Thread

🛡️
Today we are thrilled to officially announce the opening of 🎭 Shakespeare: Act II. Act II is a dramatic improvement over the first version of Shakespeare. We have completely rewritten the codebase from the ground up to make it the most freedom-respecting AI web builder available. Read more:
Soapbox
Shakespeare: Act 2 - Build Freely - Soapbox Blog
Shakespeare Act 2 transforms from a simple AI website builder into a full-featured development environment, entirely in your browser (no downloads ...
 Try it here:
Shakespeare - Open Source AI Builder
Build custom apps with AI assistance using Shakespeare, an open-source development environment
 What makes Act II different: 1. Runs entirely in the browser: Unlike Act 1, where your project files were stored remotely on our NSP servers, now all of your projects are stored locally in your browser — so you retain all control and full privacy. 2. Full AI provider and model choice: Before, you were locked into only the providers and models available from NSPs. With Act II, you can configure ANY AI provider you want. Bring your own credentials from your favorite provider like PerPerQ, Routstr, OpenAI, Z.ai, and Anthropic; use our pre-configured Shakespeare AI provider (with both free + paid tiers), or even run your own local model! This is the most choice of ANY AI web builder we’re aware of. 3. Complete development environment in the browser: We’ve built a text editor, terminal, and git version control integration directly into Shakespeare. The new changes we’ve made to the architecture of Shakespeare unlock new possibilities. One feature we’re most excited about is the ability to “Edit With Shakespeare.” Find any compatible open source project and start improving it in Shakespeare with one click! This is the power of open source, and with AI it’s finally accessible to everyone — no matter your technical level. Want to try it? Pick any app or template from our showcase and click “Edit with Shakespeare”! Showcase:
Soapbox
App Showcase - Community-Submitted Apps Built with Soapbox Tools | Soapbox
Discover and submit applications built with Soapbox tools. Browse community-submitted projects, games, social platforms, and business applications....
 Need help? We know debugging is always the worst part of building. We’ve built multiple new tools into Shakespeare to make it easier, including an in-web console to show errors, and a friendly assistant called Quilly who may make an appearance to alert you to key errors. Keep an eye out for him! We’ve also built out a detailed Shakespeare resource center with an FAQ and guides on how to get the most out of Shakespeare. Resource center:
Soapbox
Shakespeare Resources - FAQ & Guides | Soapbox
Complete resource hub for Shakespeare AI website builder. Find answers to frequently asked questions and explore tutorials on how to make the most ...
 Shoutout to our team and community who have been working tirelessly for the past 3 months on this release! @Alex Gleason @npub1jvnp...89sc @chad @MAKE SONGS LONGER @Derek Ross @npub1hee4...rsv6 @Sam @danidfra@danidfra.com @Patrick PReis @shantaram And, finally, thank you to @And Other Stuff @OpenSats and @npub17xvf...c9as for making this work possible!

Replies (48)

Let’s go! Incredible work on this. Does the updated codebase have the references in it to build out Nostr compatible login with nsec managers like Nsec.app? I think these are done through NostrConnect and bunker codes. *apologies for my non-technical lingo
I gave it a try (using the Tybalt AI), and it didn't start generating anything until my second prompt. Also, once it did generate some files, I was unable to see them (nothing showed up in the sidebar, though the browser preview did change, and I could click on individual changes in the log and see the file diffs). Any ideas what might be going wrong? image
↩ replying to npub174tp...lth2
I'm sorry that happened :( Tbh we're considering turning off the free model entirely because of this pattern. We want people to have something they can try without any cost, but the reality is that only good models work well, and good models cost money. Let us know how the Shakespeare flagship model does? Or if you use another provider, I recommend Claude Sonnet or GLM. More details on the models and what to expect from each here:
Soapbox
Which AI Model Should I Use with Shakespeare? - Soapbox Blog
Choosing the right AI model can transform your development experience in Shakespeare. With access to 40+ models across multiple providers, you have...
↩ replying to npub1jvnp...89sc
Hi MK, thanks for taking the time to respond. I gave it another try and it turns out the problem has nothing to do with the AI model at all — it was simply me misunderstanding the way the app works (i.e. a classic case of PEBKAC). I was expecting the project files to appear in the left sidebar, but apparently that's just for listing projects. The button I was looking for was up there in the right corner, and it works fine. Sorry for the confusion. I hope I didn't cause any major grief on your end. CC @Alex Gleason image
↩ replying to npub1jvnp...89sc
No problem. I do have some more questions though, if you don't mind. - If I enter an API token in the settings, where is that stored? In the browser's LocalStorage? (Same for GitHub/GitLab access token) - What exactly is the Nostr account in the settings used for? Also, I did notice a few other issues: - xAI integration isn't working (at least for me). I always get the following error no matter which model I choose (I was able to use DeepSeek as a custom provider however — would be nice though if that was added to the default list): image - The debug output isn't very useful: it only shows the error message but no stack trace or even file name / line number: image - Using the web inspector to debug the app, as suggested here:
Soapbox
Debugging in Shakespeare - Soapbox Blog
When things don
 isn't super productive since the preview seems to be running production mode (meaning all the JS is minified and squashed into a single file, and no source maps either) - When using git in the Shakespeare terminal, it does not show of the any local commits that the AI made. Since there appears to be no other way to inspect commit diffs to see what exactly was changed (clicking on the commits in the chat log or the rollback history only shows the commit message, nothing else), this seems pretty essential. Hopefully this list is useful and does not come across as overly critical (and hopefully it's also not just another case of me being an idiot). I do love the overall concept of Shakespeare and I want it to be great.
↩ replying to npub174tp...lth2
Thank for this list! It's very helpful. I will open issues for these things and we'll look into them 🙏 To answer the questions: - Yes, API tokens and Gitlab/Github tokens are stored in your browser in the same place where the project files are stored. - Nostr login can be used for a few things: 1) Use Nostr git to sync your projects (instead of Gitlab or Github) 2) If you want to use Shakespeare AI or Tybalt AI (instead of BYO AI API key), those work over Nostr so you need to be logged into a Nostr account to use those. 3) If you are logged into Nostr, Shakespeare knows things about you and your account and can use this info in your projects. For example, if you just say "Make a site about me" it will pull from your Nostr profile to get info for the site. In the future we'll likely continue to build on top of the Nostr integration for more interop across the network!
↩ replying to npub174tp...lth2
> "When using git in the Shakespeare terminal, it does not show of the any local commits that the AI made." Can you tell me more about this? I'm not able to replicate it. If you could do git log in the terminal and screenshot that and compare it to the commits you see under "rollback", that would be super helpful. For example, here is what I see, and they appear to match:
image image
↩ replying to npub1jvnp...89sc
> Can you tell me more about this? Sure! I tried importing Shakespeare into Shakespeare to see if I could have it add syntax highlighting to the code editor. The AI made a couple of commits (most were attempts to fix errors that started popping up right after the initial change), the most recent one of which can be seen in the screenshot. But when I open the terminal and type `git log`, I only see commits from the upstream repository: image Curiously, when I run `git show`, it DOES show the most recent commit made by the agent (no diff though), but when I try to inspect the previous commits using `git show HEAD~1`, `git show HEAD~2`, etc., it just says "Failed to show commit: Could not find HEAD~1."
🛡️
↩ replying to npub174tp...lth2
To be honest the free AI is very hit or miss. Sometimes it's fine for testing and works as a perfect trial and error tool for newbies to learn how to use Shakespeare and to learn how to talk to an AI. But because it's free, and uses a cheap and lean model, it sometimes takes shortcuts and doesn't complete the work. If you have API keys from one of the larger or top models, I'd recommend using that. You'll see an incredible difference.
Valentino Giudice's avatar Valentino Giudice
Did you build Shakespeare (which happens to be both a website and an app) without any coding skills? If so, why did you have to "work tirelessly" to build Shakespeare: Act II? You *just* had to "tell Shakespeare what website you want to build, and watch it" build it, am I right? If not, why not? Is it because, maybe, you are full of shit? Is that an option? Or maybe that's only the case now. Maybe it's only now with Shakespeare: Act II that no coding skills are required and, really, no work, since you can just tell it what to make and watch it build it. In that case, I expect future iterations of Shakespeare to be build without any effort (and little to no funding), without anyone having to "work tirelessly". Either that or you are full of shit. View quoted note →
View quoted note →
↩ replying to Valentino Giudice
It's probably not worth bothering, but I will answer this in good faith because it's actually very simple: We did not make changes to the models themselves. The premium models have been capable of building these kinds of apps for 6 months, at least. What we built (often with AI-assistance in the terminal and code editors) is a website that allows non-coders to use the same AI tools that coders already have access to. And yes! We did use our own tools to build it! But we also had to build, update, and maintain those tools in the first place, too :) So, if you already knew your way around a terminal, text editor, nostr libraries, AI providers, and managing agents, then all Shakespeare does is provide a convenient UI wrapper for your projects. However, if you have never opened a terminal before, don't know what AI providers are or how to choose a model, don't know how to prompt to get it to build a Nostr app... we put all of that in a friendly web interface with most of the hard stuff behind the scenes so you can have low-barrier access to the same power that coders already had. The goal is to make what WE could already access accessible to everyone. I think we've come pretty close to that. Yes, there are limitations on what AI can build (though those limitations shrink by the day). But at least now everyone can use the same tools, which is actually sufficient for a huge population of people whose projects don't have complicated or unique technical needs. We've been very open in documenting this process from the ground up, and encourage you to dig into how the tech works behind the scenes if you're curious!
Soapbox
AI-First Development with Stacks - Soapbox Blog
Stacks is a new way to share AI project templates on Nostr. Browse starter kits, remix your own, and deploy lightning-fast AI apps — all from the...
Soapbox
MKStack: Vibe Coding for Everyone - Soapbox Blog
A comprehensive guide to building Nostr apps without coding experience using Stacks, MKStack, and AI-powered vibe coding with Dork.
↩ replying to npub1jvnp...89sc
> What we built (often with AI-assistance in the terminal and code editors) is a website The exact thing which can be built with no coding skills, by just telling the model to build it and watching it. How did that take so much work? > We did use our own tools to build it! But we also had to build, update, and maintain those tools in the first place, too :) Why? If the model is the same, and if it can build websites on its own, why was it a tool and not straight up the developer? > Yes, there are limitations on what AI can build Really, because the website says "Just tell Shakespeare what website you want to build, and watch it create professional sites". How does the frontend to the existing model not qualify as a professional site?
🛡️
↩ replying to npub1n0v0...jxvn
Greetings! I've tested this scenario using my own Ollama host via docker compose. In short, what you're likely seeing is the browser blocking mixed content (http://localhost and
Shakespeare - Open Source AI Builder
Build custom apps with AI assistance using Shakespeare, an open-source development environment
). This interaction is blocked by Brave's "shields"; I was unable to replicate this in Chromium, but the same interaction is very likely to be blocked by an adblocker/privacy extension. Disabling Brave's shields and/or similar extensions on
Shakespeare - Open Source AI Builder
Build custom apps with AI assistance using Shakespeare, an open-source development environment
 is one solution, however you can also get around this by using Ollama over a valid HTTPS host, or even hosting an instance of Shakespeare locally over HTTP (thus avoiding mixed content entirely). For the last option, here is a rough example of a service that could be added to a docker-compose.yml file that clones, builds, and serves the static site files locally: ``` services: shakespeare: image: node:20-alpine ports: - "8081:3000" working_dir: /app command: > sh -c " echo 'Starting build and serve process...' && apk add --no-cache git && rm -rf /tmp/shakespeare && git clone
GitLab
Soapbox / Shakespeare · GitLab
AI-Powered Nostr Website Builder https://shakespeare.diy/
 /tmp/shakespeare && cp -r /tmp/shakespeare/* . && cp -r /tmp/shakespeare/.* . 2>/dev/null || true && npm ci && npm run build && npm install -g serve && serve -s dist -l 3000 " volumes: shakespeare_dist: ``` Hope this is helpful in some manner!
image image