Show HN: DoubleMemory – more efficient local-first read-it-later app https://ift.tt/z7yNM8F

Show HN: DoubleMemory – more efficient local-first read-it-later app DoubleMemory started as an experiment to see if I can somehow automatically save all double cmd + c, as I often do instinctively, so I don't need extensions to save links and text into an app, and avoiding flooding the capture history as regular clipboard managers does. My motivation was not to create a read-it-later app, yet it evolved into this unique yet cohesive form of a read-it-later + bookmarking organizer + clipboard manager + card based note-taking app over the last 6 months. It also launches from the menu bar with a shortcut and navigates with keyboard shortcuts. My favorite part is instead of rendering a list of article titles, everything is rendered as pretty preview cards in a translucent Pinterest-like mood board. It also has a nifty iOS app, that will allow you to swipe with your thumbs between articles just like on iOS Safari... Now that Pocket is closing, this is after Instapaper going back to indie and Omnivore and UpNext and numerous others closing over the years. All of these are cloud-hosted services, which got me reflecting: maybe this local-first architecture would be well positioned to build in this space. Here is my not-so-scientific comparison: ## Domain $10 vs $1M = 100,000x difference. ## Server running cost No servers other than what's running by iCloud vs $1M per year = 1mX difference ## Platforms Apple only (mac + iphone + ipad) vs Multi platforms (windows, linux, android also supported) = 20X maintenance cost difference ## Capturing No browser extensions required v.s. maintain all extensions for various browsers and extension stores = 5x difference ## Architecture App receives the link, Apple generates the rich preview cards for thousands of different types of links, app caches these preview cards. vs. Someone write some custom code for each link type or with Open Graph, one designer created one generic card that works for all links. = 100x cost difference. I know, Apple is coming for clipboards with more restrictions, which is basically a shared global state on Mac systems, DoubleMemory does also support other ways to capture: drag-n-drop to app/menubar icon/app icon, right click->Services menu, or Share sheet. We will add more auto-importers. Also vibe coded some importers for Pocket, Omnivore and ReadWise here: https://ift.tt/1dRAGHk Everything in the app is free with no limits. Capturing is really step 0. You giving us a chance to save your content, doesn't mean you are getting any values out of it (ain't that the typical story of read-it-later apps? save-it and never-read-it). the eventual goal is to easily retrieve these content, and eventually consuming them. I hope to eventually launch paid features that aligns with these value generating workflows. App Store link: https://ift.tt/5Lc0rxt Let me know what you think... https://ift.tt/gSkM8BZ May 24, 2025 at 12:25AM

Show HN: hcker.news – an ergonomic, timeline-based Hacker News front page https://ift.tt/WJAdpPl

Show HN: hcker.news – an ergonomic, timeline-based Hacker News front page Hi folks, I've built an alternative Hacker News front page. It is inspired by and meant to be a replacement for hckrnews.com. I built this because HN is woefully underfeatured, but most sites that try to improve it seem to assume that the visual design is the problem. hcker.news tries to maintain HN's familiarity while adding useful enhancements. There are three primary views: - Timeline View: Browse top stories by votes or comments grouped by day, week, or month (e.g., top 20 per day, top 100 per week). - Aggregate View: See top stories by votes or comments over custom time ranges. - Front Page View: The original HN front page, untouched. Feed Filtering: - Custom Keyword Filters: Include/exclude keywords (e.g., include "Rust," exclude "DOGE") or set a minimum score threshold. - No HN Algorithm: Timeline and Aggregate Views show stories usually downranked by the HN algo (e.g., flagged posts or those with too many comments). UI: - Unread Flags: Quickly spot new stories or ones you haven't seen. - Two Layouts: Classic HN style or a compact story view inspired by hckrnews.com. - Multi-column & High-density Modes: Fit more content on screen. - Themes: Light, Dark, and Manila. I'd love your feedback and suggestions. Cheers! https://hcker.news May 24, 2025 at 12:14AM

Show HN: I made an infinite gallery of AI-generated 3D skeuomorphic icons https://ift.tt/LfFc6yB

Show HN: I made an infinite gallery of AI-generated 3D skeuomorphic icons https://ift.tt/bR7rgP2 May 23, 2025 at 11:22PM

Show HN: Free Text-to-Video for Learning Anything(Inspired by 3Blue1Brown) https://ift.tt/Q6MuGC0

Show HN: Free Text-to-Video for Learning Anything(Inspired by 3Blue1Brown) I'm a huge fan of 3B1B and how he creates appealing and easy-to-understand videos. But he doesn't have a video for every single topic. Whenever I needed help in math or physics, I would try to watch his videos but the issue is that are just not specific enough to my content or curriculum. This issue applies to every single video explanation out there, they just aren't personalized. Most educational videos explain general topics, but they don’t align perfectly with the specific question I have or the way I need it explained. That’s the gap I wanted to fill. So I built a tool that generates high-quality, visually engaging explainer videos that are tailored exactly to the question you ask. Whether it's a niche math problem, a concept from your physics class, or something your textbook didn't explain well, this tool creates a custom explanation in the style of channels like 3B1B, but made just for you. The tool is free to use for some time. Me and my cofounder have dedicated a portion of our savings to this project and unless we get external funding in the near future, we would have to add a paid tier for the product or completely shut it down. Also, would love it if you show some support at our discord server. Thanks for your time! The tool is free to use for some time. Me and my cofounder have dedicated a portion of our savings to this project and unless we get external funding in the near future, we would have to add a paid tier for the product or completely shut it down. You can try it out yourself here --> https://trytorial.com/ . Also, would love it if you show some support at our discord server. Thanks for your time! https://trytorial.com/ May 22, 2025 at 11:30PM

Show HN: Pi Co-pilot – Evaluation of AI apps made easy https://ift.tt/Ne1btiS

Show HN: Pi Co-pilot – Evaluation of AI apps made easy Hey HN — 2 months ago we shared our first product with the HN community ( https://ift.tt/OBK5GqH ). Despite receiving lots of traffic from HN, we didn’t see any traction or retention. One of our major takeaways was that our product was too complicated. So we’ve spent the last 2 months iterating towards a much more focused product that tries to do just one thing really well. Today, we’d like to share our second launch with HN. Our original idea was to help software engineers build high-quality LLM applications by integrating their domain knowledge into a scoring system, which could then drive everything from prompt tuning to fine-tuning, RL, and data filtering. But what we quickly learned (with the help of HN – thank you!) is that most people aren’t optimizing as their first, second, or even third step — they’re just trying to ship something reasonable using system prompts and off-the-shelf models. In looking to build a product that’s useful to a wider audience, we found one piece of the original product that most people _did_ notice and want: the ability to check that the outputs of their AI apps look good. Whether you’re tweaking a prompt, switching models, or just testing a feature, you still need a way to catch regressions and evaluate your changes. Beyond basic correctness, developers also wanted to measure more subtle qualities — like whether a response feels friendly. So we rebuilt the product around this single use case: helping developers define and apply subjective, nuanced evals to their LLM outputs. We call it Pi Co-pilot. You can start with any/all of the below: - a few good/bad examples - a system prompt, or app description - an old eval prompt you wrote The co-pilot helps you turn that into a scoring spec — a set of ~10–20 concrete questions that probe the output against dimensions of quality you care about (e.g. “is it verbose?”, “does it have a professional tone?”, etc). For each question, it selects either: - a fast encoder-based model (trained for scoring) – Pi scorer. See our original post [1] for more details on why this is a good fit for scoring compared to the “LLM as a judge” pattern. - or generates Python functions when that makes more sense (word count, regex etc.) You iterate over examples, tweak questions, adjust scoring behavior, and quickly reach a spec that reflects your actual taste — not some generic benchmark or off-the-shelf metrics. Then you can plug the scoring system into your own workflow: Python, TypeScript, Promptfoo, Langfuse, Spreadsheets, whatever. We provide easy integrations with these systems. We took inspiration from tools like v0 and Bolt: natural language on the left, structured artifacts on the right. That pattern felt intuitive — explore conversationally, and let the underlying system crystallize it into things you can inspect and use (scoring spec, examples and code). Here is a loom demo of this: https://ift.tt/fBzjRbD We’d appreciate feedback from the community on whether this second iteration of our product feels more useful. We are offering $10 of free credits (about 25M input tokens), so you can try out the Pi co-pilot for your use-cases. No sign-in required to start exploring: https://withpi.ai Overall stack: Co-pilot next.js and Vercel on GCP. Models: 4o on Azure, fine tuned Llama & ModernBert on GCP. Training: Runpod and SFCompute. – Achint (co-founder, Pi Labs) https://withpi.ai/ May 22, 2025 at 06:01PM

Show HN: Super (YC W18) - Turn company data into answers & agents for your team https://ift.tt/HSGRO0E

Show HN: Super (YC W18) - Turn company data into answers & agents for your team Hey there, Chris here We're known for our straightforward yet powerful Knowledge Base, Slite(YCW18).We launched our AI-powered search in Feb 2023 and after getting great response and usage, we dove deeper into solving the challenge of knowledge retrieval in daily work. That's why we're now launching our second major product, Super( https://www.super.work ). Super seamlessly connects your existing tools, providing accurate answers, streamlined workflows, automated digests, and much more. You might wonder: Why not just link your apps together using something like an MCP? The problem is that MCPs can't handle complex knowledge retrieval effectively. MCPs are basically LLMs equipped with API toolbelts. If you've ever tried asking a complicated question through an MCP, one that needs data from multiple different tools, you've likely faced frustrating delays. MCPs slowly make API calls one after another, causing long waits while they collect data from each endpoint. By contrast, Super quickly searches through all the data that actually matters from all of your tools simultaneously. This means you'll get your accurate answer in seconds, not minutes. The limitations of MCP-based solutions become clear when you try to deploy them reliably within a team. They either won't index your critical content effectively, won't do it fast enough, or won't cover all your tools at once. Properly chunking, embedding, querying, and filtering data from various sources is still essential. MCPs triggering APIs can't match this integrated approach for speed and accuracy. Moreover, Super understands the value of running multiple tasks simultaneously through LLMs. For example, one step may involve identifying search filters, while another simultaneously uses an LLM to aggregate and refine information. This parallel process quickly shapes the final, accurate answer for users. Additionally, MCPs aren't designed for enterprise-grade use. Businesses need standardized experiences, fine-grained user permissions, and consistent access controls across multiple tools. Super addresses these requirements by indexing data beforehand while still respecting each user's access permissions. Super offers: - Perplexity-like search experience on your team data - A growing selection of integrations with popular data sources - Customizable AI assistants tailored to your specific needs - An extension to embed Super directly into external websites you're already using - A clear path for your company to adopt AI strategically, rather than letting individual employees scatter across different, incompatible tools. And of course... It does comes with its MCP, which makes your agentic workflows actually able to properly tap on your data. Here's a quick video showing Super in action: https://www.youtube.com/watch?v=L5A6BRW90K4 Have you hit such walls with standard MCPs? Have you try building your own solutions? https://super.work May 21, 2025 at 07:48PM

Show HN: Appwrite Sites – the open-source vercel alternative https://ift.tt/fYNxHGs

Show HN: Appwrite Sites – the open-source vercel alternative https://ift.tt/TvhIZg7 May 19, 2025 at 05:53PM