Sam Altman: How to Make an Abundant Future
Invest Like the Best with Patrick O'Shaughnessy
Build an interview brief from a guest's previous podcast appearances.
Find what your guest repeatedly discusses, what they have already answered, and where a better follow-up can move the conversation forward.
How it works
Turn previous appearances into an interview brief.
The agent handles the research sequence. You choose the guest, the angle, and which evidence belongs in the final brief.
Input
[guest][interview angle]- 01
Find real guest appearances
Exclude episodes that only mention the name.
Search - 02
Retrieve relevant transcripts
Use the guest's own words as the source material.
Transcripts - 03
Compare the evidence
Separate repeated answers, changing positions and open gaps.
Analysis
Output
Source-backed interview brief themes · previous answers · follow-ups
Just connect Claude to Audiogram, then paste the example prompt
This gives Claude the ability to find and edit podcast audio, effortlessly creating your clips.
step1
Open Connectors in your Claude settings
In Claude on the web, click your initials in the bottom-left corner and choose Settings. Then open the Connectors section.
Profile Appearance Connectors Data controls
Connectors
Connect tools and data sources to Claude.
step2
Add a custom connector
Scroll to the bottom of the Connectors screen and click Add custom connector. The button sits below the built-in connectors list.
Google DriveConnected
SlackConnect
Add custom connector step3
Paste the Audiogram server URL
Name it Audiogram and paste this as the remote MCP server URL, then click Add.
Name
Audiogram
Remote MCP server URL
https://mcp.audiogramapi.com/mcp
step4
Connect and sign in with Google
Click Connect on the new Audiogram card. A window opens on our sign-in page: continue with Google and approve access. If you're new, this creates your free Audiogram account on the spot.
Audiogram
Claude wants to access your account
Search podcasts, ingest episodes, read transcripts
Continue with Google step5
Start a new chat and paste the example prompt
In any chat, open the tools menu below the input bar and check the Audiogram switch. Then paste the example prompt below, fill in any bracketed variables, and send it.
Replace two variables and run the workflow. [guest] identifies the person. [topic or interview angle] keeps discovery and comparison focused.
Prepare an interview brief for [guest] about [topic or interview angle]. Use Audiogram to: 1. Find Apple Podcasts episodes where [guest] is actually interviewed. 2. Select the most relevant appearances for [topic or interview angle]. 3. Retrieve the available transcripts. 4. Compare the transcripts and return: - recurring themes and examples; - positions that changed or became more specific; - questions the guest has already answered; - five follow-up questions based on gaps, tensions, or unsupported claims. For every finding, cite the podcast and episode. Separate direct transcript evidence from your own inference. If there are not enough relevant transcripts, say what is missing instead of filling the gap.
Example output
Sam Altman: will AI create abundance or concentrate power?
Built from two available transcripts retrieved with Audiogram. The sources and the useful gaps are shown below.
Sources retrieved
Sam Altman on God, Elon Musk and the Mysterious Death of His Former Employee
The Tucker Carlson Show
Both transcripts returned ready. The second contained speaker-label artifacts, so individual-line attribution requires care.
- Recurring themes
- AI power concentration, personal agents, privacy, work and governance.
- What not to repeat unchanged
- Broad questions about whether AI will be beneficial or how quickly capability will improve.
- Where to press further
- Measurable evidence of broadly shared benefit, privacy limits for personal agents, and independent review of model-behavior decisions.
- What measurable evidence would show that AI abundance is being shared broadly rather than concentrating power?
- What privacy threshold should an always-on personal AI agent never be allowed to cross?
- Who should independently review decisions about how a model behaves, and what should that review make public?
Before you run it
Coverage and quality checks.
Does this work for every guest?
No. It depends on the guest having relevant published Apple Podcasts appearances and transcripts Audiogram can retrieve.
Can it compare a single episode?
One episode can be analyzed, but claims about recurring themes or changing positions require multiple relevant transcripts.
Does this process an unpublished recording?
No. This prompt is for research across published podcast episodes, not raw audio editing or transcription.
What should be checked before recording?
Confirm the guest is actually speaking, review unclear speaker labels, and verify any quotation you intend to use publicly.
Use this prompt
Start with one guest and one focused angle.
Connect Audiogram to a supported agent, paste the prompt, and replace the two variables.