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

A user shares that Grok Bot has reset usage limits for all users, with an image that appears to be a screenshot of the quota page.

At DevDay, OpenAI launched Dots, always-on agents living in ChatGPT, Slack and Teams, plus GPT-6.1 Sol, a model it says nearly matches its flagship at one-fifth the price, one day after pulling its next top model over safety concerns.

The author reviews OpenAI's new Dots agent, Space workspace and Sol model, concluding that most businesses should wait before adopting them.

At DevDay, OpenAI introduced Dots, its fifth AI agent product, along with faster developer tools and a new app marketplace, seen as an effort to catch up with Anthropic and newer agent startups.

A user shares a Grok Bot usage screenshot showing the quota fell to 1% within a day after a random free reset, prompting complaints about how fast it is consumed.

The author rounds up 20+ DevDay announcements from September 29, highlighting the always-on Dots agent on GPT-6 Astra, GPT-6.1 Sol at one-fifth the token price, and a 300 token/s Ultrafast mode.

A user shares a Cue workflow: ask the main cue agent to create a set of sub-agents from a description, then start a group chat so they can talk to each other.

A purported Xpass price list shows four tiers from $8 to $200 bundling X, Grok, Grok Bot and Cursor, with the top tier costing less than buying them separately.

The author used one OpenAI dot to run four agents (find bugs, reproduce, write patches, run tests) that cleared 386 issues overnight, 4.8x faster and 4.3x cheaper than Codex working alone.

The author says Muse handled selling unused items, work reimbursements, subscription cancellations and even recovering money owed by their college without complex setup, so they dropped Claude.

Someone posts a screenshot and tags @grok to ask what the new Grok Bot update actually changes.

Citing OpenAI Dots system card data that 8.6% of 5-step chains drift outside the task boundary and 19.7% at 10 steps, the author splits an Astra 6 agent into scanner, analyst, skeptic and executor roles with a forced 5-step checkpoint; of 47 tokens on day one, only 3 survived.