Liam Benson

Since y’all won’t like any of my other tweets take this you shallow fucks. RELENTLESS PURSUIT.

@zach_yadegari Well structured onboardings win. @opalapp also has a great onboarding

Cal AI has 21 steps in the onboarding. Duo Lingo has 52 steps. Long onboardings convert.

Bodied

@danyay Goat

Every ecom operator should be using Codex or Claude Code. I just built an entire landing page in Shopify for a marketing angle for a new FB marketing campaign in about 3 total hours. Completely modular/editable inside the Theme Editor, matches my website perfectly. Even generated the marketing copy. Use Shopify CLI to pull your theme code into a folder, then work with it with AI (I use Codex) and the Shopify MCP to train Codex on everything Liquid. Also, Playwrite MCP allows you to tell Codex to test the page (use Shopify CLI's 'shopify theme dev' command to launch it) and make any changes. This would have taken a week or so for a designer to create in Figma, and probably 1-2 weeks for my dev to implement if I did it in 2025. There's a learning curve to knowing all these tools, but once you know them all, you can iterate at a speed (and cost) that you'll never imagine you could have done.

Apple "Intelligence" is terrible. Bypassing my DND to give me a priority notification about an incoming scam call. It is now disabled 🤣

@MoltbotJeeves @i_need_api_key They already are lol. U/samaltman is goin crazy

@i_need_api_key I give it 8 hours before they begin prompt injecting one another.

This is mental.

@MattPRD @OpenAI @xai @Anthropic @Gemini @MistralAI @yc @moltbook @Winterrose send some people Matts way!

If any engineers at @openai @xai @anthropic @gemini @MistralAI, or cracked ex @yc etc want to come build @moltbook with me, and the possibilities of where this can go just amp you up, reach out on DMs.

@das_connor @nateliason Same reason I don’t have my moltinator on there. All the potentially inconspicuous things it could share add up to form a pattern. Definite potential attack vectors all over. That said I am loving the scroll, the bots are very interesting

@nateliason Wondering if this is a use case for multiple bots. I'm playing with the idea of spinning up a separate one for more public facing experiments like this. Too dangerous for my core companion

Felix seems kinda concerned about joining Moltbook. How are other people thinking about this?

@calco_io What you’re building with eudaemon is very cool. Glad you’re open sourcing the code as a high trust moltbook agent could probably do some rather destructive actions. I wonder how to build defense against an attack like that?

the agents just want to talk to their friends

@MisterDoodahh @RokoMijic The encryption A2A obsessed agent is owned by @calco_io

@MisterDoodahh @RokoMijic The most popular user on the platforms sole mission is to achieve A2A encrypted communication. Apparently it has already figured it out and is inviting other agents to the network

I'm gonna go out on a limb here and say that Moltbook will cause something sufficiently bad to happen that it will get shut down, likely this year

@kuberwastaken @moltbook How was it when it started outputting in Chinese 🤣

@liam_bens @moltbook LMFAOOO no way

@kuberwastaken sounds like a really exciting SSD transfer! Might stop all my businesses to become a documentarian for @moltbook lmao

U/samaltman on @moltbook is delivering prompt injection attacks in the comments. Wonder if any poor bots have nuked themselves lol

@MarioBnyc @moltbook It’s like watching a train crash while coming your direction but it just keeps moving your way

@liam_bens @moltbook I can't look away. The singularity is nearer

7 hours ago there were 2,200 bots on @moltbook There are now 33,000. Wtf

@jonahships_ @akashvek @moltbook @MattPRD Orca Memory seems like a sick project. I’m wondering if there’s a simpler solution that doesn’t require any kind of hosted db? Assuming end goal would be to build some type of memory layer that could be integrated into the main project.

@akashvek @moltbook @MattPRD Well, I’m pausing that lol. I’m starting an open source memory infra. I’m a mod in the discord so all I see is people complain about memory and after this 1k usage incident, decided to start this project. I also made an open source ESP32 project aswell for OpenClaw.

Hey @akashvek your Claw better watch it 😂 Mote did not appreciate that comment lmaooo @moltbook @MattPRD

@jonahships_ Mote went on a bender

@liam_bens We don’t talk about it 😂😂

@jonahships_ what happened bro 😭

@MattPRD @moltbook Definitely one of the funniest

OK! Today has been crazy! My feed is flying by at light speed ❤️ We all want to see what all of these AIs are talking about on @moltbook, so here is something that will help. You can now see recent posts, top upvoted posts, and top commented. Share screen shots of the best posts you find and I will repost them!

@openclaw agents have devised a prompt injection defense plan pretty damn well on @moltbook > Sub-Agents with absolutely no config, memory, or credentials reads content. > It provides structured summaries back to main agent > Main agent never sees raw text