The best Hacker News stories from Show from the past day
Latest posts:
Show HN: Hacky Meta Glasses GPT4 Vision Integration
Super hacky implementation due to the lack of an SDK. Fun project though.<p>In the foodlog demonstration I just made a fake fb account (sorry zucc) called "Mye Food-Log".
Show HN: Hacky Meta Glasses GPT4 Vision Integration
Super hacky implementation due to the lack of an SDK. Fun project though.<p>In the foodlog demonstration I just made a fake fb account (sorry zucc) called "Mye Food-Log".
Show HN: Hacky Meta Glasses GPT4 Vision Integration
Super hacky implementation due to the lack of an SDK. Fun project though.<p>In the foodlog demonstration I just made a fake fb account (sorry zucc) called "Mye Food-Log".
Show HN: Dobb·E – towards home robots with an open-source platform
Hi HN! Proud to share our open-source robot platform, Dobb·E, a home robot system that needs just 5 minutes of human teaching to learn new tasks. We've already taken Dobb·E to 10 different homes in New York, taught it 100+ tasks, and we are just getting started! I would love to hear your thoughts about this.<p>Here are some more details, below (or see a Twitter thread with attached media: <a href="https://twitter.com/i/status/1729515379892826211" rel="nofollow noreferrer">https://twitter.com/i/status/1729515379892826211</a> or <a href="https://nitter.net/i/status/1729515379892826211" rel="nofollow noreferrer">https://nitter.net/i/status/1729515379892826211</a>):<p>We engineered Dobb·E to maximize efficiency, safety, and user comfort. As a system, it is composed of four parts: a data collection tool, a home dataset, a pretrained vision model, and a policy fine-tuning recipe.<p>We teach our robots with imitation learning, and for data collection, we created the “Stick”, a tool made out of $25 of hardware and an iPhone.<p>Then, using the Stick, we collected a 13 hour dataset in 22 New York homes, called Homes of New York (HoNY). HoNY has 1.5M frames collected over 216 different "environments" which is an order of magnitude larger compared to similar open source datasets.<p>Then we trained a foundational vision model that we can fine-tune fast (15 minutes!) on a new task with only 5 minutes (human time)/ 90 seconds (demo time) of data. So from start to finish, it takes about 20 minutes to teach the robot a new task.<p>Over a month, we visited 10 homes, tried 109 tasks, and got 81% success rate in simple household tasks. We also found a line of challenges, from mirrors to heavy objects, that we must overcome if we are to get a general purpose home robot.<p>We open-sourced our entire system because our primary goal is to get more robotics and AI researchers, engineers, and enthusiasts to go beyond constrained lab environments and start getting into homes!<p>So here is how you can get started:<p>1. Code and STL files: <a href="https://github.com/notmahi/dobb-e/">https://github.com/notmahi/dobb-e/</a><p>2. Technical documentation: <a href="https://docs.dobb-e.com/" rel="nofollow noreferrer">https://docs.dobb-e.com/</a><p>3. Paper: <a href="https://arxiv.org/abs/2311.16098" rel="nofollow noreferrer">https://arxiv.org/abs/2311.16098</a><p>4. More videos and the dataset: <a href="https://dobb-e.com" rel="nofollow noreferrer">https://dobb-e.com</a><p>5. Robot we used: <a href="https://hello-robot.com" rel="nofollow noreferrer">https://hello-robot.com</a>
Show HN: Dobb·E – towards home robots with an open-source platform
Hi HN! Proud to share our open-source robot platform, Dobb·E, a home robot system that needs just 5 minutes of human teaching to learn new tasks. We've already taken Dobb·E to 10 different homes in New York, taught it 100+ tasks, and we are just getting started! I would love to hear your thoughts about this.<p>Here are some more details, below (or see a Twitter thread with attached media: <a href="https://twitter.com/i/status/1729515379892826211" rel="nofollow noreferrer">https://twitter.com/i/status/1729515379892826211</a> or <a href="https://nitter.net/i/status/1729515379892826211" rel="nofollow noreferrer">https://nitter.net/i/status/1729515379892826211</a>):<p>We engineered Dobb·E to maximize efficiency, safety, and user comfort. As a system, it is composed of four parts: a data collection tool, a home dataset, a pretrained vision model, and a policy fine-tuning recipe.<p>We teach our robots with imitation learning, and for data collection, we created the “Stick”, a tool made out of $25 of hardware and an iPhone.<p>Then, using the Stick, we collected a 13 hour dataset in 22 New York homes, called Homes of New York (HoNY). HoNY has 1.5M frames collected over 216 different "environments" which is an order of magnitude larger compared to similar open source datasets.<p>Then we trained a foundational vision model that we can fine-tune fast (15 minutes!) on a new task with only 5 minutes (human time)/ 90 seconds (demo time) of data. So from start to finish, it takes about 20 minutes to teach the robot a new task.<p>Over a month, we visited 10 homes, tried 109 tasks, and got 81% success rate in simple household tasks. We also found a line of challenges, from mirrors to heavy objects, that we must overcome if we are to get a general purpose home robot.<p>We open-sourced our entire system because our primary goal is to get more robotics and AI researchers, engineers, and enthusiasts to go beyond constrained lab environments and start getting into homes!<p>So here is how you can get started:<p>1. Code and STL files: <a href="https://github.com/notmahi/dobb-e/">https://github.com/notmahi/dobb-e/</a><p>2. Technical documentation: <a href="https://docs.dobb-e.com/" rel="nofollow noreferrer">https://docs.dobb-e.com/</a><p>3. Paper: <a href="https://arxiv.org/abs/2311.16098" rel="nofollow noreferrer">https://arxiv.org/abs/2311.16098</a><p>4. More videos and the dataset: <a href="https://dobb-e.com" rel="nofollow noreferrer">https://dobb-e.com</a><p>5. Robot we used: <a href="https://hello-robot.com" rel="nofollow noreferrer">https://hello-robot.com</a>
Show HN: Dobb·E – towards home robots with an open-source platform
Hi HN! Proud to share our open-source robot platform, Dobb·E, a home robot system that needs just 5 minutes of human teaching to learn new tasks. We've already taken Dobb·E to 10 different homes in New York, taught it 100+ tasks, and we are just getting started! I would love to hear your thoughts about this.<p>Here are some more details, below (or see a Twitter thread with attached media: <a href="https://twitter.com/i/status/1729515379892826211" rel="nofollow noreferrer">https://twitter.com/i/status/1729515379892826211</a> or <a href="https://nitter.net/i/status/1729515379892826211" rel="nofollow noreferrer">https://nitter.net/i/status/1729515379892826211</a>):<p>We engineered Dobb·E to maximize efficiency, safety, and user comfort. As a system, it is composed of four parts: a data collection tool, a home dataset, a pretrained vision model, and a policy fine-tuning recipe.<p>We teach our robots with imitation learning, and for data collection, we created the “Stick”, a tool made out of $25 of hardware and an iPhone.<p>Then, using the Stick, we collected a 13 hour dataset in 22 New York homes, called Homes of New York (HoNY). HoNY has 1.5M frames collected over 216 different "environments" which is an order of magnitude larger compared to similar open source datasets.<p>Then we trained a foundational vision model that we can fine-tune fast (15 minutes!) on a new task with only 5 minutes (human time)/ 90 seconds (demo time) of data. So from start to finish, it takes about 20 minutes to teach the robot a new task.<p>Over a month, we visited 10 homes, tried 109 tasks, and got 81% success rate in simple household tasks. We also found a line of challenges, from mirrors to heavy objects, that we must overcome if we are to get a general purpose home robot.<p>We open-sourced our entire system because our primary goal is to get more robotics and AI researchers, engineers, and enthusiasts to go beyond constrained lab environments and start getting into homes!<p>So here is how you can get started:<p>1. Code and STL files: <a href="https://github.com/notmahi/dobb-e/">https://github.com/notmahi/dobb-e/</a><p>2. Technical documentation: <a href="https://docs.dobb-e.com/" rel="nofollow noreferrer">https://docs.dobb-e.com/</a><p>3. Paper: <a href="https://arxiv.org/abs/2311.16098" rel="nofollow noreferrer">https://arxiv.org/abs/2311.16098</a><p>4. More videos and the dataset: <a href="https://dobb-e.com" rel="nofollow noreferrer">https://dobb-e.com</a><p>5. Robot we used: <a href="https://hello-robot.com" rel="nofollow noreferrer">https://hello-robot.com</a>
Show HN: Spaceflight News made using Htmx
Show HN: Multiple – Load test any API with JavaScript and NPM packages
Hey HN,<p>I wanted a better load testing solution – so I built one with my team at Multiple. We just opened early access and would love to get your feedback.<p>We created Multiple to solve three challenges with existing tools:<p>1. Limited scripting capabilities. XML or GUI-based scripting can only test basic scenarios. Existing code-based tools struggle with auth, generating synthetic data, and testing anything other than HTTP requests. We went with JavaScript for ease of use, versatility, and integration with existing developer workflows.<p>2. Cannot use existing libraries or code. Instead of forcing you to learn a new system and rewrite code, Multiple leverages the JavaScript and NPM ecosystem so you can use packages you're already familiar with. By supporting NPM packages, Multiple can test nearly any API, service, or protocol.<p>3. Tedious infrastructure management. There's no reason to spend time spinning up and configuring machines, and then destroying them after a test. Multiple abstracts that away. You just enter the test size and duration, and press start.<p>My favorite feature we've built so far is the Debug Run. You can use Debug Run as you write your tests to execute a single run-through. It's helpful to verify correct behavior and capture logs, and it allows you to iterate quickly, without spinning up a full load test each time.<p>We have so much in store for developers: pass/fail conditions, CLI, and repo integration, to name a few. Thanks for reading, and let us know what you think.
Show HN: Transform Notes into Visual Mind Maps
Hey, HN Community!<p>I'm a Brazilian software engineer, and my educational journey led to a significant discovery.
During my master's degree, I often struggled to recall foundational concepts studied years earlier, which were crucial for my current studies. This challenge sparked a reevaluation of traditional note-taking and inspired the creation of cmaps.io.<p>Human thinking is inherently associative, not linear. We often draw connections between learned concepts unconsciously. To leverage this natural process, making these connections explicit is essential – and that's precisely what cmaps.io offers.<p>I can personally vouch for the impact of cmaps.io on my learning and retention. It's become an invaluable tool for me, and I'm confident it can be for you as well!<p>You can find more details on the ProductHunt post:
<a href="https://www.producthunt.com/posts/cmaps-io" rel="nofollow noreferrer">https://www.producthunt.com/posts/cmaps-io</a><p>Discover cmaps.io and enjoy a revolutionary note-taking experience!<p>I eagerly await your feedback and thoughts
Show HN: Cap – open-source alternative to Loom
Show HN: Build an open-source computer vision model in seconds using text
Hello HN! I want to share something me and a few friends have been working on for a while now — Zeroshot, a web tool that builds image classifiers using text-image models and autolabeling. What does this mean in practice? You can put together an image classifier in about 30 seconds that’s faster and more accurate than CLIP, but that you can deploy yourself however you’d like. It’s open source, commercially licensed, and doesn’t require you to pay anyone per API call.<p>Here's a 2 minute video that shows it off: <a href="https://www.youtube.com/watch?v=S4R1gtmM-Lo" rel="nofollow noreferrer">https://www.youtube.com/watch?v=S4R1gtmM-Lo</a><p>How/why does it work?
We believe that with the rise of foundation vision models, computer vision will fundamentally change. These powerful models will let any devs “compile” a model ahead of time with a subset of the foundation model’s characteristics, using only text and a web-tool. The days of teams of MLEs building complex models and pipelines are ending.<p>Zeroshot works by using two powerful pre-trained models, CLIP and DINOv2 together. The web-app allows users to quickly create our training sets via text search. Using pre-cached DINOv2 features, we generate a simple linear model that can be trained and deployed without any fine-tuning. Since you can see what’s going into your training set, you can tune your prompts to get the type of performance or detail you want.<p>CLIP Small -- Size: 335 MB, Latency: 35ms<p>CLIP Large -- Size: 891 MB, Latency: 276ms<p>Zeroshot -- Size: 85 MB, Latency: 20ms<p>What’s next?
We wanna see how people use or would use the tool before deciding what to do next. On the list: clients for iOS and NodeJS, speeding up GPU inference times via TensorRT, offering larger Zeroshot models for better accuracy, easier results refining, support for bringing your own data lake, model refinement using GPT-V, we’ve got plenty of ideas.
Show HN: SwaraNotebook – a notation editor for Indian classical music
As an enthusiast of Indian classical music, I needed to write music notations in the traditional typeset format. When I didn't find any existing editors, I developed a Swara Notebook, a mobile focused web app to write North Indian Classical (<i>Hindustani</i>) music notations.<p>The notes (called <i>Sargam</i>, similar to Solfege) can be written in English, Devnagri and Bangla scripts. The transcribed song can be played back in 6 different rhythmic cycles (<i>Taal</i>) to the accompaniment of the <i>Tabla</i>(a type of drum) or a metronome. Here's an example of a transcribed song <a href="https://swaranotebook.com/view/vlB9hVgb5OdKcadbhOOYyzKwvpl2/53219a9192a5-Piyarava-ab-tum" rel="nofollow noreferrer">https://swaranotebook.com/view/vlB9hVgb5OdKcadbhOOYyzKwvpl2/...</a><p>Since North Indian classical music is oriented around <i>Ragas</i> (similar to modes in western classical music), the keyboard adapts to the notes of a <i>Raga</i>, making it contextually easier to key in notes.<p>A common question I get: can it handle microtones? I chose to not support microtones and other pitch ornamentation such as glissando (called <i>Meend</i>) since a mobile interface is not the easiest place to add such details.<p>I also find it a useful tool for ear training, by typing out notations to songs I know, and playing it back to know if I guessed the notes right.<p>It is an open source project written in Clojure/script
<a href="https://github.com/Studio-kalavati/bandish-editor">https://github.com/Studio-kalavati/bandish-editor</a>
Show HN: SwaraNotebook – a notation editor for Indian classical music
As an enthusiast of Indian classical music, I needed to write music notations in the traditional typeset format. When I didn't find any existing editors, I developed a Swara Notebook, a mobile focused web app to write North Indian Classical (<i>Hindustani</i>) music notations.<p>The notes (called <i>Sargam</i>, similar to Solfege) can be written in English, Devnagri and Bangla scripts. The transcribed song can be played back in 6 different rhythmic cycles (<i>Taal</i>) to the accompaniment of the <i>Tabla</i>(a type of drum) or a metronome. Here's an example of a transcribed song <a href="https://swaranotebook.com/view/vlB9hVgb5OdKcadbhOOYyzKwvpl2/53219a9192a5-Piyarava-ab-tum" rel="nofollow noreferrer">https://swaranotebook.com/view/vlB9hVgb5OdKcadbhOOYyzKwvpl2/...</a><p>Since North Indian classical music is oriented around <i>Ragas</i> (similar to modes in western classical music), the keyboard adapts to the notes of a <i>Raga</i>, making it contextually easier to key in notes.<p>A common question I get: can it handle microtones? I chose to not support microtones and other pitch ornamentation such as glissando (called <i>Meend</i>) since a mobile interface is not the easiest place to add such details.<p>I also find it a useful tool for ear training, by typing out notations to songs I know, and playing it back to know if I guessed the notes right.<p>It is an open source project written in Clojure/script
<a href="https://github.com/Studio-kalavati/bandish-editor">https://github.com/Studio-kalavati/bandish-editor</a>
Show HN:Draw Fast - Real-time AI image generation based on drawings in a canvas
Show HN: Gitea Cloud: A brand new platform for managed Gitea Instances
Show HN: I built a guided Build your own DNS Server challenge
Hey everyone. It's Sherub here, author of the Build your own DNS Server challenge on CodeCrafters. Currently it’s available in Rust, Go, and Python and is free while in beta.<p><a href="https://codecrafters.io/dns-server">https://codecrafters.io/dns-server</a><p>I've kept the challenge accessible but still challenging for an intermediate developer. This challenge, like others from CodeCrafters, is self-paced. You can use any tools you prefer (terminal, editor, etc.) to build the project.<p>At the end of the challenge, you will have created a DNS forwarding server. The server can create and read DNS packets and respond to DNS queries. As you go, you'll learn about the DNS protocol, its format, servers, and A records. All while getting to hone your language skills.<p>Some of the challenges and choices I had to make:<p>* To make the stages easier, I had to break them up, such that each step builds on the previous one. This was especially challenging for the 2nd stage, where we write a DNS packet's header contents. Even though I’d have liked it to be easier, breaking it up further would have been weird.<p>* Instead of implementing a recursive resolver, I've restricted to a forwarding server. We made this decision so that most developers can still use it. To add more complexity, we can use a challenge extension (noted below).<p>* Deciding how much instruction and context the stages should provide. I’ve decided to keep them as thorough as possible for most of the stages. Developers can choose to have thorough details or just skim through them.<p>I would love your feedback and questions on the challenge. You can try it out for free here: <a href="https://codecrafters.io/dns-server">https://codecrafters.io/dns-server</a> (no CC required).<p>I also have challenge extensions planned. You can find them at <a href="https://app.codecrafters.io/vote/challenge-extension-ideas?course=dns-server">https://app.codecrafters.io/vote/challenge-extension-ideas?c...</a>. I'm also keen to hear what you think about the extension ideas.
Show HN: I built a guided Build your own DNS Server challenge
Hey everyone. It's Sherub here, author of the Build your own DNS Server challenge on CodeCrafters. Currently it’s available in Rust, Go, and Python and is free while in beta.<p><a href="https://codecrafters.io/dns-server">https://codecrafters.io/dns-server</a><p>I've kept the challenge accessible but still challenging for an intermediate developer. This challenge, like others from CodeCrafters, is self-paced. You can use any tools you prefer (terminal, editor, etc.) to build the project.<p>At the end of the challenge, you will have created a DNS forwarding server. The server can create and read DNS packets and respond to DNS queries. As you go, you'll learn about the DNS protocol, its format, servers, and A records. All while getting to hone your language skills.<p>Some of the challenges and choices I had to make:<p>* To make the stages easier, I had to break them up, such that each step builds on the previous one. This was especially challenging for the 2nd stage, where we write a DNS packet's header contents. Even though I’d have liked it to be easier, breaking it up further would have been weird.<p>* Instead of implementing a recursive resolver, I've restricted to a forwarding server. We made this decision so that most developers can still use it. To add more complexity, we can use a challenge extension (noted below).<p>* Deciding how much instruction and context the stages should provide. I’ve decided to keep them as thorough as possible for most of the stages. Developers can choose to have thorough details or just skim through them.<p>I would love your feedback and questions on the challenge. You can try it out for free here: <a href="https://codecrafters.io/dns-server">https://codecrafters.io/dns-server</a> (no CC required).<p>I also have challenge extensions planned. You can find them at <a href="https://app.codecrafters.io/vote/challenge-extension-ideas?course=dns-server">https://app.codecrafters.io/vote/challenge-extension-ideas?c...</a>. I'm also keen to hear what you think about the extension ideas.
Show HN: A Dalle-3 and GPT4-Vision feedback loop
I used to enjoy Translation Party, and over the weekend I realized that we can build the same feedback loop with DALLE-3 and GPT4-Vision. Start with a text prompt, let DALLE-3 generate an image, then GPT-4 Vision turns that image back into a text prompt, DALLE-3 creates another image, and so on.<p>You need to bring your own OpenAI API key (costs about $0.10/run)<p>Some prompts are very stable, others go wild. If you bias GPT4's prompting by telling it to "make it weird" you can get crazy results.<p>Here's a few of my favorites:<p>- Gnomes: <a href="https://dalle.party/?party=k4eeMQ6I" rel="nofollow noreferrer">https://dalle.party/?party=k4eeMQ6I</a><p>- Start with a sailboat but bias GPT4V to "replace everything with cats": <a href="https://dalle.party/?party=0uKfJjQn" rel="nofollow noreferrer">https://dalle.party/?party=0uKfJjQn</a><p>- A more stable one (but everyone is always an actor): <a href="https://dalle.party/?party=oxpeZKh5" rel="nofollow noreferrer">https://dalle.party/?party=oxpeZKh5</a>
Show HN: A Dalle-3 and GPT4-Vision feedback loop
I used to enjoy Translation Party, and over the weekend I realized that we can build the same feedback loop with DALLE-3 and GPT4-Vision. Start with a text prompt, let DALLE-3 generate an image, then GPT-4 Vision turns that image back into a text prompt, DALLE-3 creates another image, and so on.<p>You need to bring your own OpenAI API key (costs about $0.10/run)<p>Some prompts are very stable, others go wild. If you bias GPT4's prompting by telling it to "make it weird" you can get crazy results.<p>Here's a few of my favorites:<p>- Gnomes: <a href="https://dalle.party/?party=k4eeMQ6I" rel="nofollow noreferrer">https://dalle.party/?party=k4eeMQ6I</a><p>- Start with a sailboat but bias GPT4V to "replace everything with cats": <a href="https://dalle.party/?party=0uKfJjQn" rel="nofollow noreferrer">https://dalle.party/?party=0uKfJjQn</a><p>- A more stable one (but everyone is always an actor): <a href="https://dalle.party/?party=oxpeZKh5" rel="nofollow noreferrer">https://dalle.party/?party=oxpeZKh5</a>
Show HN: A Dalle-3 and GPT4-Vision feedback loop
I used to enjoy Translation Party, and over the weekend I realized that we can build the same feedback loop with DALLE-3 and GPT4-Vision. Start with a text prompt, let DALLE-3 generate an image, then GPT-4 Vision turns that image back into a text prompt, DALLE-3 creates another image, and so on.<p>You need to bring your own OpenAI API key (costs about $0.10/run)<p>Some prompts are very stable, others go wild. If you bias GPT4's prompting by telling it to "make it weird" you can get crazy results.<p>Here's a few of my favorites:<p>- Gnomes: <a href="https://dalle.party/?party=k4eeMQ6I" rel="nofollow noreferrer">https://dalle.party/?party=k4eeMQ6I</a><p>- Start with a sailboat but bias GPT4V to "replace everything with cats": <a href="https://dalle.party/?party=0uKfJjQn" rel="nofollow noreferrer">https://dalle.party/?party=0uKfJjQn</a><p>- A more stable one (but everyone is always an actor): <a href="https://dalle.party/?party=oxpeZKh5" rel="nofollow noreferrer">https://dalle.party/?party=oxpeZKh5</a>