ExpertsFMMaking $$$ with Meta Muse AI
The Startup Ideas Podcast · October 19, 2026
5 Selected Lessons:
heuristic · 2 case · 1 framework · 1 priority · 1
Services arbitrage over physical goods arbitrage
Source
The Startup Ideas Podcast — Making $$$ with Meta Muse AI (October 19, 2026) · at 21:40 ▸
When this applies
when choosing between a transactional resale business model and a recurring trust/matching services business model
The takeaway
When choosing between one-off transactional arbitrage and recurring trust-based matching arbitrage, prioritize the latter because it compounds into an owned audience and repeat business.
Why it works
Services matching builds a repeatable trust relationship and recurring demand (needs recur, e.g. childcare), whereas goods arbitrage is a one-off transactional flip with logistics overhead and lower repeatability.
One level up
Favor business models that compound trust and repeat engagement over those that are purely transactional, even if the transactional model is easier to start.
Where it stops holding
This stops working when the service niche has low frequency of need (one-time only) or when physical arbitrage margins are unusually large and logistics-light.
Detail
hierarchy:
- 1. Services matching via Facebook Groups and Instagram (building trust, audience, recurring demand)
- 2. Physical goods buying/selling via Marketplace (one-off transactional flips)
rationale: Services matching builds an owned audience and trust asset; goods flipping is 'not what excites me the most' by comparison, per the host.
In their words
This is like your Gary V strategy there, where he goes into garage sales and buys and sells things. That's not what excites me the most. I'm more interested in the services component.
Giving an AI agent its own phone number to automate outreach and closing
Source
The Startup Ideas Podcast — Making $$$ with Meta Muse AI (October 19, 2026) · at 10:38 ▸
When this applies
when you want an AI agent to move from passive research into active outbound communication and transaction closing
The takeaway
Once an AI agent has identified a list of counterparties, give it its own communication channel (phone/text) so it can autonomously execute the full outreach-to-close loop, not just research.
Why it works
Connecting an agent to a dedicated phone/text line via a cheap third-party service removes the human bottleneck in outreach, letting the agent draft, send, call, and close at scale.
One level up
The value of an automated discovery system multiplies when paired with automated execution — don't stop at insight, automate the action too.
Where it stops holding
This stops working when the target audience requires high-trust human voice interaction that an AI voice agent can't yet convincingly replicate, or when local regulations restrict automated calling/texting.
Detail
procedure:
- Sign up for a low-cost agent phone plan (e.g., Bland AI's AgentLine, ~$15/month)
- Get a dedicated US number with talk and text for the agent
- Prompt the agent to call/text leads directly, draft outreach, and close sales
benchmarks: {"cost":"$15/month for the agent phone plan"}
rationale: This converts a research-only agent into a full-funnel automated agency that can close deals without a human team.
In their words
So you can literally sign up for a Bland AI account. It's I think 15 bucks for agent phone plan. All of a sudden your muse agent, you can say you can prompt your muse agent, get yourself a phone number and call me.
Platforms/pages are undervalued right now because future agents will crawl them
Source
The Startup Ideas Podcast — Making $$$ with Meta Muse AI (October 19, 2026) · at 10:38 ▸
When this applies
when a new automated discovery mechanism (search, crawler, agent) is about to start systematically scanning a category of asset that is currently cheap or ignored
The takeaway
Acquire or build the assets that an imminent automated discovery layer will search over, before the market recognizes their new role as a distribution channel.
Why it works
If millions of AI agents will soon be crawling a specific category of asset (groups, pages, listings) for matching purposes, being an indexed/high-quality source in that category becomes valuable distribution before the crowd catches on and prices it in.
One level up
Position early in whatever becomes the substrate that automated agents query, since being discoverable there becomes the new form of distribution.
Where it stops holding
This stops working when the platform changes its terms of service to block agent crawling or group transfer/ownership, or if agent adoption never reaches the scale needed to make crawled assets valuable.
Detail
rule: Buy or build groups/pages/niche accounts in categories agents are likely to crawl before the broader market recognizes the new distribution value.
threshold: Applies specifically to Facebook groups (~1.8B MAU cited), Facebook business pages, and IG niche pages, per the host.
rationale: If millions of personal agents will search these surfaces for matching needs and supply, owning/controlling them becomes a distribution advantage analogous to owning SEO real estate early.
In their words
So if you believe that we're going to have millions of Muse agents going and crawling Facebook groups, why aren't you creating a Facebook group right now?
AI agents default to safe sources, not the best sources — prompt explicitly for the highest-value niche data
Source
The Startup Ideas Podcast — Making $$$ with Meta Muse AI (October 19, 2026) · at 7:19 ▸ · at 13:48 ▸
When this applies
when using a general-purpose AI agent on a new task and relying on its default behavior instead of explicit instructions
The takeaway
Don't assume a general AI agent's default behavior surfaces the optimal source; the operator-business incentives behind the agent shape its defaults, so you must explicitly instruct it toward the highest-value (often platform-internal) data sources.
Why it works
Platform-provided agents are tuned by their makers (e.g., Meta) to avoid aggressively cannibalizing the platform's own monetizable surfaces (Marketplace, Instagram) by default, so they default to safer, lower-quality, or external sources unless explicitly directed otherwise.
One level up
Any tool built by a party with its own incentives will default to the behavior that protects those incentives — users must actively override defaults to get maximum value.
Where it stops holding
This stops working when the agent's terms of service strictly prohibit accessing certain sources even on explicit request, or when explicit prompting still can't access private/gated groups.
Detail
rule: When an agent's default search misses the richest source, explicitly instruct it to pull from that source (e.g., 'give me night nurses from Facebook groups').
threshold: Observed: default search = public listing sites + IG sweep, no groups; prompted search = Facebook groups, described as 'highest quality'.
rationale: Meta/Zuck is smart — the agent won't cannibalize Marketplace and Instagram by default, so the user has to explicitly direct it there to unlock the best data.
In their words
The default, that's really interesting. So the default is it's not going and sort of cannibalizing Facebook groups, right? But you can prompt it to go and do that. And I'm finding some of the highest quality data is coming from Facebook groups.
Night nurse search reveals agency markup built on Facebook-group sourcing
Source
The Startup Ideas Podcast — Making $$$ with Meta Muse AI (October 19, 2026) · at 2:04 ▸
When this applies
when you discover an intermediary charges a markup for a matching service that is actually just manual search of freely available public information
The takeaway
Whenever a paid intermediary's core service is manually aggregating freely available public information, that service is a target for automation-driven disintermediation.
Why it works
Agencies captured value by doing manual labor (searching Facebook groups) that a human would otherwise have to do themselves; once that search/matching labor can be automated by an AI agent, the agency's fee becomes arbitrageable away.
One level up
Look for industries whose value proposition is just 'we did the tedious search for you' — automation collapses the cost of that labor and opens arbitrage.
Where it stops holding
This stops working when the intermediary's value is really trust, vetting, or liability (not just search), which an AI agent cannot easily replicate.
Detail
situation: The host and his wife had a baby and needed a night nurse; going through an agency, he learned the agency was simply trawling Facebook groups (e.g., Miami, South Florida, New York night nurses) and charging a fee for matches.
decision: The host realized he could replicate this sourcing process himself using a personal AI agent (Muse) instead of paying the agency's fee.
outcome: This realization became the seed for the host's broader thesis that personal AI agents can disintermediate many manual-matching service businesses.
lesson: A paid matching service may just be manual aggregation of public data — a vulnerability to agent-driven automation.
In their words
The agency basically admitted that they were going through Facebook groups and looking at, say, Miami night nurses or South Florida night nurses or New York night nurses and they would bring them to here. And I was like, that's crazy that I would pay a fee for that when I could use Muse to go and do that in today's day and age.