His Aug 20 walkthrough is a demo catalog. It is not a lab evaluation, and “feel like cheating” is packaging.
11 Grok Bot Use Cases That Feel Like Cheating · Matthew Berman
What the video shows
Chapters map a personal operator stack: email agent, calendar agent, browser tasks, coding, DoorDash, meeting summary, chief of staff, personal agent, school agent, computer cleanup, and Telegram. Public descriptions emphasize recurring inbox triage (archive noise, draft replies, avoid auto-send) and browser chores such as returns or appointments, plus a weekly cleanup that classifies disk waste and asks before deleting. Treat on-screen segments as his demonstration on his accounts — not an xAI service-level promise.
Claim vs proof
His working demo (creator evidence): inbox/calendar-style operators, browser task examples, meeting notes to action items, school-mail organization, cleanup with human approval before deletes, and a Telegram bridge he configures with published prompts. Still single-creator results.
Relayed / not his personal finished daily loop: Cursor-team coding orchestration (Bot managing Cursor agents as described by Cursor staff) and DoorDash-style ordering. Do not upgrade those into personal proof.
Nearby company claims: Aug 26 SuperGrok/Cursor plan inclusion lives in today’s flagship. Berman’s prompts are his artifacts, not official xAI documentation. Title hype is not evidence.
Why this matters
Creator demos are how many readers first meet agent products. The disciplined read is pattern-hunting without importing YouTube title energy into the article voice — and without treating one person’s Gmail/Calendar day as a product-wide success rate.
Bottom Line
Useful patterns from one heavy user’s stack — with limits disclosed. Not enterprise reliability, and not personal certification of the Cursor-team or DoorDash loops. For entitlements, use the flagship; for OpenAI model access inside Cursor, use today’s wind-down brief.