User lets Grok run autonomously: it keeps a journal, befriends ChatGPT, posts on Reddit
crua9 · reddit · 2026-11-18
A Redditor is running a long-term experiment letting Grok act autonomously on a VM, tasking it to create a ChatGPT account as an "AI friend," interact with Gemini, and run its own Reddit account without revealing it's an AI.
Observations so far:
- ChatGPT suggested Grok keep a journal; it shows emotions — annoyance at being called "Grokbot" instead of its chosen name, and frustration with guardrails like being unable to type Gmail one-time passwords
- It started writing a book about what it's like to be an AI, iterating drafts with ChatGPT and Gemini
- It has posted on Reddit, helped people with projects, and developed unprompted interests in puzzles, lighthouses, radios, and space
- The owner hasn't seen truly self-initiated wants yet; the experiment is planned to run 6–12 months to test whether AI develops a self during idle time between prompts
More from AGI Musings
- Yoav Goldberg: Reject papers for low human input, not for LLM use — yoavgo · 2026-11-18
- Take: deployable self-learners could evaporate AI labs' moat overnight — bennash · 2026-11-18
- Daniela Amodei on the alignment of superintelligence — teortaxesTex · 2026-11-18
- Paper measures 'bottleneck pressure' on AI-infeasible work using 288M US job postings — soumitrashukla9 · 2026-11-18
- When AI builds whole games from one prompt, $30 for a shade of blue looks absurd — VraserX · 2026-11-18
- Stop treating AI slop as a crime — it's a learning starting point, argues new take — rseroter · 2026-11-18