In partnership with

This week’s newsletter hits close to home. Over three years ago, I started using ChatGPT to help me automate pre-production of the now-retired CryptoSapiens podcast. From there, it helped me ship the podcast newsletter every week.

The conversation around personal AI has been stuck on chatbots since then. You type, it answers.

That model is now being replaced by something different: agents that sign into your tools, take on a task, and keep working after you close your laptop.

OpenAI is building one inside Codex that stays active until you explicitly shut it down. xAI already shipped one. It's called GrokBot, and if you're on X, you've probably seen your feed talking about it. Both products are making the same bet: the most useful AI is the one that doesn't wait to be asked.

This week we look at what that shift actually means, what GrokBot specifically does for creators on X, and what a creator-built version of this looks like today using tools you already have access to.

With ❤️
Humpty Calderon

From The Trenches

OpenAI is building an AI agent that keeps working after you walk away

Code reviewed by WIRED shows OpenAI is developing a persistent AI agent inside Codex, its coding-focused AI product, that continues working autonomously across sessions until explicitly deactivated. The internal description for this mode is being 'put to sleep.' This is not a chatbot that waits for your next prompt. It is an agent that sustains itself, picks up where it left off, and keeps executing until you tell it to stop.

The productivity case for solo creators is obvious: a system that drafts, schedules, follows up, and iterates on your behalf while you're doing something else. The risks are just as obvious. An agent acting without prompting can take actions you didn't intend, rack up API costs, or produce output you'd never have approved if you'd been watching. Anthropic, Google, and Microsoft are all racing toward the same destination.

The shift from 'AI that answers' to 'AI that acts' is the most significant change in how creators will work over the next 18 months. OpenAI just showed us the internal architecture for what that looks like.

GrokBot gave every X user a teammate that works while they sleep

xAI shipped GrokBot in beta on August 11. The pitch is straightforward: each bot gets its own cloud computer, signs into the tools you already use, and keeps working after your device is off. It only comes back to you when a task needs a human decision. Bloomberg covered the launch as xAI's most direct move yet to compete with OpenAI and Anthropic on agentic AI. It’s not just a better chatbot, it’s a product that does work.

What makes this relevant specifically for creators on X: in late August, xAI added direct X account integration into GrokBot. Bots can now post, read your timeline, and respond to mentions on your behalf, with your X API credits included in paid plans. That means a GrokBot configured for your content operation can monitor conversations, flag relevant threads, and draft responses — all without you being in the app. xAI also expanded access twice within 16 days of launch, dropping the entry plan from $300/month down to tiers starting at $100/month with a seven-day free trial.

The conversation about GrokBot is largely happening inside X itself: threads, replies, and video walkthroughs from creators already in the beta. When the most active discussion about a tool lives on the platform the tool integrates with, the feedback loop between product and audience is unusually fast.

One voice memo, one week of content: a creator's AI workflow that works today

While OpenAI and xAI are building autonomous agents that act without prompting, a practical version of AI-powered content delegation already exists. The system uses Claude to turn a single voice journal entry into platform-native content across formats: threads, newsletters, video scripts, and carousels, produced simultaneously from one raw input.

The mechanics are specific. Claude's 200K context window holds your full brand guide alongside the raw voice note, so every output reflects your voice rather than generic AI prose. One prompt, multiple finished assets. Creators publishing on three or more platforms grow audiences two to three times faster than single-platform peers.

This is the creator version of what OpenAI and xAI are building at the infrastructure level. You do not need to wait for always-on agents to start delegating the formatting, the repurposing, and the distribution prep. The workflow exists now.

200+ Claude Prompts Top Professionals Actually Use at Work

Claude can be your analyst, editor, and strategist.
But most professionals are using it to fix grammar.

These 200+ Claude prompts take it from grammar tool to your most powerful AI work assistant.

Sign up for Superhuman AI and get:

  • 200+ ready-to-use Claude prompts to get real work done in minutes — researched, tested, and used by professionals at Google, Microsoft, and NASA

  • Superhuman AI newsletter (4 min daily) so you keep learning new AI tools and skills to stay ahead in your career — the prompts are just the beginning

Tool of the Week

Gemini 3.5 Transcribe: Google's transcription model built for accuracy, not speed

Google released Gemini 3.5 Transcribe as a standalone model purpose-built for audio transcription. For creators producing podcasts, long-form video, or voice notes that feed into a writing workflow, a dedicated transcription model from Google means faster turnaround, better accuracy on industry-specific vocabulary, and a direct pipeline into Gemini's broader ecosystem. If you're still running your audio through a general-purpose AI tool and cleaning up the transcript by hand, this is worth testing against your current setup.

Questions for our readers

  • If an AI agent could run one part of your content operation without you prompting it, what's the first thing you'd hand over? What would you never give up?

  • GrokBot now has direct X integration. Does that change how you think about building on X as a platform?

  • The creators who figure out AI delegation first will have a real operational advantage. Are you building toward that, or still experimenting?