The alpha, twice a week
That was 6 min.
Get the best new moments in your inbox every Wednesday and Saturday. It's free.
Check your inbox
We sent a link to your address. Click it to finish signing up. Not there in a minute? Look in spam.
or
MTS · 6 min · Oct 16, 2026
Agents automate code but still face business decisions
Han Wang
Co-founder, Mintlify

Theo JaffeeHost · Co-founder & host, MTS
One of the first members of a16z's new media team; earlier hosted the Emergent Ventures-backed Theo Jaffee Podcast.

Sophia DewHost · Host, MTS
About this segment
Han distinguishes routine implementation choices from decisions with business impact, like selecting a database. Human preferences and spending approvals still shape which providers agents use.
From “AI Agents Are Replacing Human Readers | Han Wang”, Oct 16, 2026. Watch the full episode
Thanks. We'll look at and fix what's off.
“It will at least confront the user with these high-level decisions because it still has business impact and outcomes to it.”
Takeaways
- Coding agents automate many low-level implementation choices.
- Agents still ask people about choices with business consequences.
- Human preferences influence what agents recommend.
In this episode
Full episode- 3:02Agent traffic does not make human-friendly docs obsolete3:41
7:02Agents automate code but still face business decisions6:13
- 13:15Good docs need structure, even as agents take over writing5:38Up next
Transcript
Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.
- 7:15
I think in those cases, it's really important to distinguish between the idea of a high-level decision and then a low-level implementation decision. I think what we're seeing across the board right now as coding agents have evolved is that the individual decision on how to implement an endpoint, how to write the code, a lot of these small nuances that you have to kind of, you know, is what riddled the difficulties of software engineering is now completely automated.
- 7:40
But what we have yet to see is still the high-level decisions of, hey, I want to go build this product. Agent goes ahead and does it. It's still going to have very core-level decisions of like, hey, if you're going to pick a database, is it going to be Supabase? Is it going to be ClickHouse? Is it going to be AWS? What does that look like? It will at least confront the user with these high-level decisions because it still has business impact and outcomes to it. And so... Oh, it still will ask the end user? We believe so, yes.
- 8:10
Interesting. Yes. I think the world in which the agent just fully goes in and does everything from scratch is not the behavior that we're seeing from agents. What we see from the, what we hear from the Anthropoc and Cloud team is they still, when it comes to kind of like things that involve financial decisions, business outcomes, these kind of large commitments, it still requires kind of some human input. And they try to ask humans for guidance on that. But increasingly more so that is being abstracted away. Pretty fascinating. I guess, you know, if you were to advise someone like ClickHouse or MongoDB around their advertising, you'd say it's still important that they tell humans about their brand, people understand it, so then they can prompt the agent. It's not the agent just goes rogue and makes these decisions for them completely
- 8:52
from start to finish 100 okay 100 and one of the reasons why is because if you think go back and just talking a little bit more about the geo stuff yeah right like what what leads an agent to pick one product or another it still scrapes the web or forums reddit blog posts on what people use and what people like and that information is still instilled from human opinions yeah agents aren't like living in this like reddit forum themselves discussing it they still adhere to what human preferences are. And so winning the hearts and minds of the end person is still extremely important, right? And so how do you do that from a fundamental level? How do you make sure you still get the person to know about the brand, to know how to use the product, to still pick you so that agents can go and implement it? I think it's kind of a challenge that we're, you know, seeing
- 9:40
companies solve. But on the other end, what do you think is irrelevant? Like what types of companies or tools are irrelevant for humans to know about? Because it's just, I just want to make sure that agents can index it and I don't even care if a human hears who I am. Yeah. That's a really good question. I think that I don't think that there will be the case one day where like for instance I'm just gonna tell Claude to build me an app and the next thing I know it's completely built and I just get billed from unknown vendor names and unknown brands. Really? You don't think that's going to happen i think it will i feel like that's already happening no i think it's gonna happen but there will always be some brand or some like kind of like some provider that may have abstracted away some of those like vendor relationships so a good example is like lovable maybe i'm like hey
- 10:34
lovable build me an app and it itself chooses the super bases or the databases that it prefers um and i will maybe get a lovable bill i don't know if i will want to wake up one day and just be like hey you know suddenly you know I get an invoice from this random company and that I have no idea what's for and kind of go from there. My experience is if I ask CloudCo to just build the app I don't connect it to my GitHub I don't connect it to anything I say like build the app do everything on your own come back to me when you have a live link of what's working I always get a Vercel link it's always deployed on Vercel. Always deployed on Vercel I think that's first of that's awesome. Yeah, yeah. Great for Vercel. Good advertisement for them. Exactly. Yeah. I mean, obviously Vercel is awesome. What I've noticed and what we've seen in the industry is oftentimes,
- 11:22
like a good example of this, we think of it as like, you know, like a lovable super based relationship. Oftentimes, the agent is given kind of instructions or kind of harnesses to kind of build towards specific vendors. And that oftentimes applied to kind of like the vendor relationship is abstracted from the company. But I don't know, what do you think? Do you think at one point it's going to be predominantly autonomous?
- 11:46
Well, that's what I thought going into this conversation. I thought humans are not going to be involved at all. You'll be involved a little bit of a high level taste, but when it comes to knowing the difference between Superbase or MongoDB or different types of databases, a human never- has to make that call. Yeah. But, I mean, you are the expert at it. I mean, you're with Mintlify. You're directly seeing how developers are calling this information.
- 12:08
And so, I didn't mean to go down a big rabbit hole around this, but I'm fascinated that you actually think this won't totally be the case. Yes. At least not from what we're seeing, but it might change. Yeah. And maybe you're right in the reality that, like, hey, maybe, like, at the end of the day, you know, people might not really care about the vendors or the products or the specific limitations, and more of the lower level decisions are going to be abstracted away.
- 12:31
I think that from what we've seen in our perspective as a business is that even if Claude, like, for instance, picks a Vercel as a hosting service and provider, I at one point still need to opt in to paying for Vercel. Fair, fair, fair. Even if I get the Vercel link, I'm like, oh, great. Like, let's scale it up. Let's scale it as a team. Let's bring it into the business. For now, until you just give your agent a wallet and say, spend it and go whatever whatever it needs yeah maybe maybe that makes sense as well interesting sorry to go all the way down this rabbit hole I was just it was just different from what I