I Tested ChatGPT Dot vs GrokBot (Not Even Close)

Products reviewed / compared

The video tests and compares three “personal AI assistant” agents:

• ChatGPT “Dots” (new personal assistant for OpenAI/ChatGPT users)
• Grok Bot (from Grok)
• Hermes Agent (self-hostable / technical agent)

Key features (as described)

Dots (ChatGPT)

• Access / availability: Included with ChatGPT subscription tiers (not “early access” per the reviewer). Appears in the ChatGPT app UI (left sidebar).
• Where it runs: OpenAI cloud (“OpenAI cloud computer”); usable on phone and continues running after you start it (unless restricted).
• Model dependency: Powered by GPT‑6 “Astra” and generally locked to that model (“model-bound”).
• Media / tool use demo: Uses an external media/content tool (Higgsfield API/CLI/MCP) to generate images/videos. Demonstrated creating an image and animating it into a ~15-second video.
• Customization UI: Click its face to change its look (interactive avatar/emoji).
• Learning / improvement: Improves based on feedback in the OpenAI ecosystem (reviewer claims it gets better with preferences).
• Pro limitations / regions: Not yet available for Pro in EEA, UK, Switzerland (per reviewer).
• One-per-plan note: 1 Dot per plan (more may appear later).

Grok Bot

• Access / availability: Download Grok Bot and start with a plan (reviewer cites $20 to get it).
• Where it runs: “Single shared cloud computer” running 24/7 in the background.
• Model dependency: Uses the latest Grok model (reviewer implies it’s effectively model-bound like Dots).
• Ease of use: Easy to set up and good for beginners.
• Interface demo: Multiple agent “personalities” can chat with each other in the Grok interface (sometimes “in conflict” for entertainment).
• Hard limits: Usage/rate limits are the main issue people hit.

Hermes Agent

• Access / availability: Installed by the user and run on local machine/VPS (can be used via Telegram/WhatsApp).
• Where it runs: Runs on your machine (or VPS); can also run in the background if hosted (e.g., reviewer mentions Mac mini for always-on).
• Model independence: Described as model-independent—you can choose any model per task (“run it with Opus… whatever I want”).
• Learning / memory: Permanent memory that you can edit/extend.
• Complexity: More difficult to set up; reviewer emphasizes token issues and ongoing maintenance.
• Pro audience fit: “For geeks/devs” who want control and customization; may be free software, but cost/time depends on infrastructure.

Pricing / cost comparisons (numerical)

• Dots: Requires ChatGPT Pro / subscription at about $100/month (reviewer also mentions $100 / $200 / $500 in some contexts). 1 Dot per plan.
• Grok Bot: $20/month starting price.
• Hermes Agent: $0 software cost; you pay for models you actually use (plus hardware/time).

Pros (from the reviewer)

Dots

• Convenient always-on cloud assistant experience (OpenAI cloud; accessible on phone; continues after starting).
• Strong creative media output with Higgsfield (image → ~15-second video).
• Simple to start if you already use ChatGPT.
• Feedback-driven improvement across OpenAI tools.
• Proactive messaging possible (reviewer references proactive behavior examples).

Grok Bot

• Cheap entry (~$20).
• Easy setup and beginner-friendly.
• “Multiple agents” interface to provide different angles (entertainment + finance example).
• Runs in the background via cloud computer (24/7).

Hermes Agent

• Maximum control / customization
  • Choose model per task
  • Local/offline options (when hosted on your machine; reviewer claims it can run without internet)
  • Editable permanent memory
• Runs anywhere via integrations (Telegram/WhatsApp mentioned).
• Best fit for building systems and complex workflows.

Cons / limitations (from the reviewer)

Dots

• Geographic restriction for Pro in EEA/UK/Switzerland.
• Model lock (“model-bound”).
• Only one Dot per plan (scaling limited by plan structure).
• Still maturing / behind Grok Bot (reviewer: “a couple of weeks behind”).

Grok Bot

• Hard limits / rate limits are common pain points.
• Quality depends on model and formatting, which can be imperfect.
• Cloud model is effectively locked (no real model choice claimed).

Hermes Agent

• Harder setup (install/update; token management).
• Might not be economically feasible without the right hardware—also “you pay with your time.”
• More advanced audience; not as good for casual beginners.

User experience & task performance (examples)

• “Create images/videos” test with minimal instructions
  • Dots: Clean image-to-video animation; produced a 15-second explainer-style video.
  • Grok Bot / Hermes: Also used Higgsfield to produce creative output (specific outcomes vary by test).
• Podcast editing test (Psych Podcast episode)
  • Dots: Adds logo well, but formatting off; likely needs minor guidance.
  • Grok Bot: More “traditional” editing but text missing at bottom in the reviewer’s example.
  • Hermes: Output includes logo, but overall quality didn’t impress the reviewer (depends on requests/model; didn’t match expectations).
• Tool use via connectors
  • Reviewer recommends Zapier as an authentication layer, so connections can be reused across agents without re-authenticating repeatedly.
  • Demonstrated: agents find the last email and draft a reply.
• Strategy task comparison
  • Dots: ~5-page voluminous document.
  • Grok Bot: More visual overview.
  • Hermes: Different ideas with more detail (reviewer notes quality depends on setup/model).

Comparisons summarized (who wins for what)

• Easiest to start / works “right away”: Dots or Grok Bot
• Best budget entry: Grok Bot (~$20/month)
• Beginner-friendly cloud experience: Grok Bot or Dots
• Most control / customization / model choice / local run: Hermes Agent
• Maturity expectation: Reviewer expects Dots to improve rapidly, but says it seems less mature than Grok Bot.

Overall verdict / recommendation (concise)

• Start with Grok Bot or Dots if you want simplicity and quick results.
• Choose Grok Bot if you’re budget-focused and want easy cloud setup.
• Choose Dots if you’re already in the ChatGPT ecosystem and want an always-on assistant experience (with notes on regional availability and model lock).
• Choose Hermes Agent only if you want developer-level control (model choice, local/offline use, editable memory) and can handle setup/time cost.

Unique points mentioned (deduplicated)

• Dots
  • Access tied to ChatGPT subscription; appears in the ChatGPT sidebar.
  • Runs on OpenAI cloud computer, continues in background; usable on phone.
  • GPT‑6 Astra-based and model-bound.
  • Can integrate with Higgsfield API/CLI/MCP for image/video creation; produced a ~15-second video.
  • UI avatar/face customization.
  • Podcast editing: logo added but formatting may be off.
• Grok Bot
  • ~$20/month; cloud computer runs 24/7.
  • Common pain point: hard limits / rate limits.
  • “Multiple agents” interface can show conflicting personalities for humor.
  • Podcast example: may miss text at bottom.
• Hermes Agent
  • Self-hosted on machine/VPS; can integrate via Telegram/WhatsApp; possible always-on hosting (e.g., Mac mini).
  • Model-independent and allows model choice per task.
  • Permanent editable memory.
  • Harder setup: install/update and token issues.
  • Can (per reviewer) run without internet when hosted locally.

Speaker views / roles

• Single main speaker/reviewer (“Jack”): runs the tests, provides pricing context, compares behaviors, and delivers verdicts.
• “Bots” mentioned (e.g., weight-loss bot example) are used as illustrative proactive engagement examples, still from the reviewer’s observations.