Replit

So many SaaS tools you can build for yourself in <1hr with AI + no code. I made this meeting brief bot with Replit + Zapier that - Researches attendee companies - Summarizes CRM notes - Pulls LinkedIn data 30 min before a meeting, it Slacks me my meeting brief in Notion. I host the bulk of the logic (e.g. crewAI, formatting, etc) on Replit. Zaps are temp placeholders when I want to avoid coding boilerplate (e.g. auth). Replit Deployments + Zapier webhooks 🤝 perfect for these personal Frankenstein apps.

I have plenty time to listen but never enough time to read 🤔 So I built an app! 🚀 Paper2Pod takes an academic paper and returns a podcast (audio)! It uses Google Artificial Intelligence Gemini + OpenAI TTS to convert a pdf to a audio-friendly format, then process that to an audio version in Streamlit. It's not terribly optimized, but it works! A future iteration would turn this into a service that you could simply email papers to and get an audio file. Built on Replit with Streamlit, Google AI, and OpenAI.

Boring workflow? Try Replit. Struggled with code in the past? It's never been easier to ship. ⛴️ If you find simple tasks eating up your day, basic automations like this one with Groq can work wonders (or just the opportunity to relax and take a deep breath). Automate tasks, build a web app, or deploy a scraping bot. New templates and guides are out every week to help you get started and AI-assist means you can build with minimal coding knowledge. Check out our Sheets Categorization template (with a full tutorial on getting started) that uses Groq and Google Cloud to automatically categorize customer support data in a Google Sheet.

What's life without jokes?! I recently discovered the "icanhazdadjoke" API and it's now, quite literally, my favorite API of all time. It's also great for demos— which is why I had to build a Discord Bot that tells a bad joke, on command! If you've ever wanted to build and deploy a Discord Bot, but didn't want to worry about infrastructure, Docker, deployments, or the like, this one's for you. You can go from zero to a fully deployed bot (with Replit) in under two minutes. Give it a watch and let me know what you think!

Something I've personally wanted for a few months now, but I'd refrained from building for myself and the team is a tool to speed up email responses. The reason I hadn’t built this was that I suspect we will see strong offerings from google etc soon, but the reality is the more I work with LLMs, the more I realise that what they will release will be fairly generic, and not easily tailored to achieve results which include specific details in the response to help in a broad range of scenarios. Essentially allowing say 50% of responses to be quick, but not being useful for the more complicated and a general scenario of you not being fully happy with the responses at all times. A few areas I see which may not work perfectly for a tool which is serving millions vs your specific needs and doesn’t exclusively rely on email content to build responses. - Your precise tone - Only train responses on what you would class as a good response (best practice) not your entire email inbox. - Knowledge must be easily updatable. Pricing changes, you can't get always this info from old emails.            - Dealing with specific queries - see pricing response in demo. - The knowledge you want it to use in a response and the knowledge you do not. - hook up to external knowledge base. - You need to be able to tell it what not to do explicitly - items that you want in a response are not necessarily going to be obvious from the email received, so you want to instruct. (see demo asking for review) - Work from shared knowledge resources. The tech team may have responses which support a customer success, so let's make that info available when responding. - quickly tweak responses you don't feel are a good fit. Provide examples so they never happen again. - How do you want it to build a response if there is no content or examples to respond with. What place holder response do you want to build upon. Built primarily for myself and the team to work more efficiently, but this Chrome Extension will be made available to all Answer Ai customers. Took me few hours to build yesterday evening with Anthropic 3.5 and Replit using OpenAI for inference.

The UC Berkeley AI Hackathon was INSANE. 1200 students BUILDING (on a weekend, too 🤯) I had a blast sharing how Replit at hackathons is basically a cheat-code. Standing room only for this one! The energy was amazing and it was inspiring to hear from such driven, motivated builders. Thanks to Tony Wan and the Reach Capital team for the opportunity! #AI #hackathon #UCBerkeley 

Ever get frustrated by simple syntax or typo errors with code? Our Code Repair model highlights errors and gives a one-click fix. Out to Core and Teams users today.

The first thing I did at Replit was build a Slack quickstart tutorial. It's pretty funny how looking back at old work can be like listening to a recording of yourself— nails on a chalkboard. There are always things I wish I did differently and this is one of those opportunities where I had time to polish up some old work. Today, I'm happy to share an evolution of my first project. DevRel is very much about showing simplicity and time-to-value over complexity or the "perfect solution." In that first tutorial, I used Slack Bolt. While it's a more "robust" framework, it doesn't show the speed most of our users will be after, so I rebuilt our Slack tutorial to use Webhooks instead. Now, you can deploy a Slack Bot to GCP-backed servers in under 2 minutes. Guaranteed. 🚀

Realizing you can code, build, and deploy your apps on iOS/Android with Replit mobile

🔥 Claude Sonnet 3.5 from Anthropic is arguably the top chat model, outperforming GPT-4o and GPT-4 in various benchmarks! 👉 Check out Matt Palmer's post to learn how you can launch a Gradio chatbot in just two clicks 🤯