✍️ Essay
✓ Link checked
Free
Beginner
Why we picked it
The definitive essay on where good ideas come from: notice problems you personally have, don't force it. Use it as the lens for judging whether your idea is a real problem or a solution in search of one.
From
paulgraham.com
by Paul Graham
~20 min read
- Live in the future and build what's missing.
- The best ideas look like bad ideas at first (schleps and hard-to-explain).
- Start with problems you have, in a domain you actually know.
Open
paulgraham.com →
✍️ Essay
✓ Link checked
Free
Advanced
Why we picked it
This is the direct counter to the fear in your short answer, that a platform update erases your product overnight. Chen argues the classic moats (distribution, network effects, workflow lock-in) come back exactly because the model is a commodity. Read it to understand what you must build around the model so you are not just a thin layer.
From
Andrew Chen
by Andrew Chen
20 min read
- Models commoditize, so the moat has to sit elsewhere
- Distribution and network effects still decide winners
- A thick wrapper with real switching costs is a real company
Open
andrewchen.substack.com →
✍️ Essay
✓ Link checked
Free
Advanced
Why we picked it
Before you commit, you need to know whether the value from your AI idea accrues to you or to an incumbent who bolts the feature on. Gil lays out when startups capture the upside and when the platform or the model provider does. It is a sober check against building something a giant absorbs in one release.
From
Elad Gil
by Elad Gil
12 min read
- Ask whether the value lands with you or an incumbent
- New customer segments beat head-on fights with giants
- You usually need to be ten times better to win the same buyer
Open
blog.eladgil.com →
✍️ Essay
✓ Link checked
Free
Intermediate
Why we picked it
Elad Gil is one of the most trusted operators-turned-investors on this exact question, and this piece cuts through the wrapper panic honestly. His core point is that most startups (AI or not) start non-defensible, and durable positioning is built after launch through data, integrations, and relentless execution, not claimed on day one. It is a grounding read for a founder worried their idea is too easy to copy.
From
Elad Gil (Elad Blog)
by Elad Gil
~12 min read
- Serving a real customer need well usually matters more than having a moat on launch day, and defensibility tends to accrue over time.
- Building on top of a model like GPT is fine, but the less you keep building and expanding after launch, the faster you get commoditized.
- Durable positioning comes from proprietary data, deep integrations, and execution velocity, so pick an idea where using the product compounds an advantage.
Open
blog.eladgil.com →
✍️ Essay
✓ Link checked
Free
Intermediate
Why we picked it
A map of which AI markets already have clear winners and which are still wide open, which is the practical version of asking where a newly viable idea still has room. It helps you avoid pouring effort into a fight that is effectively over. Read it to pick a lane where being early still matters.
From
Elad Gil
by Elad Gil
12 min read
- Some AI markets are already decided, others are wide open
- Pick spaces where timing still gives you an edge
- Winners in one layer do not lock up the layers above and below
Open
blog.eladgil.com →
📄 Article
✓ Link checked
Free
Intermediate
Why we picked it
This explains why niches once dismissed as too small (laundromats, vets, clinics) become real businesses when AI absorbs the labor-heavy steps and drops acquisition cost. That is the economics shift at the center of your question, shown concretely. Use it to see how a specific painful job in a small market turns into an opportunity.
From
Andreessen Horowitz
by Angela Strange and James da Costa
15 min read
- AI reopens niches that were too small to serve before
- Automating rote steps lifts revenue and cuts acquisition cost
- The product feels like leverage, not another dashboard
Open
a16z.com →
📄 Article
✓ Link checked
Free
Intermediate
Why we picked it
The core move in our short answer is decomposing a workflow into steps a model can now do. This piece frames every workflow as a vertical AI company waiting to be built, and shows where agents can take over multi-step jobs. Read it to practice breaking a real job into the steps you would hand to software.
From
Andreessen Horowitz
by Eric Zhou, Yoko Li, Seema Amble, Jennifer Li
15 min read
- Every multi-step workflow is a candidate to be automated
- Map the steps a human does, then ask which a model can own
- Agents that collaborate across parties build real switching costs
Open
a16z.com →
📄 Article
✓ Link checked
Free
Intermediate
Why we picked it
The opportunity is often a capability that already works in a raw model but nobody has wrapped into a specific job yet. This piece is about exactly that gap between what AI can do and what has been turned into a product. It helps you spot where good-enough automation exists but the painful job around it is still done by hand.
From
Andreessen Horowitz
by Angela Strange
12 min read
- The capability often exists before the product does
- Value comes from productizing a job, not exposing a model
- Look for tasks people still do manually despite AI being able to help
Open
a16z.com →
📄 Article
✓ Link checked
Free
Intermediate
Why we picked it
A structured playbook for picking a vertical where AI now delivers real leverage and building something that feels less like software and more like doing the work. It is practical about where to start and how to earn depth over time. Use it to turn a hunch about a newly viable workflow into a plan.
From
Bessemer Venture Partners
by Bessemer Venture Partners
20 min read
- Choose a vertical where AI replaces expensive manual work
- Aim for a product that does the job, not one that assists it
- Depth in one industry is the early moat
Open
bvp.com →
📄 Article
✓ Link checked
Free
Advanced
Why we picked it
Once you have spotted the opening, this gives you concrete principles for making it defensible, heavy on proprietary data and workflow depth, the exact anchor our short answer recommends. It is a checklist you can hold your idea against. Read it to pressure-test whether your idea survives a platform update.
From
Bessemer Venture Partners
by Bessemer Venture Partners
15 min read
- Own proprietary data a competitor cannot scrape or buy
- Embed in the daily workflow so ripping you out hurts
- Solve the whole job, not one visible slice of it
Open
bvp.com →
📖 Book
✓ Link checked
Paid
Beginner
Why we picked it
The single best thing ever written on customer conversations. It teaches you to ask about the customer's life and past behaviour, not your idea, so you can't be lied to. If a founder reads one thing before talking to a single customer, it's this.
From
momtestbook.com
by Rob Fitzpatrick
~130 pages
- Talk about their life, not your idea.
- Ask about specifics in the past, not opinions about the future.
- 'That's so cool, I'd totally buy it' is a compliment, not data, dig for commitment and evidence.
Open
momtestbook.com →
✍️ Essay
✓ Link checked
Freemium
Intermediate
Why we picked it
A clear-eyed take on what actually changes when a model can do a task at good-enough quality and cheaply, and where humans still hold the last mile. That is precisely the eighty-percent framing in our short answer. Read it to think honestly about which steps automate cleanly and which twenty percent you must design around.
From
Every
by Dan Shipper
15 min read
- Good-enough automation changes the economics of a job
- The hard, high-judgment slice is where humans still win
- Design the handoff between the model and the person deliberately
Open
every.to →
✍️ Essay
✓ Link checked
Free
Advanced
Why we picked it
To spot what is newly possible you need to know what the models can suddenly do that they could not before, and this essay maps the jump to reasoning and agentic capability. It points at the killer apps and new interfaces that jump unlocks. Read it to update your sense of the frontier so you are hunting for ideas that need the latest capability, not last year's.
From
Sequoia Capital
by Sonya Huang and Pat Grady
18 min read
- Reasoning models open jobs that pure chat could not touch
- New capability creates a short window of open opportunity
- The interface and workflow around the model is where apps win
Open
sequoiacap.com →
📄 Article
✓ Link checked
Free
Intermediate
Why we picked it
A broad map of where AI is actually gaining traction across infrastructure, horizontal, and vertical use cases, so you can see which workflows have crossed from experiment to viable. It is dense with benchmarks and patterns rather than hype. Use it to locate the spaces where good-enough automation already changes the economics.
From
Bessemer Venture Partners
by Bessemer Venture Partners
30 min read
- See which AI use cases have crossed into real adoption
- Vertical, service-heavy workflows are the biggest opening
- Ground your idea in where traction is real, not where noise is loud
Open
bvp.com →
