Every spring, I get the urge to clean house - and every spring, I reach for the gardening gloves instead. Creating new things, whether with code or in the dirt, is where I find my energy. But I did end up cleaning out my tabs this month, which led to some long-overdue writing. Because there's one question I get asked more than any other: what exactly is an AI answer engine?
My colleague Jimmy and I finally wrote the answer down. It's the guide I wish had existed when we started building. Here's what else is on my mind this month.
This Month's Take
Two years ago, if you asked me what Dewey was, I'd have fumbled through something about "AI-powered search" or "a personal librarian for your content." Both descriptions were technically accurate and missed the point.
Here's what I should have said: your audience has been trained – by ChatGPT, by Google, by every search experience they've touched in the last two years – to ask questions and expect direct answers. When they can't get that experience from you, they get it from someone else. Most likely, someone who scraped your work, stripped your name off it, and served it back as generic advice.
That's the problem a custom AI answer engine solves. A user asks a question in plain language. The system searches your content (and only your content), finds the relevant material, and returns a cited, accurate answer. If the answer isn't in your archive, it says so.
That last part is the whole game. The difference between a custom answer engine and a generic chatbot is the difference between a trusted advisor and a confident stranger.
Three things I believe about this space:
1. Your archive is your most undervalued asset. If you've published hundreds of articles, episodes, or newsletters, most of your audience will never find them. Not because they don't want to, but because site search is broken and browsing doesn't scale.
2. Closed-loop architecture isn't a feature. It's the foundation. There are tools that sound similar: custom GPTs, RAG chatbots, AI search widgets. The critical question is whether the system can reach outside your content. If it can, even a little, you've given up the one thing that makes this valuable: the guarantee that every answer carries your research, your judgment, and your standards.
3. The human control layer is what separates a product from a toy. Upload your PDFs to a custom GPT, and you've got a weekend experiment. Build a system where editors can review answers, pre-approve responses to sensitive questions, and route high-stakes topics to subject matter experts - now you have something worthy of your brand.
If you're evaluating AI tools, read the full guide. We go deep on the technical architecture, including content ingestion, the query pipeline, anti-hallucination verification, and more.
From Our Work
Anthropic published new labor market research this month. The headline is reassuring: workers most exposed to AI aren't showing higher unemployment than unexposed workers. But buried in the data: hiring of younger workers (ages 22 to 25) into AI-exposed occupations is slowing down fast.
The traditional logic: you hire junior people because that’s how you build the senior people you’ll need in five years. Most companies aren’t thinking about that horizon right now. They want someone ready to manage an army of AI agents ASAP. I think that’s a short-sighted decision. You can’t skip the pipeline indefinitely - especially when the supply of senior engineers who can also orchestrate AI work is already at a premium.
There’s also an underrated edge to the 22-year-olds themselves. They’ve only known this world. They’re not unlearning ten years of habits - they’re AI native. That’s an edge being quietly discounted in the rush for near-term gains. Read more if you’re thinking about your hiring strategy for 2026.
What We’re Building
The longevity space has a problem. For every peer-reviewed study on sleep or metabolic health, there are a dozen influencers hawking supplements with the energy of a get-rich-quick scheme. That’s what drew me to Livelong Media - evidence-based longevity journalism that treats readers like adults, reaching over 300,000 health-conscious readers monthly.
What we built together:
🔍 Liv — An AI guide to Livelong’s health and wellness archive. Ask about sleep, metabolic health, or longevity research, and Liv surfaces relevant articles with clear citations.
🤖 Unapologetically AI — We're allergic to building "clones" that pretend to be experts. Liv doesn't pretend to be human - they "live in the servers" and know it. This vibe isn't just fun. It sets honest expectations about what AI can and can't do safely.
🚫 No fake empathy — Liv won’t claim they’ve “been there too” or try to be your friend. They know where their boundaries are: firmly along the lines of Livelong’s coverage, not dubious personal medical advice.
🛡️ Trust-first guardrails — Every response is grounded in Livelong’s vetted content. No hallucinated health claims, no pulling from the internet average.
Health information is exactly where the stakes of AI trust are highest. Getting it wrong isn’t just annoying — it can be dangerous. Give Liv a try, and if you’re building in health, wellness, or any space where trust isn’t optional, let’s talk.
A Bit of Joy
Do you remember when the internet was full of weird humans, doing weird human things? We got a bit of that back this month with “Your AI Slop Bores Me”. It looks like an AI chatbot, except when you type something, a random human answers you or draws you a picture.
Tens of millions of people logged in within a week to give it a try.
Which tracks. Because it is so darn human. There's something almost embarrassing about how delightful it is to get a random, wrong, deeply human answer from something that looks like a machine. We spent years optimizing the internet for speed and accuracy and relevance – and somewhere in there, the weird got sanded off.
The footer says: "humans make mistakes because that's what makes us human."
Amen. I hope you enjoy it as much as we did.
One Question for You
As the internet gets more filled with AI, where do humans become more valuable?
I’m thinking about this question a ton and asking everyone I talk to right now. I’d love to hear your thoughts - hit reply and it will come straight back to me.
And if you are thinking about the role of answer engines, search, and AEO for your business, grab time on my calendar. I’d love to brainstorm with you.
See you in 2026,
Alex
You're receiving this because you signed up with Dewey Labs - where we build expert-first AI tools. We work with publishers, brands, researchers, and non-profits to create search and Q&A experiences grounded in trust and accuracy. We’ll be sending a monthly dispatch on the future of the expert economy like this one, along with occasional product updates.