VCs are Subsidizing Startups
Is this a good thing?
A solo founder can now build in a weekend what took a team six months in 2019. Claude Code, Codex, Lovable and now Cowork/OpenClaw. The cost of building has collapsed.
The cost of finding someone who wants what you built hasn’t moved.
That gap is the whole story. And VCs are papering over it with cheap money, which means founders never have to confront it.
When building was expensive, you had to be careful about what you built. You talked to users first because shipping the wrong thing cost you six months and a co-founder. Now you can ship the wrong thing every weekend. The feedback loop that used to teach founders what a real problem looks like has been severed. There are advantages to this. You can bring something to market and get some real signal(like a credit card transaction, faster), but founders are avoiding this for just as long.
Most of what I see now is founders solving problems they’ve never had, for users they’ve never met, with a distribution plan that amounts to “post on LinkedIn and hope.”
There are two ways to actually get customers, and they’ve been the same for decades.
Direct outreach to specific people who have the problem
Content that compounds over time and builds trust with a targeted audience
Both require putting yourself in front of strangers and being told no. Both feel slower than building. Both are the actual job.
Technical founders avoid them because building feels like progress and selling feels like exposure. Every hour spent on a feature is an hour you didn’t have to be visible. The AI tools have made the hiding place infinitely deep. You can refactor, redesign, rebuild, regenerate, and feel productive for months without ever having a conversation with a person who might pay you.
The other pattern I see constantly is the breadth problem. Founders building “an AI assistant for teams” or “a platform that helps you launch a startup.” No specific user. No specific job. No specific alternative they’re better than. Just a vague gesture at a large market.
If you’ve never built a successful business, you can’t see how specific you need to be. The ICP isn’t “small business owners.” It’s “solo founders who’ve been technical for ten years, have built something, have fewer than ten real users, and avoid customer conversations.” The job isn’t “help them grow.” It’s “get them to send fifty cold messages this week without freezing.”
You can’t generalize your way to product market fit. You have to narrow until it almost feels embarrassing, then narrow more.
That’s not even the hard part. Actually delivering a solution that solves the problem, in whatever rough form you can manage at first, is what founders skip over.
The deeper issue is that AI is trained on the internet, and the internet is full of nonsense business advice. The software it generates is genuinely good because software has documented best practices. The business strategy it generates is mostly recycled generic processes like “make a business plan”, “define your problem statement”, “form your founding team”, because that’s what the training data looks like. Founders who lean on AI for code get a real edge. Founders who lean on AI for go-to-market get confidently wrong answers in a polished voice.
Solve problems you’ve actually had. Sell to people you actually understand. Use the AI to build faster, not to think for you.
Failure used to be expensive enough to teach you something. Cheap money and cheap tools have made it cheap enough to repeat indefinitely. The startups that survive this era won’t be the ones with the best AI stack.
They'll be the ones who used AI as a force multiplier on real customer understanding, not a substitute for it.