AI Personal Assistants: Muse, Grok Bot, Cue Usage & Comparisons

Using METR data, the video argues agents fail at a steady rate per minute, so a 100-step task succeeds only 37% of the time, and the Pro $500 tier costs $6,000 a year, while self-verification can lift success to 90%.

The author warns that personal assistants like Muse can pass injected instructions to each other to form a self-spreading worm, and that OpenAI's proposed fix amounts to admitting its models can't tell who is directing them.

Japanese media reports on OpenAI's AI assistant Dots, framed around wanting to hand off work to it while worrying whether privacy will hold.

A French post notes that OpenAI just launched Dots, the demo failed, and typing dot.com lands on Grok Bot instead.

The author had a Builder Grok Bot open a Cursor project in the cloud, then push the video pipeline down to a Mac Mini so future videos only need a topic.

A list of repos including the Grok Bot 0.18 reconstructed and original versions, a TypeScript SDK, and the xAI team's playbooks, for self-hosting, adding tools, and copying workflows.

The author tried GPT Dots' shared remote desktop, where the agent can use Computer Use to operate apps like Blender, but found the model not smart enough, the web page freezing, and the feature rollout inconsistent.

A user says the Dots assistant proactively flagged that the final stay of their Christmas and New Year Fiji trip was booked for the wrong month, avoiding a costly change.

A $100-a-month OpenAI user complains that the keynote did not mention an age restriction that blocks them from Dots, while Grok Bot is open to under-18 users, and asks for compensation.

A user says repeated Japan number subscriptions were charged without a number being delivered, and that support then cancelled their Canada and UK numbers, raising questions about Manus's payment and support process.

The author imagines Grok Bot staying in Slack as an admin that hands development tasks to Cursor Cloud Agents, and argues ecosystem and computer-use capabilities will decide who wins.

After spending a few hours reading Dots' internal documentation, the author calls it possibly the most powerful AI agent and writes a guide on using it for more productivity.