{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreiglsljrqeg4yqy4lrhzn6txlujel2ssf6jupitvp3dgxioly7oiui",
"uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mi45h6idoxf2"
},
"path": "/t/adaptive-coding-interface-14year-old-looking-for-feedback/174720#post_1",
"publishedAt": "2026-03-28T04:29:52.000Z",
"site": "https://discuss.huggingface.co",
"tags": [
"aimlse.org"
],
"textContent": "Hello everyone,\n\nI’m Sourish.\n\nMy friends and I are building something we call an **Adaptive Coding Interface (ACI)** — a development environment designed to feel more dynamic, collaborative, and less isolated than traditional coding setups.\n\nWhile working on ML projects, especially with PyTorch, I kept running into a recurring issue: a lot of workflows feel highly repetitive. It’s not that reusable code doesn’t exist — it does — but it’s often fragmented and hard to discover efficiently (even with platforms like Hugging Face helping).\n\nThat led me to explore a different approach:\n\n * What if you could describe your project in **natural language** , and get relevant helper functions tailored to your use case?\n\n * What if the system could **learn your preferences over time** and adapt suggestions based on your workflow and similar projects?\n\n * And instead of stopping at helper functions, what if this extended to **entire templates and pipelines**?\n\n\n\n\nBeyond that, we’re thinking about the development environment itself:\n\n * **Live collaboration** , so coding feels more like a shared workspace than a solo process\n\n * A **sharing layer** , where users can publish and discover projects, helper functions, and templates\n\n\n\n\nThe goal is to make building ML systems feel less repetitive and more intuitive, while still keeping flexibility.\n\nCurious to hear — does this approach actually sound useful in real workflows, or does it risk adding unnecessary complexity? Also we are looking for people to test this platform. If you are interested in being one of the first people to try this go to aimlse.org , its limited spots. Your feedback will help us create the greatest PyTorch editor ever.",
"title": "Adaptive Coding Interface (14year old looking for feedback)"
}