World models

There's a lot of excitement right now about world models that simulate physical reality. (Here’s a fantastic overview on the latest.) But I've been thinking about a different kind of world model: a continuously running, ever-updating simulation of how a small corner of the economy actually works.

These sort of “living economic models” exist all over the place. Every insurance company and lender. Every hedge fund. All the infrastructure powering every ad you’ve ever clicked on. These companies are built on models that are constantly updated based on observed outcomes and performance. The winners in these industries aren't the ones with the best brand (though it matters to some extent) or relationships (though they matter to some extent). They're the ones with the most accurate simulation.

So the idea is not new. What's new, I think, is who can do it, and for which domains.

Historically, these systems emerged in domains with repeated decisions, accessible data, observable outcomes, and enough economic upside to justify hiring serious technical talent. Finance had all four. So did digital ads. But most of the economy never got this treatment. And the vast majority of useful context in most industries isn’t so accessible. It lives in contracts or court filings. Regulatory documents or maintenance logs. Endless amounts of unstructured data, inaccessible to anybody without a deep bench of technical talent at their disposal, at least.

AI breaks that constraint. A pipeline that ingests PDFs, extracts structured data, and feeds a continuously running model is now buildable by a small team without deep technical specialization. The commercial real estate analyst who was reading documents and entering numbers into Excel can now build the system that does it automatically, and run a model that updates in real time rather than monthly.

This opens up entire industries that looked model-resistant because their data was stuck.

There are many markets like this: title insurance, mineral rights, environmental credits, litigation finance, government procurement, specialty insurance, permitting risk. They are messy, document-heavy, judgment-heavy domains where the right model could be an edge.

Here's a specific example a founder schooled me on last week: construction liens. When a subcontractor doesn't get paid, they file a mechanic's lien at the county recorder's office. These filings are public records, but they are scattered across thousands of counties. The “lien filing rate” against a general contractor is one of the best leading indicators of whether they pay their subs. So, a model that continuously ingests those filings across the whole market knows which GCs are under payment stress before their banks do. Construction lenders would love that information. That's one customer. But the more interesting business might be to buy those liens directly from the subs who need cash immediately for operations. Indeed, why sell the model at all? The firm with the best model knows which liens are worth buying and what to pay for them. Every deal adds to the dataset and the pricing gets more accurate over time.

The obvious business is to sell the signal. The better business is to use it.

That is when these companies stop looking like AI software companies and start looking like something stranger: AI-native financial institutions? AI-native specialty hedge funds? I’m not even sure. The point is that the model is not the product. The model is the underwriting engine. And AI is enabling companies to underwrite smaller and stranger parts of the economy.

That, to me, is the interesting angle here and why I keep coming back to it. It’s not just that AI improves the productivity of some analyst stuck reading poorly digitized PDFs. It’s that AI can turn previously illegible markets into businesses with tight, compounding feedback loops feeding their own, specific economic world models. 

If I were starting a company right now, I think I’d be trying to build a world model, too. And not the Fei-Fei Li kind.

David Peterson

Could the future of software be fluid
How do we get the best of AI without losing the soul of software?

The future belongs to young missionary teams
Why it makes more sense betting on youth in the current moment

The AI-native enterprise playbook
Ten real-time observations on a rapidly evolving playing field

No more painting by numbers
It’s the end of the “SaaS playbook.

ENTERPRISE/TECH NEWS

Ineffable. Ineffable Intelligence announced a massive $1.1 billion seed funding round at a $5.1B valuation, the largest seed round in Europe ever. “Ineffable Intelligence is using  Reinforcement Learning to create a “superlearner” that can endlessly discover knowledge and skills without relying on human data. The company was founded by David Silver, the former lead of the reinforcement learning team at DeepMind, and is a professor at University College London. Silver is committing to giving away 100% per cent of the money he makes from his Ineffable equity via Founders Pledge - the biggest pledge in their history, and it is likely to amount to multiple billions.”

Big round for Helsing. German defense AI startup Helsing secured $1.2 billion in funding, pushing its valuation to $18 billion. This significant investment highlights the growing market and investor confidence in AI applications within the defense sector. The round positions Helsing as a leader in European defense AI, although information on revenue and traction is spotty.

EU Postpones High-Risk AI Restrictions Until Dec 2027. EU legislators have postponed high-risk AI restrictions until December 2027 and exempted industrial AI applications from the AI Act. This marks the first significant delay in the landmark regulation, reportedly under pressure from the US. The decision impacts the regulatory landscape for AI in Europe. “The deal marks the first significant rollback of rules in the digital space, as the EU faces pressure from the U.S. over its tech laws and amid warnings from its own industry and governments that strict restrictions had put the bloc at a disadvantage in a global AI race. Welcoming the deal, Commission President Ursula von der Leyen said it “provides a simple, innovation-friendly environment” for AI in Europe. “At the same time, we are strengthening protections for our citizens. For safe and simple AI governance in Europe,” she said on X.”

A VAST valuation in Israel. Vast Data, which has built critical data infrastructure for major AI players, announced $4B in revenue and a $30B valuation, making it the most valuable private company in Israel and nearly reaching the valuation of Wiz. “Vast set out to redesign data infrastructure for an era in which AI workloads would place fundamentally different demands on computing systems. Its core architecture, known as Disaggregated Shared Everything (DASE), was developed to address longstanding trade-offs between scale, performance, simplicity and cost. Over time, the company expanded this foundation into what it describes as an “AI operating system,” combining data storage, compute and real-time processing into a single platform. The approach reflects a growing industry view that traditional layers of computing infrastructure are converging as AI systems become more complex and integrated. According to the company, its platform is now embedded in a wide range of large-scale AI environments, supporting systems that run on millions of GPUs. Customers cited include cloud infrastructure providers, enterprises and government organizations.”

HOW TO STARTUP

Enterprise sales AI bot. A blog post by CruxOCM CEO Vicki Knott on how the company built an internal AI sales strategist (nicknamed Piper) to drive tangible value in a hard industry. “The most exciting part? This isn’t about replacing our people. In an industry built on trust and complex operations, you can’t automate a relationship. Piper is about industrial AI solutions applied with intention — better thinking, scaled. It enables our team to ask better questions, identify real pain points faster, and align our solutions to the actual business outcomes our customers care about. We aren’t just using AI to keep up with the trend. We’re building an enterprise agentic AI platform — an AI-native operating model that mirrors the precision and safety our customers expect from CruxOCM. The future of sales using AI in oil and gas industry isn’t more noise — it’s more insight.”

An assault on the AI-as-jobs-killer narrative. David George of A16Z argued that the narrative is just flat-out wrong, calling it a complete fantasy. “The doomer argument goes, “If AI can do our thinking for us, then humanity’s ‘moat’ evaporates and our terminal value goes to zero.” Checkmate, humans. Apparently, we’ve done all the thinking we’re ever going to need or want, and now that AI will carry an increasingly large share of the cognitive load, humans slide into obsolescence. Here’s the thing, though: precedent (and intuition) shows that when the cost of a powerful input falls, the economy does not politely stand still. Costs fall, quality rises, speed rises, new products become viable, and demand moves outward.2 Jevons Paradox reigns supreme. When fossil fuels first made energy cheap and plentiful, we did more than just put whalers and woodchoppers out of business; we invented plastics! Contra-doomers, there’s every reason to expect that AI will have a similar effect. Now that AI will carry an increasingly large share of the cognitive load, humans are free to tackle even more ambitious frontiers than ever before.”  Scott Galloway also wrote a piece that suggests that narrative is likely wrong but may become to a self-fullfilling prophecy: “Recently, Anthropic CEO Dario Amodei warned that 50% of entry-level tech, legal, consulting, and finance jobs will be completely wiped out within five years. Last year, he told Axios the “white-collar bloodbath” could spike unemployment to 20%. In 2023, when the AI narrative felt more optimistic, Elon Musk said, “There will come a point where no job is needed … AI will be able to do everything.” In 2021, a year before launching ChatGPT, Sam Altman wrote, “The price of many kinds of labor will fall toward zero once sufficiently powerful AI joins the workforce.” Translation: AI is an extinction-level event for workers … according to those who benefit most from AI being an extinction-level event. Their story is as old as the Industrial Revolution. In Narrative Economics: How Stories Go Viral and Drive Major Economic Events, Nobel Prize-winning economist Robert Shiller argued that fears about machines replacing human labor contributed to 19th century economic downturns. Later, science fiction reinforced the narrative, feeding the incorrect belief that automation caused the Great Depression. Fears about the rise of computers exacerbated the double-dip recession of the early 1980s. The danger, according to Schiller, isn’t labor disruption, but the narrative’s negative feedback loop. “The economic hardships created by a temporary recession or depression are mistaken for the job-destroying effects of the machines, which creates pessimistic economic responses as self-fulfilling prophecies.”

HOW TO VENTURE

In defense of cold inbound. Dan Gray wrote a brilliant piece arguing that cold inbound is a critical source of alpha for VCs. (We agree!). “Put another way, generating alpha is implicitly difficult. The easier it is to make an investment decision, the less alpha it is likely to generate. Cold inbound has the most friction, takes the most effort, and requires the greatest conviction, thus it yields the strongest outlier returns when it works….Venture capital is an outlier business. The returns are produced by a small number of companies that, by definition, cannot be systematically identified. There is no checklist that reliably surfaces a higher-potential founder, and the more an investor leans on that approach, the more they are optimising for the mediocre middle of the distribution.”

Reco has released Reco AI agent security to fill the visibility gap for AI agents across the enterprise.

Firebolt
Senior Software Engineer (Munich)

Groundcover
Backend Engineer (Tel Aviv)

Tensorleap
Head of Operations (Tel Aviv)