Show HN: ASim – generate functional mobile apps instantly on your phone https://ift.tt/UcjqvAl

Show HN: ASim – generate functional mobile apps instantly on your phone Hi HN! Daniel from YC S21 here, launching a project called aSim, which lets people generate/simulate usable native apps (called Sims) instantly from your phone. Describe an app you want and aSim will generate it. Then, edit and refine it to better suit your needs. Sims are also shareable via links, and basic app functionality is also exposed through web (though mobile is much more feature complete). A couple of my favorite Sims so far: - FridgeChef, recipes from what's in your fridge: https://asim.sh/s/419 - wordle: https://asim.sh/s/412 - HIIT workout randomizer/generator: https://asim.sh/s/418 - Pomodoro timer: https://asim.sh/s/232 Would love feedback around the experience and additional functionality you'd like surfaced! https://asim.sh/ September 4, 2024 at 11:03PM

Show HN: An open-source implementation of AlphaFold3 https://ift.tt/KZOR15E

Show HN: An open-source implementation of AlphaFold3 Hi HN - we’re the founders of Ligo Biosciences and are excited to share an open-source implementation of AlphaFold3, the frontier model for protein structure prediction. Google DeepMind and their new startup Isomorphic Labs, are expanding into drug discovery. They developed AlphaFold3 as their model to accelerate drug discovery and create demand from big pharma. They already signed Novartis and Eli Lilly for $3 billion - Google’s becoming a pharma company! ( https://ift.tt/VOq7oGH... ) AlphaFold3 is a biomolecular structure prediction model that can do three main things: (1) Predict the structure of proteins; (2) Predict the structure of drug-protein interactions; (3) Predict nucleic acid - protein complex structure. AlphaFold3 is incredibly important for science because it vastly accelerates the mapping of protein structures. It takes one PhD student their entire PhD to do one structure. With AlphaFold3, you get a prediction in minutes on par with experimental accuracy. There’s just one problem: when DeepMind published AlphaFold3 in May ( https://ift.tt/KJyp9tc ), there was no code. This brought up questions about reproducibility ( https://ift.tt/6TLtJYx ) as well as complaints from the scientific community ( https://ift.tt/q7jDrhy... ). AlphaFold3 is a fundamental advance in structure modeling technology that the entire biotech industry deserves to be able to reap the benefits from. Its applications are vast, including: - CRISPR gene editing technologies, where scientists can see exactly how the DNA interacts with the scissor Cas protein; - Cancer research - predicting how a potential drug binds to the cancer target. One of the highlights in DeepMind’s paper is the prediction of a clinical KRAS inhibitor in complex with its target. - Antibody / nanobody to target predictions. AlphaFold3 improves accuracy on this class of molecules 2 fold compared to the next best tool. Unfortunately, no companies can use it since it is under a non-commercial license! Today we are releasing the full model trained on single chain proteins (capability 1 above), with the other two capabilities to be trained and released soon. We also include the training code. Weights will be released once training and benchmarking is complete. We wanted this to be truly open source so we used the Apache 2.0 license. Deepmind published the full structure of the model, along with each components’ pseudocode in their paper. We translated this fully into PyTorch, which required more reverse engineering than we thought! When building the initial version, we discovered multiple issues in DeepMind’s paper that would interfere with the training - we think the deep learning community might find these especially interesting. (Diffusion folks, we would love feedback on this!) These include: - MSE loss scaling differs from Karras et al. (2022). The weighting provided in the paper does not downweigh the loss at high noise levels. - Omission of residual layers in the paper - we add these back and see benefits in gradient flow and convergence. Anyone have any idea why Deepmind may have omitted the residual connections in the DiT blocks? - The MSA module, in its current form, has dead layers. The last pair weighted averaging and transition layers cannot contribute to the pair representation, hence no grads. We swap the order to the one in the ExtraMsaStack in AlphaFold2. An alternative solution would be to use weight sharing, but whether this is done is ambiguous in the paper. More about those issues here: https://ift.tt/IcaJol3 How this came about: we are building Ligo (YC S24), where we are using ideas from AlphaFold3 for enzyme design. We thought open sourcing it was a nice side quest to benefit the community. For those on Twitter, there was a good thread a few days ago that has more information: https://twitter.com/ArdaGoreci/status/1830744265007480934 . A few shoutouts: A huge thanks to OpenFold for pioneering the previous open source implementation of AlphaFold We did a lot of our early prototyping with proteinFlow developed by Lisa at AdaptyvBio we also look forward to partnering with them to bring you the next versions! We are also partnering with Basecamp Research to supply this model with the best sequence data known to science. Matthew Clark ( https://batisio.co.uk ) for his amazing animations! We’re around to answer questions and look forward to hearing from you! https://ift.tt/IcaJol3 September 4, 2024 at 11:14PM

Show HN: Icebreaking AI. A free tool to help you find close friends https://ift.tt/3X57Ey4

Show HN: Icebreaking AI. A free tool to help you find close friends Hello, everyone! My name is Alex, and I'm an expat in a new country. Two years ago, I moved to Germany where I didn’t know anyone. Starting the job, I realized that becoming close friends with colleagues can be quite challenging and slow. The feeling of loneliness was gradually growing. To change that, I started attending various meetups based on my interests, hoping to make new friends. However, I often found myself answering the same generic questions that didn’t really help me understand if someone shared my values and interests. As a result, lost energy and 0 friends. SO - That’s when I began using icebreaking questions in smaller groups. Surprisingly, these questions quickly changed any conversation to a deeper, more engaging and meaningful level. I was able to identify people I genuinely connected with and started spending more time with them. BUT - preparing and using these questions was a pain. I would Google the top-100, pick a random number, and choose a question. I felt a strong need for a simple „one button“ that does everything for me. AND - After searching and not finding such a solution, I decided to build it myself. I created a vast database of questions across various topics and integrated AI to generate questions on any subject. The idea has resonated well within my communities (university and expat parties), and more people are starting to use it. I’d love to hear your stories with icebreaking questions, your experience, and your feedback so I can make the service even more useful for more people! Check it out and share your thoughts! https://ift.tt/3GQUvdp September 4, 2024 at 01:24AM

Show HN: Hestus – AI Copilot for CAD https://ift.tt/2H564Mf

Show HN: Hestus – AI Copilot for CAD Hello! We’re Kevin and Sohrab from Hestus ( https://www.hestus.co ). We're working on an AI copilot for CAD. Today we're releasing a simple sketch helper for Fusion 360 and would love your feedback. Here’s a quick demo: https://www.youtube.com/watch?v=L9n_eY-fM_E . Why we’re doing this: Mechanical engineers excel at generating initial design concepts but get bogged down translating ideas into final designs due to tedious, repetitive tasks. Our goal is to automate these mundane processes, allowing engineers to focus on the creative aspects of design. Having worked at multiple hardware companies—from medical devices to space launch vehicles—we know how often “trivial” components such as manufacturing rigging, get brushed under the table in scheduling conversations. These tasks aren’t necessarily complex, but they take time and still require the rigor of production components. From finding the perfect fastener to making sure mounting holes align, we aim to simplify and accelerate the design process from the complex to the mundane. We're tackling this problem similarly to how coding copilots help programmers work faster. Initially, rudimentary coding assistants offered simple suggestions like auto-completing variables. Now, they understand complex tasks, write entire code blocks, and help fix bugs. We're taking this step-by-step approach, starting with a beta that focuses on sketching. Our sketch helper offers design suggestions, such as applying equality constraints to similarly sized circles or adding tangent constraints between lines and curves. While designers can do these tasks manually, they often require dozens of precise mouse clicks. Our software makes suggestions that you can preview and accept to streamline your workflow. Over time we aim to improve at anticipating your needs and expand beyond sketching to other design aspects like resolving interference issues, auto-generating bills of materials with purchase links, and offering manufacturability suggestions. How this is different from other solutions: we've heard of complete generative part design solutions, but we don't believe this top down approach is the best way to assist mechanical engineers. Engineers excel at and enjoy designing new concepts—we want to focus on streamlining the most tedious aspects. Crucially, we find that generative solutions often overlook manufacturability, a key aspect of design. We invite you to try our sketch helper and share your thoughts! If you can think of any additional features that would make it more useful to you, we’d love to hear what they are. Any and all feedback is welcome! https://www.hestus.co/ September 4, 2024 at 12:19AM

Show HN: I made a REST based alternative to GraphQL for PostgreSQL https://ift.tt/zpCVYRT

Show HN: I made a REST based alternative to GraphQL for PostgreSQL Hi guys! I have had a love-hate relationship with GraphQL for a while and decided to build my own tool as a side project! It's a REST-based API generator that you can use to build and deploy on top of PostgreSQL. I am currently working on a GitHub integration so you can push your API directly to GitHub and deploy on your own infrastructure. The plan is to stabilize the features by December and then to open-source it. It's completely free to use since it's in beta, and I would appreciate any feedback I can get. Thanks: kabir@querydeck.io https://querydeck.io/ September 3, 2024 at 11:15PM

Show HN: Full Text, Full Archive RSS Feeds for Any Blog https://ift.tt/tX3F9Zl

Show HN: Full Text, Full Archive RSS Feeds for Any Blog https://ift.tt/qDyL6t9 September 2, 2024 at 06:36PM

Show HN: A YouTube videos course generator https://ift.tt/5rVuSjZ

Show HN: A YouTube videos course generator An exciting aspect of what I'm working on is that users can create their own playlists. For example, you can provide an outline of what you're learning, such as notes from your professor or any other structure, and the pipeline will create a course tailored to that. Since the content is highly customized, it will be relevant and high-quality, matching current lessons or lectures in school or university. Initially, the pipeline also generated some teaching content using AI, but due to the risk of AI hallucinations, I'm planning to pivot away from that. The new focus will be on curating YouTube videos relevant to each subject, so the courses will consist entirely of expert-created content from YouTube. This approach avoids copyright issues entirely. I'm still in the process of updating the code to implement these changes. The reason I'm posting this is to get feedback on how it sounds to you as learner. I'd love to hear your thoughts, does this solve a problem for you, would you be likely to use it if it just works ? I'm trying to find a product structure that users will find valuable and effective. The web app is completely free to use right now as I figure out the best model. Thanks! https://coursely.ai September 2, 2024 at 11:00PM