Some weeks back, we built Primo, the AI assistant inside @thepromptlab_ai website. It’s already functional there, with its own personality, conversation flow and AI infrastructure. So when I decided I wanted Primo inside our Telegram community, I had a choice: Build another bot from scratch with another personality... Or figure out how to make the same Primo work in another environment. I chose the second. The first step was understanding what we already had. I connected ChatGPT to my Lovable workspace, where The Prompt Lab website lives. From there, I could inspect the existing Primo implementation: the files, AI logic, personality instructions, chat routes, persistence and the way the website communicates with the AI. I didn't want to guess how Primo worked. I wanted to work from the actual implementation. Then I created a new bot through Telegram's BotFather. But creating the bot wasn't the hard part. A Telegram bot is basically just the interface. The real question was: How do I connect that interface to the same AI system already powering Primo on the website? This is where ChatGPT + Lovable became useful. Instead of manually rebuilding everything, I used ChatGPT to reason through the existing Lovable project and then send the implementation instructions directly into Lovable. Lovable could work with the actual project files, while I used ChatGPT to plan the architecture, inspect what already existed and tell Lovable what needed to change. The architecture we started moving toward was basically: Primo AI Core ↓ Website Primo ↓ Telegram Primo Same personality. Same AI foundation. Different interface. The goal is not to create “another Telegram chatbot.” It's to make Primo portable. We then built the Telegram side inside the existing Prompt Lab backend. Telegram sends a message → our webhook receives it → the backend passes the conversation into Primo's AI logic → Primo generates the response → the response is sent back through Telegram. We also added conversation storage so Telegram isn't just a dumb message relay. There were a few things to account for: • Telegram webhook delivery • Bot authentication/secrets • Conversation memory • Duplicate Telegram updates • Private chats vs group mentions • Response formatting • Keeping the website Primo working exactly as it already does And importantly, the Telegram version needed its own context rules because it obviously shouldn't have access to the same website-specific information a logged-in website user might have. The stack/workflow ended up looking roughly like: ChatGPT → planning, inspection & implementation instructions Lovable → builds/edits the Prompt Lab application Supabase → stores Telegram conversations Telegram Bot API → delivers messages Primo AI core → personality + intelligence The Prompt Lab website → the original home of Primo Different tools, one system. And this is where the build currently stands. The Telegram backend has been implemented and the existing website Primo is still intact. The remaining step is connecting the actual Telegram bot token, registering the webhook and testing the full message flow end-to-end. And of course... Lovable credits decided to finish right when we got to that part 😂 So I'm currently paused until the credits reset. Once they're back, I'll finish the webhook connection, send Primo its first Telegram message and see how the whole thing behaves in the wild. I'll update you guys when we get there.

I've always found it funny that we complain about taking notes, then panic the moment we need the information we didn't write down. As a student, you can sit through an entire lecture thinking “I'll remember this.” Then exam week comes and suddenly you're asking someone for their notebook because the thing you were so sure you'd remember is nowhere in your head. We write things down because we want something important to survive beyond the moment.

Tokenizing $GARRI

The Circus Telegram Bot is now open to everyone 🥳 To celebrate, we're giving away $1,000 (10 winners × $100). How to enter: 1️⃣ Open @CircusFYIBot on Telegram and send /start 2️⃣ Run /odds <site> <sport> and screenshot the result 3️⃣ Quote this post with your screenshot Good luck! 🍀

I've always found it funny that we complain about taking notes, then panic the moment we need the information we didn't write down. As a student, you can sit through an entire lecture thinking “I'll remember this.” Then exam week comes and suddenly you're asking someone for their notebook because the thing you were so sure you'd remember is nowhere in your head. We write things down because we want something important to survive beyond the moment.

🚨BREAKING: BITGET POTENTIALLY HACKED FOR OVER $100M

Friday 💛 One of the safest investments you can make is becoming someone your future self can depend on. Invest in yourself. ❤️

here’s what my Grokbot-style character looks like bet yours isn’t better, prompt in the Quoted post.

here’s what my Grokbot-style character looks like bet yours isn’t better (prompt in cs)

Apparently 4PM means “go home” for everyone else 😂 Canbo didn’t get the memo. @Canborsa_DEX is building a non-custodial perpetual DEX where you can access crypto, stocks, commodities and indices from one platform, 24/7. So yeah… Canbo is still working.

It's time to have some fun 🤪 Let's use Grok to put the Canborsa mascot in the funniest situations you can imagine! 💰 500 canton-network:native for the 5 best works 1) In the comments, hit attach media and generate your image with Grok 2) Post your result below 3) QRT this post with a short line on what Canborsa is That's it, let's see who comes up with the best one 👀

My entry for the @MoonlitRuna #MoonlitCreator contest. THE TABLE 🌙 An original piece of Moonlit lore exploring the Principle of Creation through a simple wooden table shaped by the hands of those who passed through it. A story about what we leave behind, and the space we leave for what comes next. It was fun working on this.