Hey @bubble fam: my #devday code injection sermon is now on YouTube. It's a nice crash course on how to get started with speeding up parts of your Bubble app with JS code 👇. #nocode #bubble #webdevelopment.

Needed to know world news, so I just had a chat with @grok about it. And DAMN, it is good. Won't be surprised if the future of consuming news is having a conversation with a (hopefully objective) #AI chatbot.

Up all night building an #AI receptionist that you can chat with via @WhatsApp . Did not use @bubble for this at all. Guess the #nocode stack

If more #founders posted about the things they learned and how they grew in the process instead of the goddamn ARR graphs, the world would be a better place.

Always love these kinds of emails. Just pushed an update to a @bubble plugin off a user request. It's this multi-file Zip downloader. now supports 200+ files in a single download. 👇. #nocode

Anyone know how to get in touch with @facebook developer support?. Their 'support' link conveniently goes to a 404 broken link. @MetaforDevs @Meta.

Literally the simplest @bubble plugin I've ever built. Just glad it helps someone :)

Enabled image masks on an #app I'm building. Been playing with them for too long. Time to get back to it. The #AI bit was fun to build.But I like this part more

Just saw an RFP for a complex af HIPAA compliant @bubble app for the medical sector. Budget under $5k.

Bruhhhh agent mode dropped on the @OpenAI webapp. So I tried to order ingredients for dinner. Most websites have protection against #bots accessing them (@Cloudflare got it locked down), .but this is so cool!. Totally anticipate bot-friendly versions of popular sites coming up

Guess the @bubble plugin

STFU @Lendingkart you make the most spam calls yourself

Here we go.

Can't unsee it. When you tag @sama it looks like Osama

XML formatted prompts to pretty well too. When you hear 'machine-readable prompts', think of this. #AI #promptengineering.

9⃣ Mention explicit reasoning chains in the system prompt.Minimize hallucinations by making the AI follow a reasoning chain and provide a trail of facts that it uses in its response.

8⃣ Pre-screen user prompts.A cheap model like 4.1 mini or nano can check what the user is asking your AI to do, to make sure their prompt isn't malicious. Jailbreaks are always a possibility. Avoid them!.

7⃣ Machine readable prompts do way better than just large paragraphs.

6⃣ Chain steps to accomplish complex tasks. Break complex tasks into steps and chain them together. The #AI response is WAAAY better with such an approach.

5⃣ Most important instructions should be at the TOP of the system prompt. The #AI model is more likely to see things mentioned at the start every time, compared to other stuff that comes later in the system prompt (especially when it is large).