Panic, Progress, and the Playbook
Issue #10 | Margin of Signal: August 17, 2025
Your weekly briefing for cutting through market noise to find what actually matters.
This week, we explore:
A historical lens on the present
Bret Taylor on agents, pricing, and how to build software
Dylan Patel on infra reality: moats, neoclouds, and the open-source squeeze
A couple of concrete trade ideas
Chart of the week
Incredible things are happening in the Land of the Rising Sun:
While semiconductors, ASICs, and AI are all the rage, consumers seem to prefer the Japanese sneaker company.
1910: When panic met progress (Derek Thompson)
Derek Thompson revisits the early 1900s to show how waves of anxiety and moral panics coexisted with bursts of invention.
Motorcars and airplanes collapsed distance. Culture lurched forward. Nerves frayed.
Sound familiar?
History doesn’t repeat itself, but it often rhymes. Per Philipp Blom, from The Vertigo Years, excerpted by Thompson:
Technology had created a new race of giants… and it changed the experience of space and time itself.
Public angst doesn’t foreclose breakthroughs; it often rides alongside them.
That’s a useful frame for AI’s current discourse: loud fear, real change.
Systems Thinking in the Age of Code Machines (Bret Taylor)
Bret Taylor is one of the rare builders who have done zero-to-one and scaled to enterprise maturity, which gives him a practical understanding of how AI will be productized and purchased.
Some career highlights: he co-created Google Maps (built it over a weekend), helped popularize the Like button at FriendFeed, served as Facebook’s CTO, founded Quip (acquired by Salesforce), became Salesforce’s co-CEO, and currently serves as Chairman of the Board at OpenAI and founder/CEO at Sierra (an AI agent company transforming customer service).
In this conversation, you’ll learn:
The brutal product review that nearly ended his Google career—and how that failure led to creating Google Maps
The question Sheryl Sandberg taught him to ask every morning (“What’s the most impactful thing I can do today?”) that transformed how he approached every role
The three AI market segments that matter
Why AI agents will replace SaaS products
His framework for knowing whose advice to actually listen to—and how that came in handy during the OpenAI board drama
The counterintuitive go-to-market strategy most AI startups get wrong
Sierra’s outcome-based pricing model that’s transforming how enterprise software is sold (and why every SaaS company should adopt it)
What he’s teaching his kids about AI that every parent should know
Taylor argues the unit of software is shifting from app to agent, and the unit of value from usage to outcomes. That repositioning changes careers, pricing, and GTM.
This interview is must-see TV for builders and investors in the SaaS and AI space.
AI Capex is an Economic Engine (Dylan Patel x No Priors)
In this interview, Dylan Patel (Founder, CEO, and Chief Analyst for SemiAnalysis, the preeminent authority on all things AI and semiconductors) disassembles the stack: silicon, networking, software, and the messy reality of building capacity at speed.
His thesis is counterintuitive to the doomer take: the massive data center build-out is, in effect, economic stimulus.
Dylan and Sarah Guo (Founder and Managing Partner at VC firm Conviction and formerly a General Partner at Greylock Partners) dive into topical questions around the current state of AI infrastructure.
Together, they explore why Dylan loves Android products, predictions around OpenAI’s open source model, and what the landscape of neoclouds looks like. They also discuss Dylan’s thoughts on bottlenecks for expanding AI infrastructure and exporting American AI technologies.
A few takeaways:
Open source LLMs keep raising the floor, especially for code and tool use, which pressures thin wrappers and forces inference providers to differentiate with real software and orchestration
Neoclouds that pair ruthless execution with strong software and networking will keep winning work; those that just rack GPUs drift toward commodity returns
Meanwhile, billions of dollars are spilling into power infrastructure, substations, optics, and trades labor spend—an economic tailwind with real bottlenecks
Intel: the ultimate “golden age of grift” trade
The video’s core claim: if the U.S. government takes an equity stake in Intel, it won’t be charity, it’ll be statecraft. Framed as a national-security project, the government could pair capital with policy (tax breaks, procurement guarantees, tariff leverage) to pull Intel’s domestic fabs across the finish line and compel order flow from big buyers.
In that world, Intel’s valuation story stops competing on promise alone and starts competing on mandate: fewer token “AI PC” headlines, more legally backed throughput, margins, and share capture at home.
The kicker—and where the ultimate golden age of grift angle comes in—is the gap between narrative and scoreboard.
In this era, narrative enjoys a long lead: subsidies and press confer momentum well before yields, nodes, and networking do. If Washington weaponizes industrial policy, the trade is less a classic turnaround and more a policy-sponsored rerating.
Or, to put it satirically: if hope were a process node, Intel would already be shipping 1A.
Disclosure: I own Intel shares ( INTC 0.00%↑ ) and plan to buy more. Not investment advice.
Trade Idea #2: Long vibe coding
If agents are the new app, low-friction creation becomes the on-ramp. My piece makes the case for a platform that already owns SMB demand, distribution, and payments—and now has a vibe coding wedge that could compound.
The AI Vibe-Coding Stock Hiding in Plain Sight
Disclosure: This post is for informational purposes only and is not investment advice. I may hold positions in the securities discussed.
That’s a wrap for this week. We’ll be back in your inbox next Sunday with another Margin of Signal.
Thanks for reading,
Chima