Dark

a lot of work into rebuilding our streaming repo from scratch with a high performance foundation excited to land streamdown-rn v0.2.0: github.com/darkresearch/s…

we're rebuilding github.com/darkresearch/s… to be more performant and more reliable as it makes its way to 1k github ⭐️s complete rebuild, from the ground up - strict memoization policy for performant rendering - elegant regex->remark handoff in the parser - robust incompletion handling: the stream immediately looks formatted to the user three things driving our investment in this: 1) was very inspired by @vercel's v0 ios app writeup - i'm growing more of the view that react native can compete with native builds on performance and elegance, and i see no reason why @darkresearchai shouldn't meet that standard 2) there still aren't good ai streaming libraries out there for mobile. streamdown-rn is set to be a leader here and we still have a ton of work to do - it's important we stabilize things now that the library is getting more eyeballs 3) we can't live without it internally, which is an easy reason to build something some other nugget in the branch - more proper app starter template, the beginnings of our spatial generative ui work, etc. looking forward to landing the branch

streamdown-rn v0.1.6 has landed: · a framework for streaming generative components · a much more robust set of fuzz tests · stability updates stream responsibly 🎉 github.com/darkresearch/s…

streamdown-rn: @vercel Streamdown ported to React Native.

seeing streamdown-rn growing in likes on github! if you're looking for this 👇 feature specifically, it's on the v0.1.6 branch - hasn't landed to main yet but will land asap link below

Read more: darkresearch.ai/blog/streamdow…

Every app you use should evolve itself for you, perfectly personalized to be exactly what you need at every single moment. We're building that:

We'd love to hear what you think. streamdown-rn is open source, and you can contribute today. Star the repo here: github.com/darkresearch/s…

It also features something new: we call it Dynamic Component Injection. Dynamic component injection gives LLMs a framework to arbitrarily render RN components through their responses, following a double-bracket notation that, to the LM, is as easy as standard markdown rendering. This is streamable like any other markdown, supports custom RN components registered through a whitelist made known to the LLM, and is importantly decoupled from tool use. The idea is to give LLMs complete control over when to render whitelisted RN components and how to saturate them.

Introducing streamdown-rn: A react native version of @vercel's streamdown library, with an added Dynamic Component Injection feature that allows LLMs to arbitrarily saturate and render RN components decoupled from their tool use. darkresearch.ai/blog/streamdow…

This makes it incredibly easy to start contributing and earning: 1. Go to github.com and sign up for a free Github account if you don't already have one. 2. Find one of Dark's open source repos on our Github page, and star it so you don't lose it ⭐️: github.com/darkresearch 3. Navigate to the Issues tab for that repo (like here: github.com/darkresearch/m…) and add a bug that you've found in one of our apps or a feature request you'd like to see added. You don't need to know how to code - you just need to thoughtfully contribute to the conversation. See something that isn't working in an app we built, like mallory.fun? Have an idea for an app feature or functionality you'd really love to see? All of that is fair game. 4. If your issue gets picked up by an engineer (even if they're on our team!) and merged to production, you get paid 💰. Think your feature is particularly worthy of getting picked up? Make noise about it on X and tell everyone to ❤️ your issue so it gets more attention. Payouts will be done at the end of every week.

We use @merit_systems to guide our distribution payouts. The framework is simple: the more impactful your contribution is to the repo, the more you stand to earn. We're extending this framework to non-technical contributors: anyone who raises a Github issue that leads to a pull request being merged to a Dark open source repo earns 10% of the payout towards that PR.

🔔Notifications on.

If you lost money in the markets today, build to get it back. $20,000 is on the table for contributions to @darkresearchai open source - including $10k for the @solana x402 hackathon I'm putting up a new issues on gh tonight for inspiration Learn more: darkresearch.ai/blog/project-o…

The full episode is out now: @edgarpavlovsky joined @seedclub's @11AMdotclub for 26 minutes of packed content - watch below. · Why there's Dark (the AI lab) and Scout (the product) · Why crypto & AI will unlock next-gen software, and what incumbents are missing to get this done · Why our consumer finance product, Scout, is ultimately competing with Robinhood, and why Dark has a key advantage Catch up: