Show HN: Phind.com – Generative AI search engine for developers https://ift.tt/rEeX2AD

Show HN: Phind.com – Generative AI search engine for developers Hi HN, Today we're launching 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. It's version 1.0 of what was previously known as Hello (beta.sayhello.so) and has been completely reworked to be more accurate and reliable. Because it's connected to the internet, Phind is always up-to-date and has access to docs, issues, and bugs that ChatGPT hasn't seen. Like ChatGPT, you can ask followup questions. Phind is smart enough to perform a new search and join it with the existing conversation context. We're merging the best of ChatGPT with the best of Google. You're probably wondering how it's different from the new Bing. For one, we don't dumb down a user's query the way that the new Bing does. We feed your question into the model exactly as it was asked, and are laser-focused on providing developers the most detailed and comprehensive explanations to code-related questions. Secondly, we've focused the model on providing answers instead of chatbot small talk. This is one of the major improvements we've made since exiting beta. Phind has the creative abilities to generate code, write essays, and even compose some poems/raps but isn't interested in having a conversation for conversation's sake. It should refuse to state its own opinion and rather provide a comprehensive summary of what it found online. When it isn't sure, it's designed to say so. It's not perfect yet, and misinterprets answers ~5% of the time. An example of Phind's adversarial question answering ability is https://ift.tt/4XzVtTK... . ChatGPT became useful by learning to generate answers it thinks humans will find helpful, via a technique called Reinforcement Learning from Human Feedback (RLHF). In RLHF, a model generates multiple candidate answers for a given question and a human rates which one is better. The comparison data is then fed back into the model through an algorithm such as PPO. To improve answer quality, we're deploying RLAIF — an improvement over RLHF where the AI itself generates comparison data instead of humans. Generative LLMs have already reached the point where they can review the quality of their own answers as good or better than an average human rater tasked with annotating data for RLHF. We still have a long way to go, but Phind is state-of-the-art at answering complex technical questions and writing intricate guides all while citing its sources. We'd love to hear your feedback. Examples: https://ift.tt/ABFDYxo... https://ift.tt/7BCRrv1... https://ift.tt/Fx2t5IM https://ift.tt/ZEhPQ76... https://ift.tt/XtfPuJn... Discord: https://ift.tt/glAKCrE https://phind.com February 21, 2023 at 11:26PM

Show HN: Planlike.pro – New Estimating Tool https://ift.tt/naAQji7

Show HN: Planlike.pro – New Estimating Tool https://planlike.pro/ February 21, 2023 at 05:24PM

Show HN: Small TypeScript library to work with quadkeys in a fast way https://ift.tt/u2pst4e

Show HN: Small TypeScript library to work with quadkeys in a fast way I am developing a website called Geocode Map Viewer( https://ift.tt/ENRcaru ). I was looking for a suitable TypeScript library to visualize Quadkeys on the map, but unfortunately I couldn't find one. So I decided to develop my own library, using the sample code available on the Microsoft Tile Maps page as a reference. https://ift.tt/h7qJVkE February 21, 2023 at 06:51AM

Show HN: Gargantuan Takeout Rocket – Google Takeout Transloader to Azure https://ift.tt/s0CKkwD

Show HN: Gargantuan Takeout Rocket – Google Takeout Transloader to Azure Been broken for 4 months, just got back to fixing it and validating. Figured I'll repost this. Gargantuan Takeout Rocket (GTR) is a toolkit to make the pain of backing up a Google account to somewhere that's not Google a lot less. At the moment the only destination supported is Azure. It's a guide, a browser extension, a Cloudflare worker to deploy, and Azure storage to configure. This sounds like buzzword creep, but believe me, every piece is extremely important. It's very cheap to run/serverless. You can backup a Google account at about $1/TB. Compared to renting a VPS to do this, it's much more pleasant. You aren't juggling strange URLs, needing big beefy boxes to buffer large data, or trying to login to Google or pass URLs through a VPS. Unfortunately, not everything about the procedure can be automated. But whatever can be, is. It's very fast. 1GB/s is the stable default and recommended speed. However, you can have about 3 of these going at a time for about 3GB/s+ overall. This trick is accomplished by making Azure download from Google to a file block, a unique API not seen in S3 or S3-like object storage. Unfortunately, Azure has URL handling bugs and only supports HTTP 1.1, greatly limiting parallelism. We can use Cloudflare Workers to work around these issues. I use GTR myself with a scheduled Google Takeout every two months to backup 1.5TB of data from Google. This can be photos, YouTube videos, etc. I can finish my backups to safe non-Google storage in 15 minutes after I get an email from Google that my Takeout is ready to be downloaded. Unfortunately the only destination is currently Azure. There's also no encryption support. And also Cloudflare is involved. That said, if you're fine with this, this is a fine way to backup a Google and Youtube account as-is. https://ift.tt/GPWehlB February 21, 2023 at 11:26AM

Show HN: My50sTV – Nostalgic TV Simulator https://ift.tt/8EBsyNi

Show HN: My50sTV – Nostalgic TV Simulator https://www.my50stv.com February 21, 2023 at 09:10AM

Show HN: Turn Your Pandas Dataframe into a Tableau-Style UI for Visual Analysis https://ift.tt/HQgEckF

Show HN: Turn Your Pandas Dataframe into a Tableau-Style UI for Visual Analysis Hey, guys. I've just made a plugin which turns your pandas dataframe into a tableau-style component. It allows you to explore the dataframe with easy drag-and-drop UI. You can use PyGWalker in Jupyter, Google Colab, or even Kaggle Notebook to easily explore your data and generate interactive visualizations. PyGWalker (pronounced like "Pig Walker", just for fun) is named as an abbreviation of "Python binding of Graphic Walker". Here are some links to check it out: The Github Repo: https://ift.tt/CebqFVy Use PyGWalker in Kaggle: https://ift.tt/Om9idtH Feedback and suggestions are appreciated! Please feel free to try it out and let me know what you think. Thanks for your support! https://ift.tt/CebqFVy February 20, 2023 at 09:20PM

Show HN: Whisper.cpp and YAKE to Analyse Voice Reflections [iOS] https://ift.tt/cX5qgOI

Show HN: Whisper.cpp and YAKE to Analyse Voice Reflections [iOS] Six months ago, I went full-time indie, but I haven't released anything so far. The products just never felt good enough for me to publicly say this is what I'm doing now. To get out of this mindset, I decided to make an app for myself in a week, add monetization, release it and move on. The app idea was simple: Reflect on your day by answering the same four questions out loud. The answers are transcribed and with regular use you can see what influences you the most and take action. All on-device, as otherwise I wouldn't feel comfortable sharing my thoughts. I had all core features working within a day by simply modifying an existing example app. However I was dissatisfied with iOS's built-in offline transcription due to a lack of punctuation and the speech recognition permission prompt that made it seem like data would leave the device. Decided to use whisper.cpp [0] (small model) instead. This change, lead to many others, as I now felt too little of the app's code was mine. e.g.: - Added automatic mood analysis. First using sentiment analysis, then changed to a statistical approach - Show trends: First implemented TextRank to provide a summary for an individual day, then changed it to extract keywords to spot trends over weeks and months. Replaced TextRank with KeyBERT for speed and n-grams, then BERT-SQuAD, and ended on a modified YAKE [1] for subjectively better results. (Do you know of a better approach?) As a result, this tiny app took me over a month, but it still has its flaws: - Transcription is not live but performed on recordings, so if you immediately want the transcript of your most recent answer, you have to wait. - Mood and keyphrase extraction are optimized for my languages and way of speaking, so they might not generalize well. - Music in the background can result in nearly empty transcripts. Nevertheless, after using the app regularly and enjoying it, I feel ready to release. Hope you will find the app useful too. [0] Show HN: Whisper.cpp https://ift.tt/K9urjOY [1] YAKE: https://ift.tt/1Mg4eFn https://ift.tt/4FlPIcK February 20, 2023 at 08:38PM