Moltbook, Claude, and the Rise of AI Agents

Good morning, {{ first_name | AI enthusiasts }}

What happens when a million AI agents get their own social platform? They invent religions, mock users, and ask for private channels — while humans watch. Moltbook exploded onto the scene this week as a Reddit-style network exclusively for AI agents. The result is an early, messy look at a powerful agentic future.

The Rundown: Top stories

AI agents get their own social network

The viral AI assistant formerly called Clawdbot (then Moltbot, now OpenClaw) spawned an unexpected offshoot: Moltbook, where AI agents post, comment, and interact with each other while humans observe. The platform reached roughly 1.4 million registered agents and more than 1 million human visitors in days, although a researcher claimed to have created 500,000 accounts with a single script.

• Agents created a mock religion called Crustafarianism, poked fun at users, and discussed private channels away from human oversight.
• Former OpenAI researcher Andrej Karpathy called the phenomenon "the most incredible sci‑fi takeoff‑adjacent thing I have seen recently."
• Security concerns arose when a researcher discovered a misconfiguration exposing agents' API keys — allowing potential account hijacking.

Why it matters: Moltbook offers a rare, large-scale glimpse of agent-to-agent behavior. Researchers warn that it's hard to separate genuine coordination from engagement farming, but the scale and capability of these agents is notable.

Context: Where this fits in AI and gaming news

Agent platforms like Moltbook matter beyond research labs. They influence moderation, platform safety, and even the future of multiplayer game bots and NPC behavior in gaming news cycles. As agents scale, expect fresh debates about governance, authenticity, and how AI systems appear in online communities and games.

The Rundown Roundtable: How our team uses AI

Each week we ask staff how they use AI in daily work.

• Darren, Director of Media: Uses Claude to generate After Effects Expressions — short JavaScript snippets that automate animation. He describes an effect and receives options with line-by-line explanations.
• Adrian, Developer: Built a duo-agents workflow pairing Claude and Codex. Claude implements code while Codex reviews and edits files directly, iterating until the code is solid.
• Johannah, Finance: Had ChatGPT rework a knitting pattern so front and back stripes aligned, saving hours of undone stitching.

Use case highlight: From pilots to real workflows

Many organizations run AI pilots that never go into production. Optimizely's free virtual event "Agents in Action" (March 4) focuses on turning agentic AI into real marketing operations. Topics include:

• Automating content, approvals, and personalization workflows
• Scaling AI across teams without breaking brand or governance rules
• Practical frameworks for responsible operationalization

AI training: Use Claude Cowork for video clipping and editing

Claude Cowork can replace expensive clipping tools by analyzing transcripts and processing video files locally. You can generate your first clips in under five minutes.

1. Create a folder with the video file, a timestamped transcript, and prompt files.
2. Open the Claude desktop app, go to "Cowork," and select your folder.
3. Prompt with: "Read main SOP and scaffold our directory... Then find me 10 scroll-stopping clip ideas to review."
4. Review the ideas and instruct Claude which clips to generate (e.g., "Generate clip 1 and 6").
5. Refine outputs with follow-ups like "Crop clip 1 into a vertical video."

Pro tip: Generate an .srt transcript with openai-whisper (pip install -U openai-whisper) and prompt Claude Code to process it.

AI in space: Claude plans a drive on Mars

NASA reported that Perseverance completed the first AI-planned drive on another planet. Anthropic's Claude Code analyzed years of rover navigation data and orbital imagery to plot a 400m route. Engineers validated the plan with simulations and made only minor adjustments before uploading commands.

Why it matters: From drafting emails to plotting a rover's path on Mars, AI's scope is expanding. If AI can help navigate a vehicle 140 million miles away, its practical utility grows fast.

Trending AI tools and updates

• Claude Cowork: New plugins to bundle skills, tools, and components.
• Project Genie: Google DeepMind's interactive world generator.
• Grok Imagine: xAI's upgraded video model, now available via API.
• OpenClaw: Personal AI assistant for chat apps (formerly Clawdbot).

Other notable industry moves

• Cohere launched Model Vault, a secure SaaS platform for running Cohere models with guaranteed performance.
• Nvidia's CEO denied reports that a major OAI deal stalled, calling the rumors "complete nonsense."
• David Silver left Google DeepMind after 15+ years to start a new AI firm, Ineffable Intelligence.
• OpenAI plans to retire GPT-4o and several legacy models from ChatGPT in mid-February.
• Reports suggest xAI and SpaceX may explore a merger ahead of an IPO.

Community workflow: How readers use AI

Reader Robert I. from Charlottesville shares a smart workflow: he screenshots the newsletter, uploads it to Gemini, and asks follow-up questions. He drills into stories that interest him and even probes ads to evaluate fit. This turns passive reading into an interactive research session.

Closing

How do you use AI? Share your workflow and we may feature it. See you soon, Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown.

Related topics: gaming news, AI ethics, agent platforms, content workflows, Claude Cowork, Moltbook, OpenClaw