Transcript: What matters online when AI makes software easy to build
An AI-generated audio overview (Google Gemini), transcribed by machine on Blaine's Mac (Whisper). The audio was checked against primary sources; what it gets wrong is listed at https://blainomd.com/learn#audio-when-ai-makes-software-cheap

Welcome to today's Deep Dive.
We're jumping straight into this fascinating
2026 industry report on the collapse of the build layer
and what actually happens to the web
when AI makes software production virtually free for you.
I'm your host, and okay, let's unpack this.
It is like giving everyone a magic wand
to conjure apps out of thin air, right?
But what happens to the traditional magicians?
Yeah, I am thrilled to be here,
and it really is a massive shift.
We are fully in the era of vibe coding now,
so you aren't typing syntax anymore.
Platforms like Lovable, I mean,
they just hit a $6.6 billion valuation
because non-engineers like marketers or nurses
are launching over 100,000 functional apps every single day.
Oh, wow, 100,000 a day, that's wild.
Exactly.
The barrier to building software isn't just lowered,
it is completely gone.
So if anyone can generate an app in a few minutes,
that infinite supply just creates infinite noise
for you as a user.
I mean, how do your AI agents even know what's real anymore?
Well, the web is reorganizing into these durable verticals,
and the first one is context,
which relies on proprietary data gravity
because a general AI is smart,
but it's generic until it absorbs your specific data.
I mean, look at Notion.
Users have already built 21,000 custom agents there
that just pull from their internal workspaces.
The data's gravity pulls the capabilities
into something actually useful for you,
and the second vertical is trust.
Okay, trust.
How does that work in this space?
We are seeing protocols like the Agentic Commerce Protocol
or ACP, Standardizing Machine Readable Checkouts.
So an agent can authenticate a merchant and transfer funds
without ever seeing a visual webpage.
Wait, wait.
Here is where it gets really interesting to me, though.
If autonomous agents are buying things for you,
how do we trust them not to fall for AI-generated scams
or buying 500 laptops by mistake?
What's fascinating here is the sudden rise of AI insurance.
Companies like Eleven Labs are underwriting agents
using an AIUC1 certification.
Okay, but how do you actually insure an algorithm?
That sounds like a nightmare.
It is essentially a digital crash test.
Before an agent can spend your money,
the system bombards it with over 5,000 adversarial simulations.
5,000? That is intense.
Yeah, it feeds the AI contradictory instructions,
malicious prompts, fake products.
It is actively trying to trick it.
And if the agent's logic holds up under that pressure,
it gets certified.
But insurance only covers the financial risk, right?
Yeah.
It doesn't solve how an agent discovers a product in the first place.
If I tell my AI to find the best headphones,
where does it even look?
Right, and you're hitting on the distribution crisis.
Amazon, for instance, is aggressively blocking
external AI agents right now, and it's pure economics.
Amazon relies on this massive advertising engine,
which pulled in $17.7 billion in a single quarterback in late 2025.
Ah, I get it.
Because an AI agent doesn't care about a sponsored banner ad,
it bypasses the ads to just look at raw specs and price.
Exactly.
If humans start browsing, Amazon's ad revenue vanishes,
so it becomes a walled garden.
Meanwhile, the open web has to adapt by moving to agentic SEO.
Agentic SEO, so optimizing for machines instead of humans.
Precisely.
Brands are optimizing API accessibility.
The AI just wants the raw structured data of your inventory and return policy,
not your beautiful graphics.
You know, this really reminds me of the garage band analogy from the report.
When cheap tools flooded music, the market was full of noise.
But the artists who won out were the ones with actual taste,
like human editorial judgment.
So what does this all mean for the human in the loop?
Well, humans are now basically in the business of selling liability.
A 2026 Deloitte study actually shows that 86% of people still want humans making important decisions.
Yeah, that makes total sense.
And 61% completely reject AI-led insurance approvals.
We don't just want an efficient answer.
We want someone to hold accountable if it goes wrong.
So for you listening, as coding becomes basically free,
you really have to transition from being an engineer of code to an engineer of context, trust, and taste.
If we connect this to the bigger picture, the agentic web doesn't replace humans at all.
It strips away the mechanical production and forces human accountability to the absolute forefront.
Absolutely.
Which leaves us with a pretty wild thought to end on.
Since agents now handle the discovery phase,
if your personal AI does almost all your browsing,
how long until advertisers stop trying to target you
and start trying to psychologically manipulate your AI?
We'll leave you to mull that one over.