The synthesis trap
David Peterson

I just had a first. Over the weekend, I picked up Abundance, the latest from Ezra Klein and Derek Thompson about their new vision for a "liberalism that builds." But after a few chapters, I put it down, pulled out ChatGPT, and had a 20 minute conversation about the ideas instead. Never picked up the book again.

Mind you, this is a book I wanted to read! I'm genuinely a fan of both authors. I'd probably more readily describe myself as an "abundance" liberal than anything else. But did I want to spend 10+ hours reading it? Not really. Because it quickly became clear that I could get 90% of the value from a brief conversation with AI.

As somebody who both loves reading and feels a certain pressure to produce content as part of my job (thanks for reading, by the way!), this whole episode caused me some consternation. So I asked myself: what was it about Abundance that made it so susceptible to being replaced by an AI-mediated conversation?

My answer: because Abundance fell into the synthesis trap.

Abundance is full of interesting ideas, but Klein and Thompson have been exploring them in their writing and podcasts for the past two years, which means they've already been captured by AI and are easily synthesized/discussed in conversations with any of the models.

In the age of AI, this sort of content - the synthesis of ideas that are already out there in the ether (and thus have already been hoovered up by the models) - really serves little purpose anymore.

Unfortunately, “synthesis” describes a lot of the content currently being produced. All but the best, most deeply researched non-fiction (including most, if not all, textbooks). Those business books that wrap a Tweet-sized insight inside 150 pages of fluff and anecdote. Every SEO-optimized blog post. Basically all VC content (sorry about that).

For those of us who write on the internet hoping to generate interest in what we do, whether as founders or investors, this begs the question: what content should we focus on creating? What's worth our time to produce (and other humans' time to consume)? Where do we have an edge over AI?

With that in mind, I think I've identified three categories where humans still have a slight advantage:

First, the genuinely entertaining. I still haven't read something written by an AI that made me laugh out loud or kept me genuinely at the edge of my seat. Call it taste, call it art, call it whatever you want. Most humans are not genuinely entertaining, but the best of us far exceed AI’s capabilities.

Second, the definitive take. Think academic research. Think Robert Caro, Ron Chernow, or Erik Larson. These authors don't just synthesize existing knowledge; they unearth new information through painstaking research, often in physical archives or in-person interviews. They create works that are not merely summaries but genuine contributions to our understanding.

Third, the real-time opinion. This is where platforms like Twitter and Substack shine. This is why TBPN is taking off! Novel takes and real-time opinions on matters of the day are always in high demand. AI can summarize yesterday's news, but it can't react in real-time to events as they unfold with the nuance and insider perspective that human experts can.

That’s basically it. If what you’re writing doesn’t fit into the above categories, I’m not sure it survives when pitted against AI (I write this fully aware that I really need to up my newsletter game to compete!).

We’re still in the early innings of the AI-mediated internet, but I can already feel myself unlearning how a decade plus of the internet (specifically social media) incentivized us all to engage online. So, before you hit publish, ask: is this AI-proof? Or just more fodder for the algorithm? Is this simply synthesizing what's already out there, or is it contributing something genuinely new, entertaining, or timely? Your audience's increasingly scarce attention depends on the answer.

No More Painting by Numbers
It’s the end of the “SaaS playbook.

The Age of Artisanal Software May Finally be Over
Every wave of technological innovation has been catalyzed by the cost of something expensive trending to zero. Now that’s happening to software.

Founders as Experiment Designers
David on why founders should run everything as an experiment.

When Growth Stalls
Or why to kickstart growth you should narrow your ICP.

ENTERPRISE/TECH NEWS

OpenAI’s new coding agent. A few days ago, OpenAI released Codex, “a cloud-based software engineering agent that can work on many tasks in parallel, powered by codex-1”. Dan Shipper, CEO of Every, shared a terrific review of the new coding agent. “Codex is built to turn you from programmer to manager. It felt like the next evolution of what I experienced when I first reviewed Devin—like playing online poker in college, where you can be running 3-4 tables simultaneously. And, somewhat surprisingly, it’s pretty good. It confidently one-shotted a styling fix for one of our internal apps, called Paradigm. It also one-shotted the new feature for Cora I mentioned earlier. There is a piece of the UI that you can expand and collapse. I asked it to figure out how to save that state between user sessions. and it did it quickly and well. Codex isn’t a vibe coding tool. You can tell it’s built not to replace senior software engineers, but as a tool for them.”

A new database. In his latest post, Benn Stancil makes the case for a new kind of database, one built for documents, semantic search, and AI. “When people talk about how AI could change data (and other concerns), they tend to talk about SQL-writing copilots and agentic analytics platforms. The future is agents, finding the insights in our database that we could not. I’d bet on the inverse. The potential energy in analytics is in a better data source, not in better analysis. And it’s not the analyst that needs to be different, but the database.”

Perplexity’s new round. AI startup Perplexity is reportedly raising $500M in a round led by Accel at a valuation of $14B - a 50% increase in valuation since their previous funding round announced last November. Perplexity offers an AI-native search experience with direct answers instead of traditional link lists, and is planning to launch its own browser, Comet. While Perplexity has rapidly grown its revenue and users in the past year, the cost of that growth has been steep. “Perplexity generated $34 million in revenue last year. But it burned about $65 million in cash as it spent heavily on cloud servers and AI models from Anthropic and OpenAI, which power many of the search engine’s answers, according to financial documents.”

HOW TO STARTUP

Sequoia’s AI playbook. VC Guillermo Flor wrote a breakdown of Sequoia’s recent AI Ascent event. The entire post is worth reading, but this section in particular is noteworthy: “Where the real value will be captured (hint: it’s not the models). Yes, the models are powerful. But the value? It’s still in the apps. Sequoia believes most of the value will sit at the application layer—not in the infrastructure, but in the software people actually use. That’s how it played out with cloud and mobile, and the same pattern is taking shape in AI. But it’s competitive. Foundation models aren’t staying in their lane—they’re moving into application territory too. That means startups need to build from the customer back. Focus on a specific user, in a specific market, with a real workflow. Go deep, not wide. That’s where startups can still win.”

Immigrant entrepreneurs. While restricting immigration has been a key priority for the Trump administration, new research from Stanford shows that 44% of US unicorn startups were founded by immigrants. Notably, “Israeli immigrants create unicorns at highest rate per capita” and “the relocation effect is powerful. Israeli startups that moved to the U.S. were 9x more likely to achieve unicorn status than those that stayed home.”

HOW TO VENTURE

Betting on European defense tech. Former CIA officer Eric Slesinger is going against the grain with a new $22M fund focused exclusively on European defense tech. Why did he choose to focus on Europe when defense tech investing is predominantly US-focused? “First, “Europe has individual entrepreneurs that are just as hungry, just as high conviction, and just as smart as anywhere else in the world.” Second, “European governments waited way too long to rethink what the arrangement on their own security meant, and therefore hadn’t actually taken a critical eye towards it.” And third, “Europe was quickly being seen and will, in my opinion, continue to be the site of serious gray zone competition,” meaning activities by state or non-state actors that fall between traditional peace and outright war.”

Aquant provides AI-powered tools that inform maintenance decisions, can help prevent breakdowns in factories, and save manufacturers big bucks.

Portchain reached 150 terminals on its global berth alignment network.

LightSolver’s CTO Chene Tradonsky broke down the technical aspects of how Laser Processing Units work and how organizations can utilize this new computing paradigm to solve complex mathematical problems.

DUST Identity
Senior Backend Engineer (Boston)

Groundcover
Senior Talent Acquisition Partner (Tel Aviv)

Paradime
Product Engineer - Backend - Python (Remote)

Tensorleap
Head of Customer Success (Tel Aviv)