Nov 3, 2026 · 1h 30m
AI agents face a commerce test: value, cost and capability
$AAPL + $PLTR, Value of a Search, Grok Bot Breaks the Bank | Harley Finkelstein, Bridgit Mendler, Dylan Patel & Max Kan & Andrew Megalaa
The episode connects agentic shopping, expensive inference, AI benchmarks and space infrastructure to the harder question of how useful AI becomes economically.
- 1Shopify is preparing merchants for agents that can discover products, interpret details and complete purchases autonomously.
- 2GrokBot’s usage limits expose the tension between frontier-model demand and the high cost of serving every inference.
- 3GDPX argues that AI’s commercial breakthrough depends on superhuman performance in valuable domains, not simple human equivalence.
Don't miss
The panel argues that GDPX should measure AI against valuable specialist work and teams, making superhuman economic performance more relevant than human equivalence.
The brief
The episode moves from Apple and Palantir earnings to Google’s AI search strategy, asking whether changing discovery also changes advertising, engagement and the structure of the web.
Harley Finkelstein explains Shopify’s agentic-commerce push, where Shop Pay and structured merchant data let personal agents move from product discovery toward delegated checkout.
GrokBot’s costly inference becomes a case study in AI economics: personal agents may subsidize usage only if compute costs fall sharply as infrastructure scales.
Bridgit Mendler outlines Northwood Space’s effort to expand satellite ground infrastructure, while SemiAnalysis researchers use GDPX to distinguish impressive model tasks from economically valuable capability.
The episode’s central tension is whether AI progress will arrive as visible frontier-model breakthroughs or as quiet gains in manufacturing, science, engineering and commerce.
What was said on this episode
27 statements · 17 positive · 5 negative · 1 mixed · 4 neutral
Apple reported record September-quarter revenue and Palantir nearly doubled year-over-year revenue.
“Apple reported record September quarter revenue, while Palantir nearly doubled its revenue from a year ago.”
Listen at 2:42
Palantir’s U.S. commercial revenue exceeded U.S. government revenue in Q3.
“U.S. commercial revenue surpassed its U.S. government revenue for the first time, reaching $984 million versus $953 million.”
Listen at 3:53
Google can remain strong despite declining search revenue because of its other assets.
“Even if search revenues decline every year and we're sort of at peak search, they just have so many other assets”
Listen at 8:28
Anthropic and OpenAI account for 86% of enterprise LLM API spending.
“Anthropic and OpenAI now account for 86% of enterprise LLM API spend up from 67%.”
Listen at 9:00
Google has substantial upside if it remains competitive in cloud and chips.
“as long as they can. Stay competitive around the cloud and chip business. You know, there's a lot of upside.”
Listen at 11:03
Google will not provide a solution for learning without AI search.
“Google is not going to save you at this point.”
Listen at 12:47
xAI internally adjusted GrokBot’s usage economics.
“they kind of right-sized the tokenomics internally.”
Listen at 15:30
Airplane Doppler signatures can potentially triangulate a speaker’s location.
“you can triangulate where I am.”
Listen at 17:55
Long-form outdoor vlogs can expose a creator’s location through audio analysis.
“this definitely could dox your location.”
Listen at 19:01
Rowdy is likely to become a successful snack brand.
“you can just tell already that it's going to be a hit.”
Listen at 22:39
Shopify catalog traffic converts at twice the rate of general AI search.
“conversion using Shopify catalog versus general AI search, which is effectively scraping it's two X higher.”
Listen at 28:29
Shopify Shop Pay enables agents to provide better recommendations using shopping history.
“your entire shopping history The agent now understands they give way better recommendations”
Listen at 29:35
AI agents will surface products based on product merit and disclosed attributes.
“the agent is going to surface whatever is the most the most merit based product.”
Listen at 34:18
Shopify expects agentic shopping may eventually include advertising.
“we're not assuming that agentic shopping stays ad-free.”
Listen at 38:48
Third-party satellite ground infrastructure has historically lacked high uptime.
“historically, you know, third party ground has not been able to offer.”
Listen at 44:33
Every hard-tech company is affected by supply-chain constraints.
“I don't think there's a single hard tech company that is not. supply chain impacted.”
Listen at 45:44
- Max Kanon AI for scienceNeutral58:48
Most AI-for-science spending will target engineering rather than discovery.
“the bulk of the spend is on the engineering side of things.”
Listen at 58:48
- Max Kanon Semiconductor substrate yieldsPositive1:01:30
Raising semiconductor substrate yields to 95% would create substantial economic value.
“What if those yields become 95%? The economic value that's generated is massive.”
Listen at 1:01:30
AI could reduce NVIDIA’s chip-design-to-volume-production timeline to one year.
“What if it takes them a year?”
Listen at 1:01:57
- Max Kanon AI modelsPositive1:08:58
Customers will pay for AI models that are superhuman in valuable domains.
“People are only going to pay for models that are superhuman in certain domains”
Listen at 1:08:58
Personal-agent serving costs could decline by more than 100x.
“The cost to serve will actually undergo this potential 100x plus.”
Listen at 1:09:45
Personal AI agents will likely become economical to serve.
“I think it's very likely that they are totally economical to serve.”
Listen at 1:09:58
Elon Musk is turning SpaceX into a cloud-computing provider.
“He's taking SpaceX and turning it into a neocloud”
Listen at 1:10:44
- Max Kanon Frontier and non-frontier AI modelsMixed1:16:02
Non-frontier models will process most tokens, while frontier models capture most spending.
“the bulk of tokens will be non-frontier models. We just think the bulk of dollars will be frontier models.”
Listen at 1:16:02
SpaceX’s Starfall launch indicates a large opportunity in microgravity services.
“SpaceX just wouldn't do this unless it was like they saw it as like a massive opportunity”
Listen at 1:21:59
SpaceX’s Starfall validates Varda and the broader microgravity market opportunity.
“this just feels like a huge vote of confidence in the business and the overall market opportunity.”
Listen at 1:22:16
Some election results may remain unresolved throughout the week.
“some of the elections could be like, it takes time to count the votes, so maybe ongoing all week.”
Listen at 1:24:51
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.