Most AI generated UIs are starting to look painfully similar (yeah that's what we call slop). Same purple gradients, paper+orange splash, same rounded cards, same layouts with no taste or personality. And the annoying part is that you end up spending more time fixing the UI than you did generating it. One thing I've found useful is giving coding agents actual design components to work with, rather than asking them to figure out everything from scratch. AvaraUI by @Pranto39 is an interesting library for this. It has 189 free components built on shadcn/ui, with MCP support so coding agents can work with them. A few ways you could use it: - Building a SaaS dashboard? Use existing data tables, charts and navigation components instead of generating another generic admin panel. - Working on an AI app? There are chat interfaces with streaming responses, model selectors and expandable input areas. - Need a landing page? Explore animated sections, carousels, navigation and FAQ blocks, then adapt them to your product. - Building for clients? You can reuse the components and customize their colors, typography and other styles to match each brand. - What I particularly like is that you're getting actual component code in your project. You can inspect it, change the interactions, remove unnecessary animations or have your agent modify things further. Of course, throwing 20 fancy components into a page won't suddenly give it good design. But having a decent set of building blocks gives you a lot more room to experiment than repeatedly prompting your agent to "make it look premium." Worth exploring if you're building with React and shadcn/ui: avaraui.com

One thing people building voice AI often overlook is how much information gets lost before the LLM even receives a transcript. Think about someone learning English. They'll pause mid sentence, pronounce words incorrectly, repeat themselves, or make grammatical mistakes. Now imagine your STT model quietly cleans all that up. The transcript might look great, but an AI tutor relying on it could completely miss what the student needs help with. That's why I found this benchmark from Fluently interesting. They tested real time speech recognition on low fluency English spoken by Chinese, Japanese, Korean, Thai and Vietnamese learners. And the differences between models are pretty significant. If you're building anything with voice AI, here's what I'd test before choosing a transcription model: - How well does it handle your actual user's accents? - Does it preserve hesitations, repetitions and grammatical mistakes, or quietly clean them up? - How often does it get important words wrong, especially negatives, names and numbers? - How well does it handle pauses, interruptions and spontaneous conversations? A low word error rate is useful, but it doesn't tell you everything about how well the model will work in your application. So we need to test speech recognition on the people who'll actually use your product, similar to the example of Dessa that I shared few days back.