SPLIT 1 of 1 taken: Fable wrote a Python program that solved a hard board routing problem for his hardware product (firmware diagnosis episode in same comment skipped, analysis only).
Runs LLMs to audit the math in claimed arXiv solutions for his ranked open-problems list; audits already found six faulty proofs that had caused problems to be wrongly marked resolved.
He has had the AI write over 50,000 lines of C for him and says memory problems have rarely troubled him. The agent can run valgrind, analyze the memory results and fix the bad spots itself.
The open-source 'burner' project gives the AI a real spare Android phone (Android 11+). The agent uses the phone's apps exactly like a human would — opening them, tapping, typing, scrolling — and sends the user screenshots, extending its reach to marketplaces and shopping apps with no website or API.
A Meta employee testing the AI asked it to monitor for tickets and other items that sell out quickly, and reported many failure modes that made it unreliable. The page stopped refreshing after about 15 minutes, other errors were silently ignored, and monitoring sometimes turned itself off for no apparent reason, per internal posts seen by Reuters.
This one went wrong. See what happened before you try it.
Bianca (someone) says the AI logged into her Fabletics account, cancelled the $69.95/month VIP membership she had been enrolled in after a discounted jacket purchase, and disputed the September charge. Twenty-one minutes later she had a full $69.95 refund to her original payment method, not store credit.
NYT reporter Eli Tan, reviewing the AI over two weeks with connected bank accounts and email, had it call his dental insurer Anthem. The AI supplied his member ID, answered the security question, waited on hold, then rang him through to the live representative.