2025年9月1日 の Show HN

Pocket is shutting down and I really, really liked it. So I built woomarks, an app that let's you save links with a similar UI. It's very minimal, but it's doing everything I liked from Pocket and you can bulk import your links and use the app or self-host.

- Public app that you can test: https://woomarks.com/

- My self-hosted version, where you can see my saves: https://roberto.fyi/bookmarks/

- Repository if you want to self-host: https://github.com/earlyriser/woomarks

Export links from Pocket here: https://getpocket.com/export the last day will be on October 20025.

Features: - Add/Delete links - Search - Tags - Bookmarklet (useful for a 2-click-save) - Data reads from: csv file in server (these links are public) local storage in browser (these links are visible just for the user) - Local storage saving. - Import to local storage from csv file - Export to csv from local storage. - Export to csv from csv file (useful when links are "deleted" using the app and just hidden using a local storage blacklist). - Export to csv from both places. - No external libraries. - Vanilla css code. - Vanilla js code.

Hi HN,

I've been working on AfriTales, a flutter based mobile app that brings African folktales into modern stories narrated episodes wrapped in a children and adult friendly UI player. The stories are created to cover north, south, west and east Africa. I think of it as a digital by-the-fire-side.

Why AfriTales: Cultural relevance: There is a gap in culturally-rich audio-native storytelling apps for Africans, the diaspora and people interested in African stories. Modern Influence: Modern UI makes the the app feel elegant and emotionally resonant. Retention via structure: Episodes are short (2-5 minutes) and there are stories series for premium users.

MVP features include: A launch landing page (https://afritales.org/) for early engagement and waitlist signups. I have currently sourced over 100 stories. Thanks to Google's Gemma 3 270M, users can generate stories with their own twist. Freemium model: 3 free tales per day, plus premium subscription for unlimited access. Robust Flutter structure: Architecture with TTS integration, and images for context.

I am starting in Ghana before expanding, and I'd love feedback from this community: Would you (or your child) use an audio-based storytelling app with a strong regional cultural tie? Suggestions for retention strategies or content formats that engage long-term users?

Thanks

My friend and I grew frustrated with the high cost of existing pair programming tools, and of course of grainy screens when we used Huddle or similar tools.

We believe core developer collaboration shouldn't be locked behind an expensive subscription.

So for the past year we spent our nights and weekend building Hopp, an open-source alternative.

We would love your feedback and we are here to answer any and all questions.

A simple .NET NuGet server implementation built on Node.js that provides essential NuGet v3 API endpoints.

Key Features:

* Easy setup, run NuGet server in 10 seconds! * NuGet V3 API compatibility: Support for modern NuGet client operations * No need database management: Store package file and nuspecs into filesystem directly, feel free any database managements * Package publish: Flexible client to upload .nupkg files via HTTP POST using cURL and others * Basic authentication: Setup authentication for publish and general access when you want it * Reverse proxy support: Configurable trusted reverse proxy handling for proper URL resolution * Modern Web UI with enhanced features. * Package importer: Included package importer from existing NuGet server * Docker image available

I wrote a small local tool to transcribe audio notes (Whisper/Parakeet). Code: https://github.com/bilawalriaz/lazy-notes

I wanted to process raw transcripts locally without OpenRouter. Llama 3.2 3B with a prompt was decent but incomplete, so I tried SFT. I fine-tuned Llama 3.2 3B to clean/analyze dictation and emit structured JSON (title, tags, entities, dates, actions).

Data: 13 real memos → Kimi K2 gold JSON → ~40k synthetic + gold; keys canonicalized. Chutes.ai (5k req/day).

Training: RTX 4090 24GB, ~4h, LoRA (r=128, α=128, dropout=0.05), max seq 2048, bs=16, lr=5e-5, cosine, Unsloth. On 2070 Super 8GB it was ~8h.

Inference: merged to GGUF, Q4_K_M (llama.cpp), runs in LM Studio.

Evals (100-sample, scored by GLM 4.5 FP8): overall 5.35 (base 3B) → 8.55 (fine-tuned); completeness 4.12 → 7.62; factual 5.24 → 8.57.

Head-to-head (10 samples): ~8.40 vs Hermes-70B 8.18, Mistral-Small-24B 7.90, Gemma-3-12B 7.76, Qwen3-14B 7.62. Teacher Kimi K2 ~8.82.

Why: task specialization + JSON canonicalization reduces variance; the model learns the exact structure/fields.

Lessons: train on completions only; synthetic is fine for narrow tasks; Llama is straightforward to train. Dataset pipeline + training script + evals: https://github.com/bilawalriaz/local-notes-transcribe-llm

This is the first half of my book, “Swimming in Tech Debt”. It is available at a pre-launch sale price of $0.99 (https://loufranco.com/tech-debt-book).

I have been working on it since January 2024. It is based on some posts in my blog, but expands on my ideas quite a bit.

In September 2024, excerpts appeared in Gergely Orosz’s Pragmatic Engineer newsletter, which helped me get a lot of feedback that expanded the book from my initial idea. This half is about what I expected to do before that —- the rest of the book goes into team and CTO practices.

I love the simplicity, expressibility and extendibility of Nunjucks.

But I was not able to find something with similar for Java, especially with the same syntax.

So, built one. And it's pretty fast too.

https://github.com/freakynit/Blueprint

Not sure this is the right way to post this, but I'm sure quite a few people are as frustrated as I am by the AI enshittification of Google search and would like to know this.

I accidentally discovered in a fit of rage against Google Search that if you add an expletive to a search term, the SERP will avoid showing ads and also an AI overview.

The good thing is that it works also with the "-" (minus) operator, so you can make sure the expletive is actually not included in the result pages.

Try it yourself: search for a fairly generic query that gives you ads and AI overview, and add "-f*k" at the end, uncensored of course.

Enjoy a much better search experience. It might be placebo, but it feels like the results are actually better sorted.

Edit: edited to avoid HN pro-expletives filter :D

I recently made the game that is bundled in the Comet browser by Perplexity. It's a sort of replacement for Chrome Dino game for when you lose network connection. You can read about my process of prototyping and game design philosophy, with accompanying videos, in this blog post.

Happy Labour Day,

For your consideration is a rendering of the Moon with changing Lunar phases, Earth-Moon distance, and librations all in ASCII. The inspiration was asciimoon.com that previously appeared on HN but did not accurate phases.

Most orchestration frameworks today still behave like fragile chains — they break when faced with contradictions, long-term memory, or dynamic routing.

Neuron is a cognitive multi-agent architecture that thinks in circuits instead of chains. Multiple agents collaborate in parallel, adapt their pathways in real time, and keep persistent context across extended interactions.

Key components

Agents: Intake, Reasoning, Response, Memory

Circuits: Dynamic routing instead of linear chaining

Memory: Episodic + contextual persistence

Monitoring: Full reasoning traces for observability

Why it matters

Handles contradictory inputs without collapsing

Maintains state across extended sessions

Parallel coordination for complex reasoning tasks

Transparent logs for debugging & trust

GitHub repo: https://github.com/ShaliniAnandaPhD/Neuron

Evaluation Notebook: https://www.notion.so/shalini-ananda-phd/Neuron-Evaluation-N...

Tutorial Series: https://www.notion.so/shalini-ananda-phd/Neuron-Framework-Tu...

About me / context: https://www.notion.so/shalini-ananda-phd/Shalini-Ananda-PhD-...

Would love feedback from the HN community — especially if you’ve run into the same breakdown points with traditional tools.

Phispr is an ephemeral chat application designed for anonymous, temporary conversations that vanish without a trace. Built with Go, it offers both web and terminal user interfaces.

A weekend project exploded into a two weeks project, and with a funny origin. https://github.com/bnkamalesh/phispr/blob/main/docs/genesis....

Hi HN, I did this project over the Labor Day weekend.

My family runs off our shared Google Calendar. This project was to use a ESP32 and three-color e-ink display to create a wall-mounted daily updated calendar, weather and news dashboard for us to coordinate our lives.

We built StripeMeter, an open-source usage metering platform that plugs directly into Stripe. It solves the classic SaaS pain of “why is my bill higher than expected?” by giving both developers and customers real-time usage tracking with live cost projections. Why it matters:

- Transparency: Customers see exactly what Stripe will bill them (within 0.5% parity).

- Exactly-once guarantee: No double billing, ever.

- Fast & scalable: Sub-minute freshness with Redis + Postgres counters.

We’d love feedback from SaaS builders, especially if you’ve struggled with Stripe’s metered billing. Does this solve a real pain for you? What would you need before trusting it in production?

I've been using Substack for my tech blog and newsletter for a while.

As a developer, I truly believe that sharing knowledge through a newsletter is one of the best ways to grow an audience — and eventually monetise your efforts.

Substack is built for a general audience, not devs. I was missing code highlighting and minimalist design.

I wanted to build something dedicated for developers: - open-source, free to use, self-hosted - beautiful code highlighting - markdown-first writing - clean, minimalist design - dark mode support

I believe that an open-source model is a great way to build a product. IndiePubStack is: - fully open-source and free to use (MIT licence) - designed to be self-hosted (Docker image distribution) - built on top of reliable services (Kinde for auth, Resend for email delivery)

All you need is an inexpensive VPS and a domain name, free Kinde and Resend accounts (that will be enough for a long time).

I'm already using IndiePubStack for my personal blog and newsletter.

The future depends on the community feedback, if the open-source project gets some traction, the next logical step will be to build a SaaS platform on top of it.

Hi HN! I’m excited to share my first product launch, HTML Capture Compare. This is also my first time posting on Hacker News, so thank you for any feedback you might have.

I recently founded qaas.work, a company that helps small teams without automation expertise. We get embedded with a client’s team to set up automation frameworks, CI/CD flows, and reporting from scratch. Once the client's team is comfortable and up to speed, we step back. This approach helps smaller companies get robust automation without the high cost of a full-time engineer.

In my work, I've constantly seen a huge time sink: debugging flaky tests caused by changes in the UI. Automation teams waste countless hours manually inspecting HTML to figure out why a locator broke.

That's why I built this Chrome extension. It’s a simple, visual tool to quickly capture and compare HTML structures, making it easy to find what’s causing test failures.

Here are the main features:

Side-by-Side Compare: Get a clear view of the old and new HTML structures in a separate window to pinpoint exactly what changed.

Merged Compare: See the updated structure directly on the webpage for a more streamlined view of the changes.

Manage Baselines: Keep track of all your captured HTML baselines in one place, with the ability to view and delete them as needed.

My goal is to help you debug tests faster and keep your automation pipelines running smoothly, which is a key challenge for every team I've worked with.

You can try it out directly from the Chrome Web Store: [https://chromewebstore.google.com/detail/html-capture-compar...]

I’d really appreciate any thoughts or feedback from the community. If you’ve dealt with this problem before, I'd love to know what other features would be helpful!

SARPRO is a fast Rust toolset for turning Sentinel‑1 GRD products into GeoTIFFs and high‑quality synthetic‑RGB JPEGs, with CLI, GUI, and a typed Rust API. The latest release focuses on cloud‑friendly remote I/O with proper authentication and robust batching.

What’s new in 0.3.2:

Authenticated CDSE/ASF downloads (CLI) to fetch GRD .SAFE.zip over OAuth2 (CDSE) or Earthdata cookies/netrc (ASF) and process directly.

Stream or materialize: Reads via GDAL VSICURL/VSIZIP for streaming, or materialize into a cache with progress bars and ASF resume support.

Non‑auth remote SAFE: Process unpacked SAFE over HTTP(S) behind a Range‑capable server (e.g., nginx, Python RangeHTTPServer)

Quality & speed: Auto‑CRS (--target-crs auto), CLAHE autoscale, synRGB modes; 400–500MP → 2048px typically completes in ~1–2s on a modern laptop; native full‑res warps remain tens of seconds.

Links: - Repo: (https://github.com/bogwi/sarpro) - Docs: (https://docs.rs/sarpro) - Previous HN: [Sarpro – Ultrafast Sentinel‑1 GRD → GeoTIFF/JPEG, SynRGB, Clahe, Auto‑CRS (Rust)](https://news.ycombinator.com/item?id=44961747)

I tried setting up an AI git commit tool this weekend and it was hell.

All the tools are either not maintained, outdated, or make it a total pain to customize the system prompt. All I wanted was simple AI commit messages, that I can customize!

Introducing Giz: A tiny, drop in replacement for `git commit`, with ai commit messages.

`giz commit` behaves exactly like `git commit` with one exception:

The commit message is AI generated when no `-m` or `--message` argument is passed. User is asked for confirmation first. Its behavior is otherwise identical.

Easy to modify system prompt:

The prompt is stored in a textfile located at: `giz prompfile`. The diff (`git diff --staged`) is pasted two newlines below whatever you put in this file.

Quickstart:

- `pip install giz`

- `giz set_openai_api_key <your-api-key>`

- `giz commit ...`

Go to the link to see a demo!

We built a minimal browser extension for vulnerability readers: it highlights CVE IDs anywhere and on hover shows CVSS/EPSS, KEV, and references/exploits count. No login, no paywall. Looking for feedback on edge cases and data you want in the hover.

Hey HN,

I built something I've been wanting for years: a way to access the power of Google Apps Script inside Sheets on demand, without having to switch to the code editor and remembering Apps Script APIs.

## Why bother?

Apps Script is actually underrated: it lets you do arbitrary structured logic (when formulas get unreadable/convoluted), procedural transformations, and even add custom UIs (menus, dialogs, sidebars). It goes further than macros, with triggers, etc.

But the pain is real: remembering the APIs, figuring out deployment, debugging code in an unfamiliar environment, etc. For non-technical folks, it’s simply inaccessible.

## What Kyran does

It lives as a Sheets add-on (sidebar). You describe what you want, it generates Apps Script code, installs it into your Sheet, and you can run it right away.

You can iterate in the "spreadsheet-native" way: not by reading the code, but by trying it on some data and checking the result. If it's wrong, you tweak the description.

The code it generates is short (usually <50 lines), uses a very limited API surface, and can be tested quickly. That makes the usual risks of AI hallucination far less of an issue in practice than in general-purpose coding.

## What I love (technically) about this

It's a narrow domain where AI codegen actually works rather reliably: small programs, well-defined APIs, immediate execution in a data-rich context.

It matches how people already use spreadsheets: experiment -> check results -> adjust. People asking for formula help on the internet don't usually try to decipher the monstrosities given to them, and you don't need to read the code here either. The code is running in a very tight sandbox Google spent a lot of time improving, and it's even more restricted for our use case.

For power users, it's essentially "vibe coding" Apps Script: you still end up with real code, just without the boilerplate and deployment tax. For others, it just makes possible something they could only dream of.

I've been trying it on Google Sheets forum posts asking for help, and it works great.

## Who might care

- Spreadsheet power users who regularly outgrow formulas.

- Engineers who hate chasing docs just to write 30 lines of Apps Script for a one-off task

- Ops/analyst teams that hack Sheets together as their internal "apps."

## Limitations

The scripts you create inherit the permissions of the Kyran add-on itself. Apps Script can control a lot of other Google Apps, including Drive, but I don't want to scare people away with a huge permission prompt.

This is therefore currently limited to manipulating the current sheet only. I hope the permission model for add-ons evolves so I can ask for any permission (like calendar access) at the exact point where user generated code needs it.

## Links

[Homepage / docs](https://blog.darkaa.com/kyran-ai-app-builder-for-sheets-user...)

[Add-on install](https://workspace.google.com/marketplace/app/kyran_ai_app_bu...)

You get some free requests, and can add your Gemini API key for more.

Yes, this has massive platform risk. Sanity is not something I can afford right now ;)

Would love feedback, especially from folks who've wrestled with Apps Script before.

I'm a solo founder launching ExeTrace, a tool that introduces a new cybersecurity category: Executable Drift Monitoring (EDM). Unlike AV, FIM, or EDR, EDM focuses on detecting behavioral drift in executables, subtle changes that signal tampering, misuse, or repurposing. ExeTrace monitors how binaries behave over time, catching threats that traditional tools miss. I built this to solve a real-world problem: how do you know when a trusted executable starts doing something it shouldn’t? ExeTrace answers that with forensic clarity and modular design. Would love feedback on: - The concept of EDM as a category - Messaging and onboarding clarity - How it compares to existing endpoint tools Landing page: https://sites.google.com/view/exetrace/home Thanks for checking it out, excited to hear what you think.