing: HNSW (approximate nearest neighbor), FlatIndex

Batch search, advanced metadata filtering, GPU acceleration

Hi HN! I'm North, DevRel from the CREAO team. We are a 15-20 person startup, and we spent a couple of months building this vibe coding product. We've been exploring the possibilities of how agents should assist people's work for productivity in various scenarios.

Three months ago, we noticed that all small teams and companies are struggling with subscribing to multiple SaaS tools that never fit their exact needs. Pretty much all founders and SMB professionals need customized productivity tools that fit their daily work routines, but they don't have developers to build these lightweight tools for them. So we thought we needed to build a vibe coding product that allows them to ship their own tools in minutes and get rid of these SaaS subscriptions.

For our product features, we built it differently from all the other vibe coding products. We want our users to build something actually useful instead of just a product prototype. But the current well-known code agents are performing poorly at building complex data schemas and integrating APIs. So we fine-tuned a small model for the following features:

1.Automatically builds complex relational data schemas

2.Automated MCP/API connections with existing systems

With these core capabilities, our code agent can then build internal tools that connect with existing systems with state persistence. For example, I built a dashboard to track my X engagement and export results to a CSV file.

https://drive.google.com/file/d/1FBBOdrtonlj_56jgQuAoPpGnABB...

This productivity boost is specifically designed for small team founders—you can build your own CRM workflows, data dashboards, revenue tracking, etc. Each of these could cost you $20-50/month in SaaS subscriptions.

We are still in the early stage and improving our user experience. Our team is delivering new features pretty much every week, and we would love to hear from HN to give us some suggestions on how we can do better. You can always reach me at [email protected]. I love to shape the product together with community users. I hope you can find our work very helpful for your productivity.

Thanks!

AgentState to solve a problem I kept running into: managing state for multi-agent AI systems is surprisingly hard. When you have multiple AI agents that need to coordinate, persist their state, and query each other's status, you typically end up with a mess of Redis/Postgres setups, custom queuing, and manual synchronization code.

The whole thing is ~3MB, written in Rust for performance and safety, runs in Docker, and handles 1000+ ops/sec. I've been running it in production for AI workflows and it's been rock solid.

I work at a so called "hyperscaler" in billing and invoicing and learned that LLM use is metered and billed in "tokens". I had never really used LLMs at the time. Now that I do (including to make this site, with heavy oversight), this is obvious because they all tell you how many tokens some query is using.

This reminded me of "Token Mania" of 2020-2021 so I decided to combine the two concepts in an absolutely useless way - enter the HypeVortex.

I made a token on Polygon (contract here: https://polygonscan.com/address/0xf06bEedB6B21B3072b424919be...). This token CANNOT be exchanged, it can only go from me to users and back to me. It's worthless.

You login with email (sorry), and get a code emailed to you (again, sorry). Then a wallet is made for you, you claim HypeVortex tokens and spend them on AI. Your email is stored as a hash, I don't know what it is and I can't see what anyone does.

All in a geocities style page. With midi.

I have no profit motive here, it's entirely silly fun. For those of you born early 90s and before, I hope you have some nostalgic fun with it.

After years of using screeners that were missing features or were expensive, I built this free alternative available on desktop and mobile. You can filter and sort by all the essentials:

Symbol, Name, Price, Market Cap, Change %, Change, Volume; P/E and Forward P/E ratios

Real example that saved me hours: I was searching for (cheap & solid) stocks to sell options. I filtered with PE and Forward PE under 25, volume above 1 million, price minimum $25, and market cap over $10 billion and discovered a lot of new stocks that nobody talks about on youtube or elsewhere.

Also I tried to make the design as good as I could because a lot of financial website are totally outdated in term of styling. Would love feedback from the HN community - contact us through the website if you have suggestions or need specific tools that would be useful to you

Thank you for reading & I hope you get some value from it :)

Hi HN

I’ve been maintaining more than 20 open-source Laravel packages over the years, and I quickly ran into the same pain many of you probably know: keeping track of dependencies, versions, vulnerabilities, issues, and contributor activity across dozens of GitHub repositories is exhausting.

So I built [RepoSentinel](https://reposentinel.com) – a dashboard that connects to your GitHub account and gives you a single place to:

- Monitor dependencies and outdated packages - Get alerts on security vulnerabilities - Track repo activity and contributor stats - Spot potential risks before they become incidents

It started as a tool I built for myself, but I decided to make it available to others. It’s not another “yet another GitHub client” – the focus is really on *maintainability and risk management* for people juggling multiple repos.

Would love feedback from the HN crowd: - Does this scratch an itch you’ve had? - What’s missing or annoying in your current workflow? - Any blind spots you see in the approach?

Repo: [https://reposentinel.com](https://reposentinel.com)

Thanks for reading!

I had always found meal tracking too boring. Healthify had huge forms that are boring to fill after each meal. Then it let you take pictures but then I always forget it to take them before eating.

Meal tracking was such a pain, so I built an app where you can simply say what you ate and it will break it down into macros immediately. Now I'm that annoying guy who talks to his phone every time he ate.

I built a lightweight GIF decoder in pure C, ideal for embedded or performance-critical environments. It’s header-only, zero dynamic memory allocations, and fully platform-independent. Supports both static and animated GIFs, with turbo and safe decoding modes. Works great on microcontrollers, IoT devices, and anything with a framebuffer. Would love feedback or ideas where this could be useful.

GitHub: https://github.com/Ferki-git-creator/TurboStitchGIF-HeaderOn...

Hi Everyone,

I built a website to make finding flats and flatmates super easy! You can post your properties or create a request for a place that matches your needs.

No need to juggle multiple Facebook or Telegram groups or subreddits. Our site has filters and sorting options (by price, location, gender, etc.) to help you find exactly what you're looking for.

The standout feature? Location-based searching! Just enter any landmark available on Google Maps, apply additional filters and sorting, and get nearby tailored results that fit your preferences.

We prioritized privacy: you can browse without logging in, and a phone number is optional when creating an account.

It's completely free, with no limits on how many contacts you can make.

We hope you love our project! We're considering adding a "Support Us" page to help cover server costs.

Would love to hear your feedback!

I am running a AI Read Team challenge with a bug bounty of $100

gl

I review my code in multiple passes before committing. First pass goes to "reviewed", thorough second pass moves files to "ready".

    git cl add reviewed utils.py parser.py  # First review  
    git cl add ready tests.py               # Double-checked
    git cl status                           # See review progress
    git cl commit ready -m "Add feature"

Built this because Git's staging area is binary but my review process isn't. Need to track which files I've reviewed once vs twice. Single Python file that adds multiple named staging areas to Git.

https://github.com/BHFock/git-cl

I built a web app called *AI Image Description Generator* that uses AI to instantly turn any image, you upload into rich, context-aware text.

You can choose from 11 different description modes—such as *detailed narrative*, *concise alt-text* (WCAG-compliant), *tweet-style caption*, *Instagram post*, or even *AI art prompts* suitable for tools like 4o, The tool also supports over 35 languages and allows you to tweak text length (15–1000 chars).

You will know the detailed content of the picture, such as the product title when writing e-commerce shelves, or the SEO of the picture

More features will be added in the future, so please feel free to discuss and use them.

Try it here: https://www.image-description-generator.net

A third iteration of my very low compute, novel prompt architecture chatbot, that treats the prompt as a programmable information waveform - with graphical sequencing and mixing interface.

I made a new library to detect malicious and harmful behaviour in Python scripts. It uses static analysis with semantic modeling. Even when the code is pretty obfuscated, it can still detect it.

For example, it can infer that

    getattr(sys.modules["built"+"ins"], "".join(reversed(["al","ev"])))("1+1")

Is basically:

    eval("1+1").

Currently, I'm testing it on public files where some of them implement malicious behavior, as well as past malicious packages on PyPI.

You can see some of the detection examples here: https://github.com/rushter/hexora/blob/main/docs/examples.md

I'd love to hear your feedback and ideas on how to improve this and identify missing rules.