Investing in the second-order effects of AI

Opportunities beyond performance, how to write with AI, and growth after outbound

---

Read to the end for a how-to guide for writing well with AI and an invite to a private dinner in New York.

I’m back after a nice two-week break. This is the longest pause I’ve taken in my newsletter in ≈16 months.

---

Second-Order AI - models get better, so what?

At both the app and model layers, everyone is basically chasing some version of “max token throughput.” That is, the single plotline: “How do I make AI better and get it to do more things?”

This is great and makes sense given the civilizational scale of the technology we’re already operating at. But it does wind up feeling a bit hollow. Sort of a brute force “more, more, more” approach to summoning The Machine God.

The more interesting question isn’t what AI can do. It’s what happens after it does it. The first-order effects of AI are mechanical and obvious: cheaper digital production, personalization, observation, coordination. So what? What happens next?

This is probably the single most exciting opportunity set in 2026 and beyond: investing in the second-order impacts of improved AI.

AI Breaks Things. Now Fix Them.

AI amplifies volume until existing filtering/routing mechanisms collapse and need to be rebuilt.

• Hiring: Applications are free to send. Volume explodes. Recruiters can’t distinguish candidates. Resume screening stops working. The whole funnel breaks.

  • Slow company: MeritFirst, Tofu, Stealth

• Outbound sales: Sophisticated campaigns available to everyone, flooding channels. The well gets poisoned and new GTMs replace it.

  • Slow company: Memelord, Stealth

• Code collaboration: Git was built for scarce, human-written code. Whatever replaces it will be built for abundance.

  • Slow company: Stealth

• Trust and security: Everything on the internet is fake. Content, identities, credentials. Systems that assumed human-scale production and human-verifiable authenticity stop working.

  • Slow company: Outtake, Sublime Security, Stealth

• Deployment: more apps, faster releases, more integrations, etc. SIs are critical chokepoints and will become very powerful this cycle.

  • Slow company: Stealth, Stealth

These problems are already real and the solutions are being built today or will be soon.

Obviously we’ve already invested in these ideas and are looking for more. The full list of what will get broken and needs to be rebuilt is much longer, maybe infinitely so. I’m curious where you’re looking and where else this dynamic is going to play out.

Net-New Opportunities

Then there’s whole categories that only make sense once AI capabilities are mature.

• The complement shift: Intelligence commoditizes. Its complements become the new scarcity: trust, taste, judgment, access, meaning. As production gets cheaper, people will pay for what can’t be produced.

• Org compression/the end of companies: Coordination costs drop and small teams can do big things, dropping minimum company sizes. We might even see a return of temporary, purpose built companies (like film productions or joint-stock companies).

These are further off and squishier based on proliferation and capabilities that we don’t yet QUITE have.

So what?

The social, economic, and political upheaval and change this civilizational technology wreaks will be epochal.

There’s a ton of money to be made improving AI right now (duh) but there’s a lot more creative ways to build businesses using AI in novel ways and fixing what it breaks. The most exciting opportunities and n-of-1 businesses will be in the second order impacts.

Growth after outbound

I’ve written before about how AI is going to kill outbound performance:

The scale of outbound sales will increase by orders of magnitude and the well will become permanently poisoned. Your bot’s emails and calls will just go unanswered. Your AI generated SEO blog posts will go unread. The only way to build a business in that future will be bottoms up demand generation brand, audience, captive distribution, virality, reputation, network.

As outbound becomes less performant it will be part of a channel mix/halo strategy not unlike display advertising before.

At the company level, we’re seeing founders respond in a bunch of ways: becoming media brands, showing up to more conferences, building creator-led companies, acquiring their own distribution outright. All reasonable.

But how do you bet at the theme- rather than company-level? There’s historical precedent. Masa built SoftBank by buying and building tech magazines and tech conferences, most famously paying Sheldon Adelson $700 million for COMDEX. If software was gonna be huge, he’d own the places where software people gather. [Aaron Witt is running a version of this today with BuildWit (vertical software for construction) and Dirt World (a construction conference).]

But obviously conferences and media brands aren’t great standalone venture investments in 2026. So what’s the contemporary equivalent?

If trusted networks and warm intros become everything when cold outreach dies, then the investable layer might be platforms that aggregate or route through those networks. Influencer marketplaces and audience platforms. Paid intro routing (Boardy and Superconnector are building versions of this).

Every company is going to have to respond to the deterioration of traditional outbound. There are obviously going to be platform-level bets to make on that theme.

How I use AI (and How I Don’t): Bicycles for the mind vs motorcycles for the mid

I’ve been an avid fan of LLMs as a writing copilot and editor since ChatGPT launched. My usage has massively accelerated over the last few months since I largely can’t type on a computer (wrist injury)

1. Speak notes aloud (via wispr bc no desktop Claude voice mode) and brain dump, often with screenshots and/or prior essays. This is very useful to just get the most raw versions of an idea together without self conscious constraints.

2. Claude asks me clarifying/challenging questions (prompt: “interview me asking questions one at a time to help flesh this out”) to organize/substantiate the idea, after which I give an outline/flow

3. Claude generates a first skeleton draft putting my notes into the flow

4. I move to a word processor (google docs or bear) and re-write paragraph by paragraph (also via wispr).

5. I send back to Claude for comments and feedback before repeating step #4.

6. Once I’m satisfied, I send it to Pangram to make sure I’ve sufficiently re-written everything and hold myself accountable.

7. Publish

My goal with AI as a writing partner (and remember, writing is thinking) is to do more and better, rarely just to do it “faster”. A bicycle helps you go further faster but still requires a great deal of effort and provides conditioning. A motorcycle is much faster but relying on it will atrophy your muscles and lungs.

I am very worried about how many smart people are just writing longer because is now cheap. They’re using intellectual ICEs instead of burning calories; reading and writing that makes us all dumber and less informed.

If you’re not using AI to write/think, you’re not maxing out what you can do. If you’re only using AI to write, you’re part of the problem.

AI is the new plastic; it can enable novel processes and beautiful designs or cheap, disposable crap. It’s up to you.

Cyber Dinner

​AI is breaking $1.5T of security market cap.

​For 30 years, cybersecurity has worked because attacks have had known limitations. Businesses have also had a concept of control: how to secure within a perimeter, how to vet vendors, and how to remediate gaps or accept known risks. The $1.5T security market cap was built on the assumption of visibility and control. That known landscape is gone with AI.

This winter I’m teaming up with ex/ante and Outtake to host a dinner and discussion in NY about the changing threat landscape and how new and existing security companies will meet that.

Sign up here!