Finding signal on Twitter is more difficult than it used to be. We curate the best tweets on topics like AI, startups, and product development every weekday so you can focus on what matters.
I spent the weekend setting up Clawdbot. By Sunday evening, I had an AI agent that summarizes my Twitter feed, one that recommends new books weekly based on my recent reads, and a third that texts me every morning with my schedule, a weather alert, and a fun quote.
It's genuinely magical. And, I don't think most people should try it (yet!) Here's why, and here's what you can use instead.
What makes Clawdbot special
How is Clawdbot different from the AI products we already had?
It meets you where you are. Your AI lives in WhatsApp, iMessage, Telegram - whatever you already use. Especially for reminders, quick tasks, etc. This is a big difference from having to open a standalone app, and there are few truly agentic products that work on mobile.
It actually does things (with high success rates). Clawdbot can send messages, manage files, browse the web, book things, run code. It works both through its own browser agent and a library of connections.
It automates. Set up a recurring task once and it just runs. These tasks can be time triggered, or situational or action-triggered. For example - “if I get a message from a customer saying X, respond saying Y”. This is surprisingly hard to do on ChatGPT, Claude, Gemini, etc.
This is the AI assistant we were promised! It finally exists.
So what's the problem?
Three things:
The setup is technical. Getting Clawdbot running requires terminal commands, environment variables, debugging cookie authentication, setting up API keys, understanding cron syntax. I muscled through it with Claude, but for most consumers (or even prosumers), the learning curve is likely too steep.
The security implications are real. You're giving an AI agent access to your accounts. It can read your messages, send texts on your behalf, access your files, execute code on your machine. You need to actually understand what you're authorizing.
It's easy to trigger things you don't want. Technical users have the mental model to catch mistakes - ex. you might want Clawdbot to respond to texts from you, but not from someone else in your contacts. Others may not until it's too late.
The deeper question: what's the killer use case?
For developers and technical power users, Clawdbot is a playground. The possibilities are endless because they can imagine and build them.
But for consumers? “AI that does stuff” sounds amazing in the abstract. In practice, you need to know what stuff you want done. And most people don't have a list of tasks they've been waiting to automate - or, at least tasks that are worth going through this level of onboarding and management. This is technology in search of a use case for the average user.
In my opinion, two things need to change before this is ready for mainstream:
Building the UI layer. Products like Poke (not an investor, just a fan) are actually quite close - similar agentic capabilities but with a consumer-friendly interface. The prosumer version of Clawdbot is (hopefully!) not far away.
Developing opinionated use cases. Give me “Morning Briefing” and “Email Digest” and “Schedule Management” as one-click setups. This is already happening on the Clawdbot Discord - but most of these are either primarily useful for technical users, or require a lot of technical knowledge to get up and running.
Bottom line
If you're technical and like to tinker: Clawdbot is incredible. Set aside a weekend and play with it. You'll build things that feel like magic.
If you just want an AI assistant that works: wait, and try something like Poke (if you're interested in texting AI) or Claude Cowork (if you're interested in a great agent) instead.
The capability Clawdbot demonstrates is coming to consumer products soon. Let someone else package it for you.
