contents
There’s a website called Linktree. It lets you put more than one link in your Instagram bio. That’s the entire product, worth $1.3 billion.
Grammarly checks your spelling. Valued at a measly $13 billion with over $700 million in annual revenue.
Iubenda is a tool that generates privacy policies for websites. Just a privacy policy. $24 million in revenue, 160 employees, acquired by team.blue in 2022.
The “soundwave on top of album art” format you might have seen a thousand times on X/Twitter? That’s a tool called Wavve. A solo founder built it, ran it for a few years, and sold it for an indie millionaire exit while it was throwing off $100K a month in profit.
Meanwhile, you and I are spending our time thinking about whether to fine-tune a model, build an agentic workflow, or move to SF for the AI summer.
There’s a name for this: schlep blindness.
Paul Graham wrote this 14 years ago
The term comes from a Paul Graham essay published in January 2012. “Schlep” is Yiddish for “a tedious, unpleasant task.” PG’s claim is painful:
Your unconscious won’t even let you see ideas that involve painful schleps. That’s schlep blindness.
The mechanism is not deliberate, and you’re not consciously rejecting hard ideas. Your brain is filtering them out before you even consider them, the same way it filters out the hum of an air conditioner you stopped noticing an hour ago.
PG’s main example was Stripe. Online payments were a $20 trillion problem in 2010. The technical solution wasn’t novel. The reason nobody had built it was that it required dealing with banks, fraud, regulations, chargebacks, and the headaches of integrating with COBOL-era banking infrastructure. So nobody did. Until two brothers from Limerick decided the schlep was worth it. Now Stripe is worth more than $50 billion.
The essay is 14 years old and has been read by every founder who’s ever taken Y Combinator’s Office Hours seriously, even today.
A $29 billion company built on this one essay
In 2016, Alexandr Wang dropped out of MIT to start Scale AI with a simple pitch. AI models need huge volumes of high-quality labeled data, and labeling data is mind-numbing manual work that no AI researcher wants to do. Scale AI would do it.
Wang has been explicit about where the idea came from:
One of the secrets to Scale AI — and I think this applies to almost every industry — was that the problem we were solving of building really high quality data sets was something that most machine learning teams knew was important but wasn’t necessarily the sexiest problem that every AI scientist wanted to work on.
The company is now valued at $29 billion, powering training data for OpenAI, Meta, and the U.S. military.
PG wrote about it in 2012. Wang acted on it for AI in 2016. By 2026, schlep blindness should be a solved problem. Founders should have read the essay, internalized the lesson, and now be elbows-deep in unsexy work.
But we aren’t. The 2025–2026 cohort of founder energy is overwhelmingly chasing AI moonshots: personalized medicine, electronics in space, inference chips for AI workflows, AGI bets. All those are worthwhile efforts. Meanwhile, every existing schlep just got cheaper to solve, because the tedious parts (data entry, boilerplate, manual review) are exactly what AI is good at.
What 2026 schlep blindness looks like
The frontier of schlep blindness in 2026 is old schleps that AI just made tractable.
Backlinks for SEO are a schlep. One well-known tactic involves finding broken links across the web, identifying their original content, recreating that content on a client’s blog, and writing personalized outreach to dozens of webmasters per backlink. Done manually, it’s $2-5k a month for a handful of links. AI can do most of the steps now, but nobody’s built the AI-native version yet.
Privacy policy compliance is a schlep. Updating one when laws change across 30 jurisdictions is a schlep on top of a schlep. AI can handle most of it. There’s room for an Iubenda 2.0 that’s 5x cheaper and also automatically updates for you when you add a new tool, because AI does the work.
Cold email is a schlep. The new 2026 tactic is reversing the script: do the hard part first and build a personalized deliverable for the prospect (a draft blog, a custom report), then send a single email pointing to it. You still need lots of volume, but conversions are higher for high-quality deliverables that are already made. The schlep is producing the deliverables at scale. AI can eat that schlep for breakfast before you drink your oat latte.
The cocktail-party version of “AI is changing everything” usually points at the moonshots. But what the average founder should be excited about is that thousands of boring problems just became economically viable to solve.
How to spot your own schlep blindness
If schlep blindness is unconscious, you can’t introspect your way out of it. But you can use other people’s annoyance as a proxy. Some questions:
- What’s the most-skipped task in your workflow? Look beyond the things you’d describe as “I hate this” if asked. What do you do regularly that you avoid thinking about at all.
- What problem do you assume someone else is solving, but can’t name a product for? Obvious problem, no obvious product. Schlep city.
- What’s been broken for 10+ years that nobody seems to fix? If something’s been broken that long, you know there’s something there.
- What’s gated by “this is annoying” versus “this is technically hard”? The annoyance bucket is hugely under-supplied.
- Which of your daily complaints could a 50-person company solve, if they were willing to do the unglamorous work? That’s a schlep waiting for the right person.
The honest test, though, is the gut one. Look at the products on this list (Linktree, Wavve, Iubenda, Grammarly, Scale AI, Stripe) and ask: would you have built any of these? In your top-of-mind list of startup ideas right now, are any of them close in shape to “more than one link in my Instagram bio”?
If the answer is no, it’s probably because your unconscious is filtering them out.
The bet
To be fair, several opportunities in 2026 really are downstream of big AI breakthroughs and moonshots. But many more are in the schleps you’ve been walking past for years, with new economics because AI took the worst part of the work off the table.
PG saw the pattern in 2012, Wang acted on it in 2016, and many more since. The list of companies in this post is what happens when people take it seriously. Plenty of schleps are still out there.