The Price Of Knowledge Is Going Towards Zero

Did you see how core wholesale prices rose like crazy?

What did?

Core wholesale prices.

No, I didn't.

So core PPI, which excludes volatile food and energy prices, increased seasonally adjusted 0.8%, more than the 0.6% gain in December and well ahead of the Dow Jones consensus estimate of 0.3%.

So are we cutting interest rates?

The 10-year went down. So usually if the 10-year goes down, that means they believe that rates are going to start going down, but they're worried about stagflation at Yeah, I mean, the world's crazy right now.

But honestly, like, okay, we can say that, right? But honestly, like, we should just say, okay, the world's crazy right now. Like, here's what we would do about it. So here's what at least I would do about it. I only think there's one way through it. Like, if you're curious and you're actually out there, you're curious, you're learning, you're experimenting, and you're actually even— a bonus would be if you're out there teaching, you will be unstoppable no matter what. I think all the stuff that people are talking about, oh, there's AI, there's layoffs right now. All you can worry about is yourself. And the only way out is through. So I'm looking over a hill right now. There's a hill over there, okay? So there's this giant tidal wave coming over that hill right now, and you can choose to surf that wave, or you can choose to like not do anything and just stare at it and just get hit by it, right? Um, but I would say the more curious you are, the more entrepreneurial you are— and you don't have to be an entrepreneur, you can be like an intrapreneur— um, you're gonna do just fine and you're gonna surf this wave. And that's why during this period of chaos right now, as the world adapts, I think there's gonna be some job displacement, but I think net-net Long-term, we're going to be okay.

Yeah. And when you look at like the overall wholesale prices, which is just another version of inflation, for anyone who's not sure what I mean by core wholesale prices, think of a business buying a product that they are going to resell to someone else. So you're buying, you know, I'm Best Buy, which is electronics company if you don't live in the US.

Wow.

And they go buy a TV from Sony and then they go and sell it to end consumer. Right? So they get a wholesale price. They don't pay the sticker price for a Sony TV. If they did, there would be no margin, no profit for them. And on an annual basis, core wholesale prices accelerated 3.6%. Now, is that some of that caused from tariffs and other stuff? I don't know. But the market didn't like it.

We should talk about this because I actually have a tweet that's related to this. So we're talking about Inflation, we're talking about stack. Actually, no, we're talking about stagflation, which is actually you have no growth plus inflation. Really bad. Okay, stagflation is terrible. Now, I we're talking about okay, you know blocks doing these layoffs over here. I just pulled this up. I saw this this morning as I was walking back from the gym. So this guy Jeff previously at Co2 and Goldman Sachs, so finance guy. I think we actually work with this company. This is a crypto company. Yeah, we did. We we worked with this company before. Say Network.

Okay.

Um, so spoke with this guy Jeff, says spoke with several YC founders planning to lay off all engineers below staff/principal, basically everyone under L5. This only became viable after Opus 4.5 in December. The block layoffs are a signal the floor just collapsed. If you're early in your career, the next few years are everything. Your edge will be how well you integrate AI into the value you create. The fastest learners are about to compound at absurd rates. Maybe that's part of the reason why, Neil, I, I'm excited for You, me, our other kind of friends who love to learn. But like, point is, you're talking about inflation going up and you're talking about these layoffs over here. So the headline I had for this one is the price of knowledge work is going to continue to trend towards zero because a lot of this stuff can do the work. But the people who are able to kind of converge their roles into multiple roles, they will make a lot more money and they'll be in even more demand. So if you, again, if you keep compounding yourself, you keep learning, you're, you're constantly curious, you're experimenting, you're eating the poo poo, you will be fine. Yep. Quick break. I want to tell you about something that we've been building for years at Single Grain. We've been doing SEO and content for clients, and the biggest problem is always execution. Everyone knows what to do, but doing it at scale is brutal. So we built ClickFlow. It's an AI SEO product that creates content that actually sounds like your brand, not just generic AI garbage. It also helps with other areas such as internal linking, creating FAQs, and more. We use it ourselves. Our clients use it. One user reported saving over 90 hours a month alone. If AI SEO matters to your business, you can try it for free for 14 days at clickflow.com. All right, back to it. All right, so we have this interesting one over here, Neil, on the new distribution channel is agents. So Google, I think recently just launched their Web MCP. So Neil, when you look at the ClickFlow website or you look at the Carrot website, Ubersuggest is a good example, or AnswerThePublic is a good example because you have— I think for both of those, correct me if I'm wrong here, but you have versions where they can just use it, right? Like freemium version, they can get a couple uses, correct?

Uh-huh.

Okay. So, or we can talk about booking flights on Expedia, for example, right? So if you're going to be running these agents, whether you're using like— I know Perplexity just came out with a computer use one. You can use Claude Cowork to do work for you and you can browse the internet. You want these agents, you want to make it very easy for them to ideally not do web browsing, but call on MCPs and just book a flight using whatever travel website you're using, or call on data from Ubersuggest or AnswerThePublic or ClickFlow or Carrot, right? Or spin up ads very quickly. So Web MCP was launched by Google, okay? And I think, what are we talking about here? Google, I think Stripe also launched something, I think called, like, to allow agents to interact with each other. So I think you're going to have a lot more with, I think you even You even had a post on how like a lot of traffic coming to your website now is like what, 50% bots? Is that what it is?

51% of website traffic in general are bots now.

Yeah. So, you know, this guy Akush is just saying that Andrej Karpathy said that the new channel, the news distribution channel for software is agents. Agents don't just browse your marketing site. They watch your demo video. They don't watch your demo video. They don't click through to your onboarding flow. They call your CLI. So that's your command line interface.

Okay.

That's like using a terminal. They hit your MCP server, they read your docs programmatically. If none of those surface areas exist, your product is invisible to them. Look at how fast this moves. MCP went from 0 to 97 million monthly SDK, so software development kit, uh, downloads in 12 months. 10,000+ active servers. OpenAI, Google DeepMind, Microsoft, Cloudflare all adopted it. By December 2025, Anthropic donated MCP— oh, I didn't even know that— to the Linux Foundation because the standard had already won. Running an MCP server is now compared to running a web server. That's the new baseline for product discovery. What do you think about this?

Yeah, I agree.

You're just too tired right now.

I am tired, dude. I'm like not daydreaming, but I kind of am.

I can tell you don't have it anymore. It's okay. I'll carry us through the rest of this, these 30 minutes. So, okay. I'll talk about it more. This is good. Neil can't—

I've been up for 27 hours at this point.

Did you say you just landed, right?

No, I landed a while ago, but I'm working both time zones, so it's just—

That's on you. That's on you. Yeah. See, by the way, we should, I'll take a little quick detour here. So Neil, did, did you sign up for the peptide company I was telling you about? Literally, it's called The Peptide Company from our mutual friend Jacob.

Mm-hmm.

Yeah. So did he give you the sleep peptides? Pinolon, epiphalin? Are you trying them right now?

I'm traveling. Did you bring some with you? No.

Oh, that's unfortunate. Okay. So anyway, I'll give this to everyone. We can take a little detour here and then come back to talking about distribution channels. This is not health advice, by the way, but I was taking Pinolon and that's— and I'm not even sure if I'm pronouncing it correctly, but that is a peptide. And basically, it gives me super deep sleep and it leveled up my HRV, which is my stress level. So, you want this to be higher, so to 107, which is pretty good. And it brought my heart rate down from 45 resting heart rate down to 40, which is really good. So, all that to say, with people that are tired like Neil who does traveling, there's a lot of travel hacking that you can do, right? And one of my other travel hacks too is to also not fall asleep until you're in the time zone and then fall asleep in the right time. So, that way you don't have jet lag. But Neil never listened to me for that one. He just kind of works both time zones. So, that's why he's dead right now.

And you have an app that helps with it. I don't know what it's called.

It's called TimeShifter. So, it tells you the right times to sleep. and all that. Sometimes it's pretty brutal. It tells you to fall asleep immediately when it's like in the middle of the day. I'm like, how do you do that, right? Um, sometimes you need special stuff for that, but, um, I'll leave that part alone. Okay, so coming back to this distribution channel piece, okay, so 85% of enterprises are expected to have AI agents deployed. These agents need structured programmatic access to your product, okay? They need CLIs, MCP endpoints, and machine-readable documentation. A beautiful React dashboard is worthless to an agent trying to pull data into Workflowy at 3 AM. So the way I look at this right now is Yes, think about MCPs. Yes, think about building for agents. These, um, yes, also if you want to distribute your, your repos from a marketing standpoint, put it on GitHub, right? That's what Corey Haynes did. He did a really good job with that. But at the same time, you still need a dashboard for humans. So the stuff we're building for Pay for Performance right now, Neil, we do have a dashboard for like the outbound emails we're sending, the outbound sales calls that we're doing. Okay, that's, that's one side of things. But even internally, We need a dashboard to manage all of our agents. You need some type of dashboard, some type of control system to see what they're doing and how they're performing, right? And I will tell you, these agents are just continuing to compound every day. And like, even by the way, Neil, even when I went to the gym today, I didn't realize I was running 10 different workflows. So while I was at the gym and when I was sitting next to my CTO yesterday when we were flying back, by the way, we're sitting row 26 economy, okay? It was great. So He, like, as he was talking to me, he spun up a couple different things, but by the time he landed, he's like, dude, I just pushed something to GitHub that would've taken 2 days to do, and I got it done in the middle of our conversation. So I think one, AI efficiency is crazy. That's one. The second one is build dashboards for your agents. The third one is, yes, you want to be looking at building discoverability. You want to make discoverability easy for agents because you're going to have a world where agents are doing a lot of the work.

You do. But the key here is, because dude, I was just in a heated debate about agents, like, uh, uh, 24 hours. It was, uh, it's a big company. I was at their offices. I can't mention the company, but they own some of the biggest cars that we all see in Los Angeles, uh, and in all places all over the world. So, and, and they own more than just cars. They're a really massive company.

Yep.

So when you think about it, we were talking about agents and deploying at mass scale, and they were just like, oh, we want to do agent for this, we want to do agent for that. And they were showing me some of the stuff that they're building, and their biggest problem wasn't the ideas, it wasn't actually even building the agents. They didn't have checks in place for how the agents were doing their work. I thought it was a smart idea to have a team continually go and build all these agents and get a lot of work done. And this group would help other departments internally create agents so then that way they could be more efficient. So think of it like as a centralized hub that just helps a lot of different teams in an organization build agents. And I was like, the biggest problem here isn't the building part. If you're not actually checking the quality of the work, it could be costing you a lot of money. I'm like, you guys need to figure out how to put that within your processes and systems.

Dude, that's okay, Neil. I did this like 2 days ago. Okay, so I've been building my— my OpenClaw has been sitting in my Telegram, so it's really only been working with me as almost like my chief of staff, you know, assistant, if you want to call it that. It has a lot of context around the business, the goals. It's hooked in with a bunch of MCPs.

Okay.

I added it into Slack 2 days ago. I added it into some of the private channels that we have, and people are working with it back and forth now. They're asking questions, they're querying data, they're asking for strategies, they're asking it to build things, right? I'm like, holy crap, I've just like opened up this machine. And so to Neil's point, I do think you need a human in the loop to kind of fact-check and check the data as well. You need a human in the loop to continue to fine-tune it. The problem is I was building all these things out of my own entertainment, but now it's time for me to kind of release some of my babies into the wild and let my team work with it, right? I'm still like, I think there's one thing to have a personal assistant, but when you have the personal assistant that knows what you want to hit and your goals and knows how you work and your team is querying data, like for example, Oh, hey, we created all these programmatic SEO pages around ChatGPT ads a few weeks ago. How's that going right now? Give me a 4-week table. Boom. Like literally in a couple seconds, it spun out and everyone can see how things are going. Oh, what should we do about that? It seems like, what are next steps here? Well, you know, it's not performing over here and you're not driving any conversions here. You should do this. And I'm like, oh, hey, SEO person over here, what do you think about this idea? And it's just happening in real time. Right. Um, and it's amazing. And then I added to the sales channel too. Hey, how have we been doing with, with, uh, look at our Gong and look at our HubSpot. How have we been doing in the last 60 days or so? What trends are you seeing? Right. And then if it starts to repeat something, report something wrong, the, the, it's okay because the salespeople jump in. But the key thing, Neil, is that now they're starting to engage with it more and they're having conversations with it and it's moving things along.

That is great.

Yeah.

So on a side note, I just got this email. It's one of the best marketing emails I've gotten in a long time. Check this out.

Look, I'm so boring, guys, that Neil's starting to check his email in the middle of the conversation.

No, no, no, no, this just popped up. So I have an OCD where I always have to have zero inbox. I don't know why. So this is to my dev email. You know, I get forwarded all my dev stuff when it comes to all the tech emails so that way I can see what my team is doing without asking them what they're doing. And we use Claude, of course, but look at this subject line, reduce your Claude API costs. They're trying to send me an email that helps me save money, which would hurt them, but in the long run, it would make me more loyal to them and cause me to spend more money over time.

You would use it more too.

I would use it more. Uh, so build efficiency with Claude. You're building with Claude. Now let's optimize your implementation. Here are 4 ways to reduce costs without sacrificing quality. Learn more. And then they just break down a lot of different examples. I was like, That's a great marketing email. It's very rare that a company does what's beneficial for you and hurts them in the short run, and this hurts them. This would make them less money, but in the long run, it would make them more money and have a much loyal customer.

You know what it is too, Neil? The subject line is really good. When it says reduce your API costs, any entrepreneur that's looking at it, any finance person that's looking at it, it's like, hell yeah, right? Dude, my API cost for Anthropic this month already, it's like $3,500. I'm like, oh damn. So, so I need to do something about that to optimize it. But you know what's going to happen in the future, Neil? You might get that email as a human and say, hey, and then there's going to be a link saying, hey, do you want me to just do it for you? And you just click it and it does it for you, right? Or the future is going to be, it's just going to be sent to an agent and the agent's gonna be like, hey, I spotted this over here. There's a way to save costs by 80% or so. Would you like us to do that? And actually my agent does that now. It's like, hey, you're spending on all these stupid costs over here. You're right. This model is too expensive. You should use this model up here. I'm like, that's good, right?

So yeah. Yeah, no, it's great as long as it does it accurately and good. And I think that's a problem that we face right now with AI in which it can make recommendations, it can help you save time, it can help you save money, but in that process it makes a lot of mistakes from what we see. But I bet if you compare Jan 1st, 2026 to Jan 1st, 2027, the amount of errors will drastically get reduced.

Yeah. And by the way, the cost per token is going to drastically reduce as well. You know what's kind of crazy to me? It's Anthropic or OpenAI, they're kind of raising for the next round, like the next training run, right? They're spending all the money because here's the thing, they know the price per token is going to drop, but they know they also have to have the best model, like a very competitive model. So they're just raising for the next round, next round, next round. Like, I don't know when that's going to stop, but it's exciting to watch from the side. And I actually, Neil, I I've been messaging Apple. I've been trying to do the math on whether it makes sense to buy the Mac Studio Ultras, like the new ones that are coming out, I think next week. And so you can buy a fully loaded one for basically 10 grand. And because I want to run these models locally, right, for a lot of the things, because it's going to be a lot cheaper, right, if you do it that way, especially the work that I'm talking about doing now. So I'm like, okay, well, if I buy 2 of these Mac Studios running KIMI, which is a Chinese open source model locally, I'm going to save a ton of money. Like, I'm going to be making money like pretty quickly, right? But if I were to buy 5 of these Mac Studio Ultras, that probably would cost me 50 grand or something like that, like out the gate. And so, you know, do I want to buy it? I'm like, well, the chips update every like year or 2 years or so. Like the chips in 2 years are going to be even crazier, right? And so I'm like, okay, well, instead of paying the 50 grand upfront, I can pay basically 8 grand to lease 5 of these. I'm just giving you a simple example, 8 grand to lease 5 of these Mac Studio Ultras for 24 months, so 2 years. Okay, are you following?

Yeah, I didn't know they have leasing programs.

That's a way better deal.

You're not gonna keep this stuff for 5 years.

It's gonna be way better tech in a year, right? And so like, um, I was working with my Open Claw on like coming up with the math where it's like, yeah, you should just lease these things. Because my— one of our mutual friends yesterday is like, why don't you just buy it? I'm like, because this stuff becomes outdated so quickly and I want the latest technology running my stuff, especially if it's for our own stuff. And also if it's helping Clients, because clients are going to want stuff local, isolated to one computer.

So yeah, I think leasing's smart unless it slows you down a lot, and then it's just a cost analysis. Because to do leasing, typically you got to fill out financial paperwork and all this kind of stuff.

It's really easy. It's really easy.

Oh, okay. If it's really easy and it doesn't slow you down, but like if something was going to slow you down a month or two, it's just like, yeah, I'm not saying spend the $50 grand, but maybe you go and you buy one of them for $10 grand and you lease the rest.

That's exactly what I'm going to do. So I'm going to buy one for 10 grand and then the other ones I'm probably just going to lease. It just depends on how this scales. Like if this one, this initial pilot test goes well, like we're going to be off into the races. Guys, thank you for joining us and we will see you. We'll see you next time. Bye.