开源日报每天推荐一个 GitHub 优质开源项目和一篇精选英文科技或编程文章原文，坚持阅读《开源日报》，保持每日学习的好习惯。
今日推荐英文原文：《Future-proof Your Components With This Game Changing React Pattern》
查看文档 ： http://ui.shadcn.com/docs
今日推荐英文原文：A Journey into Wasm
The Fediverse Scales Up — RSS Readers 2.0?
It was a wild end of the year for the fediverse, which up till recently had been a loose collection of relatively small federated websites and applications built on the W3C ActivityPub specification. But it’s fast becoming a large and formidable federation. When Meta’s Threads app began connecting to the fediverse at the start of December, the scale of the fediverse suddenly went up a notch or two. Soon after, the social media app Flipboard also announced it will federate. More apps are expected to join this year, so it’s highly likely that 2024 will be the year the fediverse goes mainstream.
So what will happen over the coming months as more people use the fediverse? It’s hard to predict, but one thing many of us are looking for is a better way to interconnect with our online friends. ActivityPub is often referred to as “two-way RSS,” because it allows us to interoperate across different social networks. Whereas RSS is a protocol that enables publishers to distribute their content, ActivityPub is for both publishers and consumers of content. As the spec puts it, “It provides a client-to-server API for creating, updating and deleting content, as well as a federated server-to-server API for delivering notifications and content.” Note that the “activities” can be any kind of digital information — text, images, video, audio and whatever comes next.
One thing I’d love to see in the fediverse era is a renaissance in RSS readers, except this time for multiple types of content. I commented about that on Mastodon recently and got some intriguing responses. Brent Simmons, creator of NetNewsWire, wrote a post explaining that while his product isn’t suitable for fediverse content (he said it’s “a bad fit for a Mastodon reading experience,”) he thinks that someone else could create “a nice unified experience” that covers both RSS feeds and ActivityPub content.
Another commenter pointed out that the Vivaldi web browser includes both an RSS reader and a client for its own Mastodon instance, Vivaldi Social, in its desktop application. There will be a lot of other ideas that get explored in 2024. As Mike McCue, founder of Flipboard, said, “I think RSS and ActivityPub are like peanut butter and chocolate, and it will be awesome when they are combined.”
Doubling Down on AI Dev Tools
Understandably, many of the forecasts about web development this year revolve around the increasing use of AI.
According to Lee Robinson, vice president of product at Vercel, there will be an increase in AI-generating user interfaces. Of course, Vercel has its own such product (v0), but Robinson thinks it will go beyond that.
- Dev News: Vue 3.4, Nuxt 3.9, and React Server Components Analyzed
- New Open Source Standard Brings Consistency to Webhooks
- How to Build a Server-Side React App Using Vite and Express
“To accelerate the creation of the first version (v0) of projects and user interfaces, we’ll see more generative UI tools that enable instant creation of UI code from screenshots, drawing, voice or prompts,” he said. “Critically, the tools that embrace established industry tools (like React) for their outputs will lower the barrier for shipping generated code in real product use cases. Generative UI will not replace developers — rather, it’ll advance accessibility and augment the developer workflow.”
Rita Kozlov, senior product director for the developer platform at Cloudflare, took a broader view of AI tools for developers. She said that “frontend developers must redefine their role in the AI era.” She thinks that “established frameworks hold an edge over newer ones” when it comes to how effective they will be with AI-assisted coding. She elaborated in a follow-up email:
“There is more existing content out there, including documentation, and open source examples for more established frameworks like React, Angular and Vue.js that AI is trained on and can pull from. We can assume that AI tools will be more accurate in the responses they give regarding these languages, and that anyone relying on these responses can also find more supporting context outside of the AI chatbots as they look to confirm and augment the AI responses they receive as a starting point.”
Also: Get Ready for Multimodal AI
At the AI Engineer Summit last October, OpenAI said that 2024 will be the year of multimodal. The company shared a demo of GPT-4V (GPT-4 with vision), which converts images to text, and then uses that text to generate new images. According to OpenAI, text will be the “connective tissue” for multimodal large language model (LLM) development.
The number of multimodal LLMs will likely increase in 2024. According to Jacob Marks, a machine learning engineer at Voxel51, now “models like GPT-4V, Google’s Gemini, and open source IDEFICS from Hugging Face and LLaVA can natively ingest both images and text.”
Eleventy creator, Zach Leatherman, is also an advocate for a simpler web. Recently he hit back at Netlify’s state of web development report, which I covered in our 2023 web dev wrap-up. Leatherman said the Netlify report “looks bad for Eleventy” (and Next.js). He claimed that “Netlify has a vested interest in elevating Astro, because Astro is best poised to dethrone Next.js.” But it was the following comment that might be the biggest takeaway: