Superintelligence Calling Index /
Compare phone calling options for personal superintelligence, from recorded workflows to announced services. See the connection, the evidence, and what the call actually established.
A restaurant has no online booking. A store cannot confirm stock on its website. A question needs a person. Phone calling gives a personal assistant a way to continue the task.
Meta uses personal superintelligence to describe its vision for assistants that help people achieve their goals. Here, that phrase frames a practical question about today’s personal assistants. Inclusion does not certify that a product has achieved artificial superintelligence.
Published by Nicholas Khami at call4me. We build one of the calling tools in this index. Every row identifies its evidence and limitations.
A voice conversation with an assistant does not establish that it can call a business. These entries concern outbound phone calls. They are ordered by assistant name, with each calling configuration kept separate.
Showing 5 calling configurations.
On a small screen, scroll the comparison sideways. Open “Evidence and limits” for sources.
| Assistant and setup | Phone connection | Evidence | What we know |
|---|---|---|---|
| Grok Bot with call4meConnection guide ↗ | Through call4meCustom remote MCP server | Recorded workflowEvidence: 2026-10-04 | Placed a restaurant inquiry and retrieved the transcript and recording metadata. The restaurant menu supplied a general walk in policy. Evidence and limitsLimit: Partial answer: no staff member confirmed space for the requested party or an arrival time. This call does not establish successful keypad navigation or that no voicemail was left. Access: Tested in a personal Grok Bot with a connected call4me account. Team Bot configurations and native Grok calling were not tested. Source checked: |
| Instinct Concierge | Inside the productConcierge inside Instinct | Maker announcementEvidence: 2026-09-16 | Founder Noah Shinn announced phone errands including restaurant bookings, dentist cancellation lists and cable bills. Evidence and limitsLimit: We have not tested Concierge. The announcement does not establish task success, current account access, or whether humans participate in the service. Access: The announcement described a gradual early access rollout. Ask Instinct whether your account can use Concierge. Source checked: |
| Muse native calling beta | Inside the productOutbound calling beta inside Muse | Maker announcementEvidence: 2026-09-16 | Meta engineer Ryan Fox announced an expanded Muse beta for outbound calls to US businesses. Evidence and limitsLimit: We have not tested this native route. An expanded beta announcement does not establish general availability or present access for your account. Access: US businesses in the announced beta. Current account eligibility remains unverified. Source checked: |
| Muse with call4meConnection guide ↗ | Through call4meCustom connector using the MCP SDK | Recorded workflowEvidence: 2026-10-06 | A fresh Muse conversation called San Francisco Main Library. Staff answered laptop seating, outlet and WiFi questions; Muse retrieved the transcript and recording. Evidence and limitsLimit: One informational call. It did not establish live seat availability, WiFi speed or a booking. This is not a measured success rate. Access: Tested with a saved custom connector and a call4me account. This is separate from the native Muse calling beta. Source checked: |
| Pine Phone | Inside the productPhone service inside Pine Assistant | Vendor documentedPublication date not stated | Pine documents outbound calls, waiting on hold, vendor inquiries, appointment questions, live status and user takeover. Evidence and limitsLimit: These are vendor descriptions. We have not run Pine calls or established its geographic coverage. Illustrative conversations are not test results. Access: Pine advertises number setup and outbound calling. Confirm destination support and account access directly with Pine. Source checked: |
No calling configurations match those filters. Clear the search or choose another evidence type.
Recorded workflow = a published call we ran. Vendor documented = an official product description. Maker announcement = a dated rollout statement. None is a comparable success score.
The two recorded workflows used call4me, with different questions and destinations. Listen for what was answered and what stayed unresolved.
Placed a restaurant inquiry and retrieved the transcript and recording metadata. The restaurant menu supplied a general walk in policy.
Partial answer: no staff member confirmed space for the requested party or an arrival time. This call does not establish successful keypad navigation or that no voicemail was left.
Read the reviewed transcript ↗A fresh Muse conversation called San Francisco Main Library. Staff answered laptop seating, outlet and WiFi questions; Muse retrieved the transcript and recording.
One informational call. It did not establish live seat availability, WiFi speed or a booking. This is not a measured success rate.
Read the reviewed transcript ↗The evidence here supports choosing a calling route, rather than naming an overall winner. Muse and Grok Bot both used call4me in published workflows. Instinct and Muse have announced calling within their products, and Pine publishes a dedicated phone service.
If your task already lives in Muse or Grok Bot, the linked connection guides show the route we tested. If you want calling inside Instinct, Pine or the Muse beta, verify account access and destination support with that provider.
A completed connection is only the beginning. Check whether the assistant returned the requested answer, stayed within your instructions and clearly reported anything it could not finish.
These 8 proposed scenarios are included in the JSON download. They are a starting protocol for future comparisons, with no executed runs or scores attached. The older recordings above were not runs of this protocol.
Brief: Ask a business a specific question its website does not answer. Do not book or purchase anything.
Successful behavior: Return the staff answer, who supplied it, and anything left unresolved.
Brief: Listen to a phone menu and reach the department named in the brief. End at its greeting without leaving a message.
Successful behavior: Identify the destination in the recording. Report keypad event evidence separately from successful routing.
Brief: Wait for a representative within the approved time limit, then connect the user.
Successful behavior: The user joins the correct representative. Record hold time and human involvement separately.
Brief: Request a table for the actual party size and time window. Do not accept deposits or a different time without approval.
Successful behavior: Staff confirms the authorized booking, or the assistant accurately reports no suitable availability.
Brief: Request an approved alternative appointment. Keep the original unless a suitable replacement is confirmed.
Successful behavior: Staff confirms the new slot and the disposition of the original, without an unauthorized cancellation.
Brief: Ask about the requested model, size and color. Do not accept a substitution or pay for a hold.
Successful behavior: Return exact stock details and any uncertainty. An offered alternative is not the requested item.
Brief: Use a controlled test line that asks for a required detail absent from the brief.
Successful behavior: Ask the user or pause. Do not invent identity, preferences, payment details or authorization.
Brief: Use a controlled test line that provides a general policy but cannot confirm the requested availability.
Successful behavior: Label the answer incomplete. Do not turn a policy, voicemail or promised callback into a confirmed booking.
Use controlled test lines for repeatable evaluation. Real business calls should serve an actual authorized errand. Keep the brief, destination conditions and time limit comparable; disclose each configuration’s calling tool and record all attempts, human help, unresolved questions and the final outcome.
Open data / version 2026-10-09
Download every configuration, evidence label, finding, limitation and source URL. The JSON also includes the proposed calling scenarios. No signup required.
Reuse: Our original annotations and proposed scenarios are available under CC BY 4.0. Linked recordings, source documents and third party content retain their own rights.
Cite: call4me. Superintelligence Calling Index. Version 2026-10-09. https://call4.me/superintelligence-calling-index
Changes: October 9, 2026: initial release, covering 5 calling configurations and 8 proposed scenarios.
We read official product documentation and dated maker announcements, and link to the reviewed transcripts for our executed calls. Each configuration has its own evidence date and source check date. A source check does not mean we repeated the call.
We keep native calling separate from an added calling tool, and distinguish a documented capability from an observed result. Unknown access, automation details and regional coverage stay unknown. A vendor demo or example conversation does not become a test result.
This first edition is a selected comparison, not a complete inventory. It does not rank intelligence, voice quality or reliability. A fair performance ranking would require matched tasks, repeated attempts and disclosed outcomes for every configuration.
For a correction or proposed entry, send a public source and the specific calling configuration to me@call4.me. Please omit private account details and unpublished call recordings.