Deep Tide Guide: The author of this article, Steve Blank, is very famous in the Silicon Valley startup circle and is known as the "Father of Lean Startups." He wrote The Four Steps to the Epiphany and is the proposer of the Customer Development methodology.

Eric Ries's The Lean Startup was developed based on his theory. He has taught entrepreneurship courses at Stanford, Berkeley, and Columbia University, and the U.S. National Science Foundation's I-Corps program is also built on his methodology.

Steve Blank recently had coffee with a founder he had invested in and found that the founder had been working hard for six years without realizing that the outside world had changed.

He wrote this article, and the core point is straightforward:

If your company was founded more than two years ago, your business plan is likely outdated. AI is reshaping development speed, team size, pricing models, and competitive moats. Founders still running with a 2024 playbook are unlikely to make it to the next round of funding.

For readers who are currently starting businesses or are interested in the tech and venture capital scene, firsthand observations from across the ocean are worth reading.

Below is the full translation.

If your company was founded more than two years ago, many of the original assumptions likely no longer hold.

You need to stop what you're doing—whether it's coding, building products, hiring, or fundraising—and first look around to see what's happening. Otherwise, the company will die.

A Cup of Coffee-Induced Anxiety

I just had coffee with Chris. Chris is a founder I invested in six years ago, and he has been working hard ever since, focusing on:

1) A complex autonomous systems problem,

2) In an existing market,

3) With a unique business model.

Chris is now preparing to launch his first large-scale funding round. I reviewed his investor deck and found a problem: While he was heads-down working, the outside world has undergone earth-shaking changes.

The autonomous systems software moat he spent five years building is becoming less and less unique. Autonomous drones and ground vehicles in Ukraine have spawned dozens, if not hundreds, of companies with larger teams and more funding working on the same things.

While Chris has been pushing for customer adoption in his niche market (which indeed needs disruption but is still controlled by incumbents), demand for autonomous technology has exploded in an adjacent market: defense.

Over the past five years, VC investment in defense startups has surged from zero to $20 billion annually. His product is suited for logistics and medical evacuation in contested environments. But he was completely unaware of these opportunities in the defense market.

Chris's team did achieve impressive system integration (deep integration with an existing flight platform, making his solution different from most competitors), so there is still a business, but it's no longer the business originally envisioned.

After talking with Chris, I realized: Most startups founded more than two years ago have outdated business plans, and their tech stacks and team configurations are likely obsolete.

If you haven't been looking up recently, here's what you've missed.

What Has Changed

VC money is heavily tilting toward AI. In 2025, AI projects took two-thirds of total VC investment. This means if what you're doing isn't AI-related, you're competing for a smaller pool of capital. Non-AI startups must answer a question: Why can't a better-funded, AI-native competitor simply eat your market?

For software founders, AI has completely rewritten the old formulas for cost, speed, and manpower. Using tools like Claude Code or OpenAI Codex for "Vibe Coding," an MVP (Minimum Viable Product) can be built in days or even hours, not months. This also means the MVP itself no longer proves your team's capability.

These tools are changing the composition of development teams: Fewer engineers, and the types of engineers are changing, with a split between "business process engineers" and "deep tech engineers."

Work that used to require a development team can now be done by a few people, sometimes just one. Data, once a source of differentiation and a moat, is being commoditized by foundation models (ChatGPT, Gemini, Claude) for public data sources.

The very concept of agile development needs rethinking.

The old bottleneck was: Can we afford to build and release this product? The new bottleneck is: Do we know what to test? Can we reach users fast enough to learn? Agile is no longer a serial process. AI Agents can run multiple things in parallel at the same or lower cost.

You can now test multiple versions of the same business simultaneously, or even test different business directions at the same time. You can run five pricing models, ten marketing messages, twenty UX flows concurrently. And the "user interface" might no longer be a screen; the test goal might become: finding the prompt that gets the AI Agent to deliver the desired outcome.

The bottleneck is no longer engineering capability but has shifted upward to judgment, insight into customer desired outcomes, and distribution.

AI Agents Will Rewrite Every Software Category

AI Agents will change every software category, including yours.

Today's software applications work like this: Show information to the user and wait for the user to act through interfaces like dashboards, alerts, workflow tools, and reports. But customers buy software to get a job done, not to look at more screens. Actually getting the job done is what AI Agents (orchestrated by tools like OpenClaw) will achieve autonomously.

What does this mean?

If your product currently tells the user "what to do next," an AI Agent will eventually do that step for the user. If a competitor's product automatically completes the task while yours is still waiting for a mouse click, you are no longer competitive.

The next generation of applications won't just display information on screens; they will act like an employee: resolving tickets, booking meetings, qualifying leads, auto-replenishing stock. As products shift from "software as interface" to "software as outcome," pricing will shift from per-seat fees to pay-for-outcome fees: per ticket resolved, per meeting booked, per lead closed.

(The search for Product/Market Fit will become the search for AI Agent/Customer Outcome Fit. The Minimum Viable Product (MVP) will become the Minimum Possible Outcome (MPO). I'll expand on this in a future article.)

Hardware Isn't Spared Either

For hardware founders, the changes are equally dramatic. Hardware is still bound by the laws of physics, capital, supply chains, and manufacturing cycles. You can't skip machining metal, building prototypes, or chip tape-outs.

But AI lets you kill bad ideas faster. You can now simulate more design variants, create digital twins, and stress-test assumptions earlier and cheaper before building physical prototypes. The result is accelerated learning and discovery (and sometimes faster failure), and in a startup, failing faster is an advantage, not a drawback.

Once AI is embedded as part of the system, the product itself changes. Add an AI backend to a camera, and it becomes a surveillance system, a vibration sensor, a machine failure prediction system. Robots become factory workers. The moat is no longer just the hardware itself but what the hardware can sense plus what decisions and actions can be made with that data by the AI.

The Sunk Cost Trap

Companies founded before 2025 typically have tech stacks optimized for a world where software development was expensive and customized. Agile development and DevSecOps made us lean, but they operate serially, and team sizes were structured accordingly.

Companies that spent years building "proprietary code and feature moats" are discovering that AI is commoditizing much of their tech stack. This puts fundraising startups in an awkward position: their business model may be partially or wholly obsolete.

These changes aren't necessarily visible when you're heads-down building product and searching for Product/Market Fit.

Tech stack, product features, user interface, headcount—these sunk costs become reasons not to pivot: How can we throw away years of work? Our VCs invested based on this direction. Customers still want the UI. The team believes in this roadmap. Our customers aren't ready yet.

(Chris is a classic example. He built something truly impressive that likely still has competitiveness, but the business model around it needs to change.)

Some sunk costs are actually assets: deep domain knowledge, customer relationships, proprietary data, hard-won regulatory approvals, physical integrations. These are worth keeping. Chris's flight platform integration falls into this category.

The sunk costs that are truly liabilities are: large engineering teams built for slow software cycles, per-seat pricing models, product roadmaps built around features rather than outcomes. These are the "Dead Moose on the table"—the obvious problem no one wants to address.

The founders who survive are those who can look at what they've built and ask: If I were starting over today, in today's market, with today's tools, what would I actually build?

This is an uncomfortable question when you've already raised funding for a specific direction. But it's nothing compared to the discomfort of having your investors tell you they're not funding the next round and you shutting down with an obsolete plan.

Summary

· You can't run a 2026 race with a 2024 (or earlier) playbook. Funding, technology, and business models have all changed. Agile development is becoming parallel development.

· The search for Product/Market Fit will become the search for AI Agent/Customer Outcome Fit. MVP will become MPO (Minimum Possible Outcome).

· A sunk cost mindset will put you out of business.

· Defensible moats may still exist in: proprietary data, deep understanding of customer outcomes, regulatory lock-in, or becoming a Program of Record.

· If you're still sleeping soundly, you haven't figured out what's happening.

· The founders who survive will get out of the building, see the landscape, pivot, and correct course.