Show HN: dreamGPT: What if LLM hallucinations were a feature and not a bug? https://ift.tt/CrxTlfE

Show HN: dreamGPT: What if LLM hallucinations were a feature and not a bug? The first GPT-based solution that uses hallucinations from LLMs for divergent thinking to generate new and novel ideas. Hallucinations are often seen as a negative thing, but what if they could be used for our advantage? dreamGPT is here to show you how. The goal of dreamGPT is to explore as many possibilities as possible, as opposed to most other GPT-based solutions which are focused on solving specific problems. https://ift.tt/bhnrUHM May 16, 2023 at 04:32AM

Show HN: Legend-State 1.0 – The fastest React state library https://ift.tt/9WMsVnd

Show HN: Legend-State 1.0 – The fastest React state library After almost a year of development and iterating, we just released Legend-State 1.0. It's the fastest React state library and is very easy to use, based on Observables (Signals) with fine-grained reactivity and built-in persistence. I'd love to know what you think, and I'm also happy to answer any general JavaScript performance questions if you want since I've gone very deep into optimizing . https://ift.tt/PsJZX1A https://ift.tt/YIBUwz0 May 16, 2023 at 04:36AM

Show HN: Hat-syslog – Syslog Server with real time web UI https://ift.tt/9WMTgbf

Show HN: Hat-syslog – Syslog Server with real time web UI https://ift.tt/Jxb5RBO May 16, 2023 at 03:35AM

Show HN: Openlayer – Test, fix, and improve your ML models https://ift.tt/zn6tUXa

Show HN: Openlayer – Test, fix, and improve your ML models Hey HN, my name is Vikas, and my cofounders Rish, Gabe and I are building Openlayer: http://openlayer.com/ Openlayer is an ML testing, evaluation, and observability platform designed to help teams pinpoint and resolve issues in their models. We were ML engineers experiencing the struggle that goes into properly evaluating models, making them robust to the myriad of unexpected edge cases they encounter in production, and understanding the reasons behind their mistakes. It was like playing an endless game of whack-a-mole with Jupyter notebooks and CSV files — fix one issue and another pops up. This shouldn’t be the case. Error analysis is vital to establishing guardrails for AI and ensuring fairness across model predictions. Traditional software testing platforms are designed for deterministic systems, where a given input produces an expected output. Since ML models are probabilistic, testing them reliably has been a challenge. What sets Openlayer apart from other companies in the space is our end-to-end approach to tackling both pre- and post-deployment stages of the ML pipeline. This "shift-left" approach emphasizes the importance of thorough validation before you ship, rather than relying solely on monitoring after you deploy. Having a strong evaluation process pre-ship means fewer bugs for your users, shorter and more efficient dev-cycles, and lower chances of getting into a PR disaster or having to recall a model. Openlayer provides ML teams and individuals with a suite of powerful tools to understand models and data beyond your typical metrics. The platform offers insights about the quality of your training and validation sets, the performance of your model across subpopulations of your data, and much more. Each of these insights can be turned into a “goal.” As you commit new versions of your models and data, you can see how your model progresses towards these goals, as you guard against regressions you may have otherwise not picked up on and continually raise the bar. Here's a quick rundown of the Openlayer workflow: 1. Add a hook in your training / data ingestion pipeline to upload your data and model predictions to Openlayer via our API 2. Explore insights about your models and data and create goals around them [1] 3. Diagnose issues with the help of our platform, using powerful tools like explainability (e.g. SHAP values) to get actionable recommendations on how to improve 4. Track the progress over time towards your goals with our UI and API and create new ones to keep improving We've got a free sandbox for you to try out the platform today! You can sign up here: https://ift.tt/JVA2Rml . We are also soon adding support for even more ML tasks, so please reach out if your use case is not supported and we can add you to a waitlist. Give Openlayer a spin and join us in revolutionizing ML development for greater efficiency and success. Let us know what you think, or if you have any questions about Openlayer or model evaluation in general. [1] A quick run-down of the categories of goals you can track: - Integrity goals measure the quality of your validation and training sets - Consistency goals guard against drift between your datasets - Performance goals evaluate your model's performance across subpopulations of the data - Robustness goals stress-test your model using synthetic data to uncover edge cases - Fairness goals help you understand biases in your model on sensitive populations https://ift.tt/9eST8RW May 15, 2023 at 11:05PM

Show HN: Sha2git brings code hosting to secure SHA-2 Git repositories https://ift.tt/9b1XJNk

Show HN: Sha2git brings code hosting to secure SHA-2 Git repositories https://sha2git.com/ May 15, 2023 at 08:44AM

Show HN: Terminal Notifications for Long Processes over Slack and Discord -Nudge https://ift.tt/QDbNEe3

Show HN: Terminal Notifications for Long Processes over Slack and Discord -Nudge Hi HN! Nudge is an effortless notifier for long-running terminal commands. No prompt needed, you can set it up to automatically notify you for completion of commands running over X minutes over Mac, Slack, and Discord. It can even notify you of commands running over ssh without installing it on the remote host. I built it to notify me of long, monolithic builds at my workplace. Check out Nudge Notifier at NudgeNotifier.com https://ift.tt/OQHUZ96 May 14, 2023 at 11:51PM

Show HN: Online and CLI Tool to backup password protected data with QR codes https://ift.tt/Hqmladn

Show HN: Online and CLI Tool to backup password protected data with QR codes https://ift.tt/yafvNOH May 14, 2023 at 11:47PM

Show HN: Run AWS Cedar Policy Like OPA https://ift.tt/ipejmnt

Show HN: Run AWS Cedar Policy Like OPA https://ift.tt/iP9IYk7 May 15, 2023 at 02:55AM

Show HN: I built my first Cyberdeck https://ift.tt/JHxRcm8

Show HN: I built my first Cyberdeck https://ift.tt/xGBviKP May 15, 2023 at 12:08AM

Show HN: Tack, a fast lightweight scripting language for games and embedding https://ift.tt/1L9ztUa

Show HN: Tack, a fast lightweight scripting language for games and embedding https://ift.tt/i2HoaCt Hi HN! Tack is a scripting language I've been working on sporadically for the past year or so, and intensely for the past few weeks. It originated out of a desire for something that was like Lua, but with a more familiar syntax, and without some of the other surprises in Lua such as the 1-indexed tables. It's also been a great learning project, and a very satisfying challenge! While the current version is early beta at best, I hope to continue working on it and maybe see some adoption. Despite the relative lack of optimization, I'm very pleased with the performance so far - although I haven't done a huge amount of benchmarking, it seems to be significantly faster than the stock Lua 5.4 interpreter for the quicksort test, and the btrees test (copied from the Computer Language Benchmarks Game). The language is designed for embedding in C++ programs, and is written in C++ more or less from scratch including the handwritten recursive descent parser, and a register-based compiler/interpreter. The only dependency other than the standard library is my C++ adaptation of the khash library used for the object type - a from-scratch hashmap seemed not worth the trouble. Quick code example - more examples in the repo! fn quicksort(arr) { const n = #arr if n <= 1 { return arr } " find the midpoint " let l = min(arr) let r = max(arr) if r == l { return arr } const mid = (l + r) / 2 " split array into upper and lower " const upper = filter(arr, fn(x) { return x < mid }) const lower = filter(arr, fn(x) { return x >= mid }) " recursively sort the upper and lower sub-arrays and join the result" return quicksort(lower) + quicksort(upper) } let A = [] for i in 0, 1000000 { A << random() } let before = clock() let B = quicksort(A) let after = clock() print("Time taken: ", after - before, "seconds") Building requires just cmake and a C++20 compiler - tested with MSVC 2022, g++11 on WSL and Clang 15 on M1 https://craftinginterpreters.com was a great help with implementing closures, as I had gone down a blind alley with my first approach for locating the closed-over variables. However I have taken a slightly different approach towards boxing. As I do intend to use this for some small games myself going forward, there is a standard library already, and plans to expand it. I also intend to release a GLFW-based mini game framework along with precompiled binaries, so hobbyists (and younger relatives!) may use it without needing a full compiler toolchain. I would love if anyone is interested enough to try it out! James https://github.com/PlumeCat/tack May 14, 2023 at 11:09PM

Show HN: Torquigen,create symmetrical animated GIFs from your images https://ift.tt/Xg0d1sE

Show HN: Torquigen,create symmetrical animated GIFs from your images This is the first code I've written in WebGL2. It supports Chrome, Firefox, and Safari (macOS or ipadOS). https://torquigen.app May 14, 2023 at 10:09PM

Show HN: ts-npm-template – Template to bootstrap NPM package with TypeScript https://ift.tt/Arpk9LN

Show HN: ts-npm-template – Template to bootstrap NPM package with TypeScript https://ift.tt/6wmqlnp May 14, 2023 at 06:06PM

Show HN: WhyBot, making GPT-4 question itself https://ift.tt/K9CzdNc

Show HN: WhyBot, making GPT-4 question itself Hi HN — we’re John and Vish! We built WhyBot, a tool to help you deeply explore a question or topic. You ask a question, and WhyBot responds by building an ever-expanding knowledge graph. It does this by recursively generating answers and follow-up questions. You can change its persona to change the flavor of the generations (try toddler mode!). We originally built this for the AngelList Agent Hackathon ( https://twitter.com/AqeelMeetsWorld/status/16502799744050421... ) and got a lot of interest from folks asking to play around with it. So we thought it’d be fun to brush it up and release it as a web app! It’s a work in progress and we plan on adding more features, such as saving, sharing, focusing on one branch and potentially executing code. We hope you enjoy playing around with it and would love to hear any of your feedback or thoughts. https://ift.tt/z6XGEpD May 14, 2023 at 09:47AM

Show HN: Bytebase – a GUI-based database schema change tool for developers https://ift.tt/93qtcKH

Show HN: Bytebase – a GUI-based database schema change tool for developers https://ift.tt/BrKJOlP May 13, 2023 at 11:36PM

Show HN: MotivationalSMS – Daily Inspirational Quotes by SMS https://ift.tt/1moeuLE

Show HN: MotivationalSMS – Daily Inspirational Quotes by SMS https://ift.tt/Gdb1rTj May 13, 2023 at 05:31PM

Show HN: React 18.2 and Express 4.18 TypeScript SSR ESM stack https://ift.tt/HXKUnRF

Show HN: React 18.2 and Express 4.18 TypeScript SSR ESM stack https://ift.tt/K6zrPGj May 13, 2023 at 12:55PM

Show HN: Kaizen, music updated over time like software https://ift.tt/eavSJip

Show HN: Kaizen, music updated over time like software Hi, I'm co-founder and CTO of Kaizen. The project started as just a weekend project with me and a music producer friend. It has since grown into a community of artists looking to share their music more frequently and engage with their fans more consistently. We've just launched on ProductHunt: https://ift.tt/TL34fN9 We would love any thoughts or feedback! https://kaizen.place/ May 13, 2023 at 01:16PM

Show HN: A game about guessing which YT video is the most popular https://ift.tt/0XrFngb

Show HN: A game about guessing which YT video is the most popular https://ift.tt/bQ89xGf May 13, 2023 at 06:36AM

Show HN: React.js LLM Agent (open-source) https://ift.tt/IrqWsKa

Show HN: React.js LLM Agent (open-source) I've been working in the couple of months on an experiment, trying to make GPT-4 much more useful for web development / React, writing production code that is relevant to any repository without copy pasta from ChatGPT or having small snippets of auto-complete from Copilot that are not in your context. The agent is taking a user story text and generating and composing multiple react components to generate the relevant screens, based on atomic design principles, with Typescript, TailwindCSS and RadixUI. Is is still experimental but very interesting results, I would like to get your feedback on it! It is completely open-sourced, looking for contributors! https://ift.tt/CeFmGPM May 12, 2023 at 03:06PM

Show HN: Infinity Whiteboard, Designed for Teachers https://ift.tt/jGo8DSL

Show HN: Infinity Whiteboard, Designed for Teachers I've created a whiteboard which I use every lesson when teaching maths, though it can be used for anything. It currently has a few hundred teachers using it daily. It's designed for use with touch-screen interactive whiteboards in classrooms, and stays in sync with your phone/tablet/whatever without signup/login. You can also find me on Twitter where I post updates etc: https://twitter.com/jakegmaths Some features and cool things: * Sync devices without signup - offline by default, just hit 'sync devices' and use the same code on multiple devices to sync * Touch-first - 1 finger draws; 2 finger pan/zoom; 3 finger gestures like changing pen colour * Add images - when teaching, this is usually photos of student work taken on my phone and auto-sync'd to the whiteboard at the front of the room * Add PDFs - when teaching, these are usually past paper exams which I then annotate over with the class * Zen mode - 3 finger tap or hit the ∞ icon to hide the UI; something I use every lesson so students can focus on the actual maths (there's also a fullscreen button when not on iOS) * Visualiser - often when teaching we'll work on paper with a webcam aka 'visualiser' pointing down at it; this projects that to the main whiteboard, with optional cropping, freeze-frame and snapshots * Screencast - many teachers use eg PowerPoint to teach; instead, I'll use PowerPoint in edit/design mode rather than slideshow mode, with a locally-cast cropped portion of that on the main whiteboard at the front of the room. This enables me to eg edit my PowerPoint as I go and use all the PowerPoint tools not available in slideshow mode * Instant replay - hit the play button to play back all the scribblings currently showing on the screen * Magnet mode - when sync'd with another device, use the magnet icon so the other device follows you. Most of my teaching is now via a tablet-with-stylus anywhere in the room, and as I pan/zoom around with the tablet the main whiteboard comes with me... but only when I want it to by activating the magnet * Student mini-whiteboards (MWBs) - if my students have devices and I want them to use them, I 'sync devices' then enable student MWBs and each student has a live copy of the whiteboard, and I can see what they write and can showcase any student instantly on the main board * PWA support - install as a PWA and you can download whiteboards as .iwb files which can then be double-clicked to open/edit on desktop * Free - I have no plans to charge for this Other things you may find interesting from a tech perspective: * The client is a single <5,000 lines HTML file, with JS, CSS, SVG-favicon all inlined (plus PDF.js lazily loaded if you add a PDF) * This is vanilla Javascript with no frameworks or libraries (except PDF.js) and no minification or build scripts - just view-source and check out how ugly all my code is! * 77.6kB for everything (except PDF support)... the size of 'modern' websites frankly disgusts me * The server is just a single ~500 line Javascript file and runs on Deno (also ported to Bun but unstable for now) and really just serves some static files, deals with websockets and temporarily stores images people add * Costs ~£5/month on Heroku * There's no database or any long-term persistence - Heroku servers restart every 24h and nothing is saved beyond that; it's all ephemeral https://www.mathsuniverse.com/infinity May 12, 2023 at 06:56PM