Peter Yang Tests Grok Bot’s X Integration, Email and Scheduled Reports

Grok Bot is my favorite personal agent at the time of this video. I test its X integration, an email address and a change to its default model. My examples include research dashboards, recurring reports, a customer-support bot and a generated music video.

I tag the bot in the replies to a post about television shows and ask it to identify underrated and overrated recommendations. Unlike a public Grok reply, the mention starts a task in my Grok Bot account. I can follow up privately and ask it to add a show to my personal taste file.

I keep a file of films and shows I have rated, then use it to request recurring recommendations. I also describe the X access limits available in my test. For me, access to posts is useful for research, though the private follow-up does not add a visible answer for other people on X.

For one research task I ask for a sentiment analysis of two personal-agent products, then request an interactive comparison page. It displays sentiment over time, share of conversation, criteria and links to posts used in the analysis. The results are the bot’s interpretation of the data it retrieved.

I ask the bot to analyze my own posts, sorting them by impressions and video views and breaking them down by format and topic. A recurring version could help me examine which subjects to make more or less often.

For video ideas, another report combines X posts with YouTube channels I monitor. A matrix compares recent social discussion with the number of videos on a topic, and the report links to examples. I use this to look for possible gaps, then inspect the sources myself.

My direct-message test does not work as hoped. The bot can identify some senders but says message contents are encrypted and unavailable; it also cannot pull the contents of message requests. I plan to give the team feedback.

Grok Bot now has an email address. I suggest giving it a distinct assistant name so it makes sense when copied on a conversation. In my scheduling test, it checks availability and drafts a reply for a proposed meeting.

The bot asks me to confirm the message and invitation rather than completing the whole booking without a prompt. I would prefer less prompting for this kind of task, although the prepared reply and calendar action are still useful to me.

I find recurring emails more useful. My examples include a business report using connected channel, banking and course data; a weekend activity report using local sources; and a health report using my own connected measurements. I read email more consistently than separate chat threads.

My weekend planner runs every Friday and uses instructions about nearby towns and my viewing preferences. I refined those instructions with the bot rather than writing them all at once. A market-research report can also arrive by email.

I reply to a research email and ask the bot to draft a video outline in my project-management tool. It creates a draft with a hook and outline. I still plan to review it manually before deciding whether to make the video.

Multiple bots can participate in one thread. I ask my video producer and adviser bots to resolve overlapping research reports; after a discussion, they propose a consolidated approach. I use different personas for recurring jobs.

I say the service now uses an Anthropic Opus model by default and cite a friend’s comparison in support of my preference for it. I then ask the bot to make a short rap-battle video from linked social posts, with generated music. I review lyrics and iterate on the voices.

A community-support bot reviews member requests daily, including billing and membership issues. I say it can process some requests and show a draft response to a request for a lifetime plan that asks for my approval.

I think of my bots as organized helpers for support, video research, family planning and health. In my experience the combination of connected research, email and distinct roles makes the product useful, while the failed direct-message and scheduling examples show areas I would like improved.