The best Hacker News stories from Show from the past week
Latest posts:
Show HN: WikiBinge – discover how all things are vaguely connected
Connect two articles on Wikipedia, but do it the long way. I've always been a fan of the theory of six degree of separation, but it's an overused concept when exploring the Wiki-graph.<p>Instead of showing the shortest path, which in my opinion is "boring" and ends up connecting super-important central articles, I came up with my own method: WikiBinge selects the smaller, less represented articles on Wikipedia. In a WikiBinge path, the underdogs are the kings!<p>How does it work? It's pretty straightforward! Compute PageRank on the Wiki-graph and assign as weight of each edge the PageRank value of the destination node. A WikiBinge path is then simply a shortest path using these weights: the algorithm will then favor paths passing through articles with lower PageRank values.<p>More on the motives to build this here: <a href="https://www.jamez.it/project/wikibinge/" rel="nofollow">https://www.jamez.it/project/wikibinge/</a><p>This is an older project of mine, but it never got much exposure, so I'm humbly submitting it now.
Show HN: Google Analytics alternative with the most generous free tier
Hi HN,<p>As an indie hacker, the new Google Analytics (GA4) coming motivated me to look for a straightforward alternative that would also be affordable. I had a few basic product requirements and didn’t want to spend too much to replace a free product. There are a lot of great Google Analytics alternatives out there, but the pricing didn’t seem right. As someone who likes to just build things, many of which aren’t businesses yet, it didn’t make sense to pay for options like Plausible and Fathom out of the gate.<p>So I joined with a friend to build Beam Analytics. Beam gives you all the standard web analytics. It also comes with easy to create funnels so you can see how users move through your site. And we have a great proxy for cohort retention that doesn’t need you to log any data with us. It’s cookie-less and GDPR compliant.<p>The free tier is 100k page views per month so hopefully you’ll give it a try. There’s also a Wordpress integration to make integrating with WordPress sites as easy as a single click - <a href="https://wordpress.org/plugins/beam-analytics/" rel="nofollow">https://wordpress.org/plugins/beam-analytics/</a>.<p>Appreciate your feedback. You can also email us at hi (at) beamanalytics.io or DM me on twitter @TheBuilderJR.
Show HN: Twitter API Reverse Engineered
endpoints: /graphql /1.1 /2 /i
Show HN: Personalized book recommendations with Librarian AI
Show HN: GPT-4-powered web searches for developers
Hi HN,<p>Today we’re launching GPT-4 answers on Phind.com, a developer-focused search engine that uses generative AI to browse the web and answer technical questions, complete with code examples and detailed explanations. Unlike vanilla GPT-4, Phind feeds in relevant websites and technical documentation, reducing the model’s hallucination and keeping it up-to-date. To use it, simply enable the “Expert” toggle before doing a search.<p>GPT-4 is making a night-and-day difference in terms of answer quality. For a question like “How can I RLHF a LLaMa model”, Phind in Expert mode delivers a step-by-step guide complete with citations (<a href="https://phind.com/search?cache=0fecf96b-0ac9-4b65-893d-8ea5708db222">https://phind.com/search?cache=0fecf96b-0ac9-4b65-893d-8ea57...</a>) while Phind in default mode meanders a bit and answers the question very generally (<a href="https://phind.com/search?cache=dd1fe16f-b101-4cc8-8089-ac56defb228d">https://phind.com/search?cache=dd1fe16f-b101-4cc8-8089-ac56d...</a>).<p>GPT-4 is significantly more concise and “systematic” in its answers than our default model. It generates step-by-step instructions over 90% of the time, while our default model does not.<p>We’re particularly focused on ML developers, as Phind can answer questions about many recent ML libraries, papers, and technologies that ChatGPT simply cannot. Even with ChatGPT’s alpha browsing mode, Phind answers technical questions faster and in more detail.<p>For example, Phind running on “Expert” GPT-4 mode can concisely and correctly tell you how to run an Alpaca model using llama.cpp: (<a href="https://phind.com/search?cache=0132c27e-c876-4f87-a0e1-cc48f07ccc20">https://phind.com/search?cache=0132c27e-c876-4f87-a0e1-cc48f...</a>). In contrast, ChatGPT-4 hallucinates and writes a make function for a fictional llama.cpp.<p>We still have a long way to go and would love to hear your feedback.
Show HN: Skip the SSO Tax, access your user data with OSS
As the former CTO of an Insurtech and Fintech startup I always had the “pleasure” to keep regulators and auditors happy. Think of documenting who has access to what, quarterly access reviews, yearly audits and so on…<p>Like many others we couldn’t justify the Enterprise-plan for every SaaS tool to simply get access to SSO and SCIM/SAML APIs. For Notion alone the cost would have nearly doubled to $14 per user per month. That’s insane! Mostly unknown to people, SSO Tax also limits access to APIs that are used for managing user access (SCIM/SAML).<p>This has proven to be an incredibly annoying roadblock that prevented me from doing anything useful with our user data:
- You want to download the current list of users and their permissions? Forget about it!
- You want to centrally assign user roles and permissions? Good luck with that!
- You want to delete user accounts immediately? Yeah right, like that's ever gonna happen!<p>It literally cost me hours to update our access matrix at the end of every quarter for our access reviews and manually assigning user accounts and permissions.<p>I figured, there must be a better way than praying to the SaaS gods to miraculously make the SSO Tax disappear (and open up SCIM/SAML along the way). That’s why I sat down a few weeks ago and started building OpenOwl (<a href="https://github.com/AccessOwl/open_owl">https://github.com/AccessOwl/open_owl</a>). It allows me to just plug in my user credentials and automatically download user lists, including permissions from SaaS tools.<p>Granted, OpenOwl is still a work in progress, and it's not perfect. At the moment it's limited to non-SSO login flows and covers only 7 SaaS vendors. My favorite part is that you can configure integrations as “recipes”. The goal was for anybody to be able to add new integrations (IT managers and developers alike). Therefore you ideally don’t even have to write any new code, just tell OpenOwl how the new SaaS vendor works.<p>What do you think? Have you dealt with manually maintaining a list of users and their permissions? Could this approach get us closer to overcoming parts of the SSO Tax?
Show HN: Supavisor – a Postgres connection pooler written in Elixir
hey hn, supabase ceo here<p>this is a postgres connection pooler. it’s similar to pgbouncer, but built with Elixir and specifically designed for multi-tenancy.<p>it’s still under development, but it’s at a stage where we can gather a feedback from the community and you can try it yourself. we aren’t using this in production yet, but aiming to deploy it for a subset of databases in the next 2 months.<p>We have the following benchmarks (details in the readme):<p><pre><code> - Elixir Cluster maintaining 400 connections to a single Postgres database
- 1_000_000 clients connecting to the Elixir cluster
- Sending 20_000 transactions per second
- Consuming 7.8G RAM and ~50% CPU on a 64vCPU machine
</code></pre>
supavisor can be run as a cluster or a single node/binary. It’s handling 90%+ of the throughput of pgbouncer on a local machine (running pgbench)<p>we will place this in front of all supabase databases. It will eventually be able to handle multiple types of connections: traditional TCP connections, and HTTP connections for developers who are connecting to Postgres in serverless environments using Prisma, Kysely, Drizzle, etc<p>the proxy will serve as a connection buffer while we scale databases: scaling up compute with zero-downtime, and for scale-to-zero - triggering a server restart when a connection is initiated<p>finally, i want to shout out to Jose and the Dashbit/elixir team. They were extremely helpful with the design & architecture. they have been valuable partners, and elixir continues to be an amazing language for tools like this and our Realtime server.
Show HN: BrowserBox – do stuff with browsers that you can't normally
Show HN: ChatGDB – GPT-Powered GDB Assistant
ChatGDB is a tool designed to superpower your debugging experience with GDB, a debugger for compiled languages. Use it to accelerate your debugging workflow by leveraging the power of ChatGPT to assist you while using GDB!<p>It allows you to explain in natural language what you want to do, and then automatically execute the relevant command. Optionally, you can ask ChatGPT to explain the command it just ran or even pass in any question for it to answer. Focus on what's important - figuring out that nasty bug instead of chasing down GDB commands at the tip of your tongue.<p>See it here: <a href="https://github.com/pgosar/ChatGDB">https://github.com/pgosar/ChatGDB</a>
Show HN: ADHD-friendly Pomodoro web app
Show HN: Tabby – A self-hosted GitHub Copilot
I would like to introduce Tabby, which is a self-hosted alternative to GitHub Copilot that you can integrate into your hardware. While GitHub Copilot has made coding more efficient and less time-consuming by assisting developers with suggestions and completing code, it raises concerns around privacy and security.<p>Tabby is in its early stages, and we are excited to receive feedback from the community.<p>Its Github repository is located here: <a href="https://github.com/TabbyML/tabby">https://github.com/TabbyML/tabby</a>.<p>We have also deployed the latest docker image to Huggingface for a live demo: <a href="https://huggingface.co/spaces/TabbyML/tabby" rel="nofollow">https://huggingface.co/spaces/TabbyML/tabby</a>.<p>Tabby is built on top of the popular Hugging Face Transformers / Triton FasterTransformer backend and is designed to be self-hosted, providing you with complete control over your data and privacy. In Tabby's next feature iteration, you can fine-tune the model to meet your project requirements.
Show HN: Tabby – A self-hosted GitHub Copilot
I would like to introduce Tabby, which is a self-hosted alternative to GitHub Copilot that you can integrate into your hardware. While GitHub Copilot has made coding more efficient and less time-consuming by assisting developers with suggestions and completing code, it raises concerns around privacy and security.<p>Tabby is in its early stages, and we are excited to receive feedback from the community.<p>Its Github repository is located here: <a href="https://github.com/TabbyML/tabby">https://github.com/TabbyML/tabby</a>.<p>We have also deployed the latest docker image to Huggingface for a live demo: <a href="https://huggingface.co/spaces/TabbyML/tabby" rel="nofollow">https://huggingface.co/spaces/TabbyML/tabby</a>.<p>Tabby is built on top of the popular Hugging Face Transformers / Triton FasterTransformer backend and is designed to be self-hosted, providing you with complete control over your data and privacy. In Tabby's next feature iteration, you can fine-tune the model to meet your project requirements.
Show HN: Want something better than k-means? Try BanditPAM
Want something better than k-means? I'm happy to announce our SOTA k-medoids algorithm from NeurIPS 2020, BanditPAM, is now publicly available! `pip install banditpam` or `install.packages("banditpam")` and you're good to go!<p>k-means is one of the most widely-used algorithms to cluster data. However, it has several limitations: a) it requires the use of L2 distance for efficient clustering, which also b) restricts the data you're clustering to be vectors, and c) doesn't require the means to be datapoints in the dataset.<p>Unlike in k-means, the k-medoids problem requires cluster centers to be actual datapoints, which permits greater interpretability of your cluster centers. k-medoids also works better with arbitrary distance metrics, so your clustering can be more robust to outliers if you're using metrics like L1. Despite these advantages, most people don't use k-medoids because prior algorithms were too slow.<p>In our NeurIPS 2020 paper, BanditPAM, we sped up the best known algorithm from O(n^2) to O(nlogn) by using techniques from multi-armed bandits. We were inspired by prior research that demonstrated many algorithms can be sped up by sampling the data intelligently, instead of performing exhaustive computations.<p>We've released our implementation, which is pip- and CRAN-installable. It's written in C++ for speed, but callable from Python and R. It also supports parallelization and intelligent caching at no extra complexity to end users. Its interface also matches the sklearn.cluster.KMeans interface, so minimal changes are necessary to existing code.<p>PyPI: <a href="https://pypi.org/project/banditpam" rel="nofollow">https://pypi.org/project/banditpam</a><p>CRAN: <a href="https://cran.r-project.org/web/packages/banditpam/index.html" rel="nofollow">https://cran.r-project.org/web/packages/banditpam/index.html</a><p>Repo: <a href="https://github.com/motiwari/BanditPAM">https://github.com/motiwari/BanditPAM</a><p>Paper: <a href="https://arxiv.org/abs/2006.06856" rel="nofollow">https://arxiv.org/abs/2006.06856</a><p>If you find our work valuable, please consider starring the repo or citing our work. These help us continue development on this project.<p>I'm Mo Tiwari (motiwari.com), a PhD student in Computer Science at Stanford University. A special thanks to my collaborators on this project, Martin Jinye Zhang, James Mayclin, Sebastian Thrun, Chris Piech, and Ilan Shomorony, as well as the author of the R package, Balasubramanian Narasimhan.<p>(This is my first time posting on HN; I've read the FAQ before posting, but please let me know if I broke any rules)
Show HN: Live coaching app for remote SWE interviews, uses Whisper and GPT-4
Posting from a throwaway account to maintain privacy.<p>This project is a salvo against leetcode-style interviews that require candidates to study useless topics and confidently write code in front of a live audience, in order to get a job where none of that stuff matters.<p>Cheetah is an AI-powered macOS app designed to assist users during remote software engineering interviews by providing real-time, discreet coaching and integration with CoderPad. It uses Whisper for audio transcription and GPT-4 to generate hints/answers. The UI is intentionally minimal to allow for discreet use during a video call.<p>It was fun dipping into the world of LLMs, prompt chaining, etc. I didn't find a Swift wrapper for whisper.cpp, so in the repo there's also a barebones Swift framework that wraps whisper.cpp and is designed for real-time transcription on M1/M2.<p>I'll be around if anyone has questions or comments!
Show HN: Ermine.ai – Record and transcribe speech, 100% client-side (WASM)
Show HN: We are building an open-source IDE powered by AI
Show HN: Hocus – self-hosted alternative to GitHub Codespaces using Firecracker
Show HN: Unknown Pleasures, a tiny web experiment with WebGL
Show HN: Unknown Pleasures, a tiny web experiment with WebGL
Show HN: Coursemate – connect with other self learners
Hey Hacker News!<p>My name is Collin, 18 years old and doing a gap year after finishing high school last year.<p>This was my first real project after starting to learn web development around 5 months ago.<p>I came up with this idea as it was a real pain for me to find other people from my country and especially my age, learning and taking online courses about the same stuff online. Lots of these online courses include their own discord communities and forums, but I still found it very hard to connect with other people in there.<p>Thats why I built Coursemate.<p>I would love to get your feedback on it! :)