Nate Herk | AI Automation · YouTube · 07:43

Anthropic Just Dethroned OpenAI. Here's What Happens Next.

A 7-minute essay arguing that the AI coding war is a free sample phase — and your $200/month is a 12–24 month exemption from real prices.

Posted

May 13th 2026

2 days ago

Duration

07:43

Format

Talking Head

educational

Channel

NH|

Nate Herk | AI Automation

§ 01 · The Hook

The bait, then the rug-pull.

Two tweets, one hour apart. One from Sam Altman offering a free month of Codex, one from Anthropic announcing 50% more Claude Code usage through July 13. Nate Herk saw both land on the same morning Anthropic officially passed OpenAI in business adoption — and instead of reading it as a win, he read it as a warning: you are not the customer, you are the training data.

§ · Stated Promise

What the video promised.

stated at 00:34“That's what I wanna spend some time talking about with you guys today.”delivered at 07:00

§ · Chapters

Where the time goes.

00:00 – 00:37

01 · The Trigger

Ramp AI Index article drops: Anthropic at 34.4%, OpenAI at 32.3%. Sam Altman tweets free Codex, Anthropic counters with 50% more Claude Code through July 13.

00:37 – 02:18

02 · The Free Sample Frame

$200/month gets you output comparable to a $5–15k/mo engineer. Companies are deliberately eating cost because adoption and proprietary usage data matter more than subscription revenue right now.

02:18 – 03:20

03 · The Data Moat

Usage patterns are the biggest moat. Altman admitted OpenAI was losing money on Pro subs. Real API cost is 5–25x what you pay on subscription.

03:20 – 04:17

04 · What the Numbers Actually Mean

Ramp adoption metric is real but narrow. Three headwinds for Anthropic: misaligned incentives (profit from expensive models), Claude regression reports, compute and cost problems.

04:17 – 05:03

05 · The Historical Pattern

Every subsidy era runs four phases: land grab, habit forms, competition thins, the reset. FB Ads, AdWords, Uber, DoorDash, Netflix, AWS all ran this playbook.

05:03 – 06:07

06 · Three Reads

Bullish: use it like crazy. Bearish: don't build on sand, your data is the product. His view: both — burn credits, invest in portable patterns not vendor lock-in.

06:07 – 07:43

07 · Closing Frame

Don't panic at the Twitter war. Build tool-portable projects. The $200 is a 12–24 month exemption from real prices.

§ · Storyboard

Visual structure at a glance.

Ramp chart open

hookRamp chart open00:00

Sam Altman tweet

hookSam Altman tweet00:14

Claude Code 50% tweet

hookClaude Code 50% tweet00:37

Free Sample Phase slide

promiseFree Sample Phase slide01:00

You are the training data

valueYou are the training data01:42

Ramp chart detail

valueRamp chart detail03:45

Four phases slide

valueFour phases slide04:18

Three reads + closing frame

ctaThree reads + closing frame05:23

§ · Frameworks

Named ideas worth stealing.

01:00concept

The Free Sample Phase

AI companies are subsidizing usage because they need adoption and usage data more than revenue right now. The pricing you see today is not the pricing you will see in 12–24 months.

Steal forany video about SaaS pricing, AI tools, or the own-your-stack argument

04:18model

The Four-Phase Subsidy Cycle

Land grab
Habit forms
Competition thins
The reset

Every platform that subsidized early adoption eventually normalized to real market prices once habit lock-in was achieved. FB Ads, AdWords, Uber, DoorDash, Netflix, AWS all ran this cycle.

Steal forexplainer video about why SaaS prices always go up, why you should own your stack

05:03list

Three Reads (Bullish / Bearish / Where I Land)

Bullish: Use it like crazy — cheapest AI will ever be vs what it can do. OpenAI is paying you to switch. Subsidy users get a multi-year head start.
Bearish: Don't build on sand — reset has already started, image tokens 3x'd, every loss-leader phase ends, your data is the product.
Where I Land: Both at once — burn the credits, run both, invest in patterns not vendors, treat this as a 12–24 month exemption.

A clean three-column decision matrix for how to respond to the AI pricing war.

Steal forany commentary video that needs a bullish/bearish/synthesis structure

§ · Quotables

Lines you could clip.

01:42

“You are not the customer. You are the training data.”

Self-contained, punchy reframe. No setup needed.→ TikTok hook

03:12

“You would be paying five to 25 x more money on those tokens compared to what you're actually being charged on your subscription.”

Concrete stat that reframes subscription price as a deliberate subsidy.→ IG reel cold open

06:07

“If one day Anthropic just disappeared... if you could move everything over to a different tool in the matter of like an hour and the rest of your week is completely unchanged, that's a win.”

Clear, quotable, actionable benchmark for tool independence.→ newsletter pull-quote

07:00

“You're not paying $200 for AI. You're paying $200 for a twelve to twenty four month exception from real prices.”

The entire video thesis in one sentence. Standalone.→ TikTok hook

§ · Pacing

How they spent the runtime.

Hook length14s

Info densityhigh

Filler5%

§ · Resources Mentioned

Things they pointed at.

00:05toolRamp AI Index ↗

00:14linkSam Altman Codex tweet

00:37linkClaudeDevs 50% limits tweet

§ · CTA Breakdown

How they asked for the click.

07:30subscribe

“If you guys enjoyed and learned something new, please give it a like.”

Soft single ask at the very end. No mid-roll CTA, no product pitch.

§ · The Script

Word for word.

HOOKopening / re-engagementCTAthe pitchmetaphoranalogy

00:00HOOKSo today, for the first time ever, Anthropic has passed OpenAI in business adoption, and this research article dropped this morning. You can see the overall trend where OpenAI has been leading in business adoption since pretty much, you know, early twenty twenty three. But you can see right here in April, Anthropic has now passed OpenAI. Right after that article dropped, Sam Altman tweeted, Codex is the best AI coding product, and we want to make it easy to try. For the next thirty days, we are giving companies that wanna try switching over from Cloud Code two months of free Codex usage. This tweet dropped about an hour ago. And then about forty five minutes after that tweet, we saw this one from Claude that said, Cloud Code weekly limits are increasing 50%

00:37HOOKall the way through July 13. So that's two months from now. So Codex says, hey. Free Codex for a month. And Cloud Code says, hey. You're gonna get 50% extra usage on Cloud Code for the next two months. And I was reading these two tweets, I was like, that's pretty funny. I mean, they're obviously going right back and forth at each other. But this made me think about something else, of the whole way that we should be thinking about AI. And that's what I wanna spend some time talking about with you guys today. And before you guys ask, I am currently working on a codex versus cloud code video, so that will be coming very soon. Appreciate you guys' patience. So I have this document here, which I'm calling the free sample phase. Kind of thinking about this AI coding war. So we got these two tweets. Right? Codecs free, extended Cloud Code usage for the next two months. And I want you to really think about, okay, let's say you're paying $200 a month for Cloud Code or whatever amount you're paying. The amount of output that you can actually get from $200 a month of actually using Cloud Code every week hitting your limit is absolutely absurd. I'm not a formal software engineer or anything, but I guarantee you, you could be getting that same amount of output from a full time software engineer hire, which might be anywhere from 5 to 10 to 15 k a month. And if you took these coding agents away from companies and from

01:41HOOKorganizations, they would feel the pain because they have now gotten so used to defaulting to Claude code or Codex for pretty much everything. If you surveyed a ton of people, thousands of people that are using Claude code every single day or Codex every single day and said, hey. How much would you actually pay for what you're getting out of this? I would honestly be shocked if the vast majority of them said something less than $200. Like, $200 feels like a steal. So anyways, you are not the customer. You are the training data. Now I obviously don't know this for a fact, but I've seen some conversation online and it really feels like we're kind of just in a free sample phase. Because there are two major things that come to my mind when I think about what these companies need right now, which is adoption

02:18and data. Adoption, obviously, because they want people using their stuff. They want people to be addicted to their stuff, which I'm sure a bunch of you guys are addicted to the AI tools that you use every day to the point where if it got taken away from you or if the API is down or whatever, you feel the pain and you almost think like, how do I even work normally now? And that's exactly what they want you to feel. And then of course, the other thing is data. That proprietary data that they're getting is one of the most valuable resources to them. And ultimately, it's probably the biggest moat because these tech companies that are trying to build better AI models that don't have data as far as how people are using it and how they converse and what they're asking the model to do better at, All of that data, all of those patterns are making the models better and better and better. And it's interesting to think about. We saw this tweet from Sam Altman, you know, back, like, a year and a half ago that basically said, we're currently losing money on OpenAI Pro subscriptions.

03:03People are using it much more than we expected. And we all know that if you took your typical usage for the month and you were routing that straight through API rather than using a, you know, fixed cost subscription, you would be paying so much more money, five to 25 x more money on those tokens compared to what you're actually being charged on your subscription.

03:20And, obviously, these companies are eating costs on their own end. All of that compute that they have to actually use in order to send you a response back from their model, they're eating that cost. And they're okay with it because right now they need the adoption and they need the data more than they need your $200 a month. So yes, Anthropic rose 3.8%

03:37in April to 34.4% of businesses, and OpenAI fell 2.9% to 32.3%. And overall adoption across businesses was still rising. Now this is purely an adoption

03:49graph and metric. This doesn't actually mean that Anthropic is the definitive leader in business adoption or that they have the best AI model or the best coding harness. This article pointed out a few other things, like Anthropic's incentives are misaligned with those of business customers. They obviously make more money when more businesses use more tokens. And they're incentivized to drive users to their more expensive models. We've also seen some other weird stuff like people saying that cloud has gotten worse, as well as a bunch of stability problems with their compute and

04:17all of these new feature release. And this isn't a new thing. If you always try to look at, like, what's happening in the industry or in your, like, line of work and look at the patterns across other areas of business and other, like, you know, ebbs and flows of of cycles that we've seen in history, this has happened with a lot of other things because companies need adoption. Now, obviously, there's just like general inflation, but basically the idea is like land grab, the habit forms, competition thins,

04:42and then the reset. So things like Facebook ads have gotten way more expensive. Google AdWords, Uber, DoorDash, Netflix, AWS.

04:50There's lots of other things that go into the pricing for all of these businesses, so I'm not saying it's like a exact correlation and that's exactly what they did and that's exactly what OpenAI, Anthropic, and all these other AI models are doing. I'm just saying look back at some of these other patterns. I mean, like, this is exactly what I'm doing with my YouTube channel. I'm gonna start charging for every single YouTube video I post ever. I'm just kidding. Wouldn't that be crazy, So really what I'm trying to say is use it like crazy. Now this bullet point, I don't necessarily agree with because you also have to think like, okay. Open source models are getting better. They're also getting smaller, which means that I'm gonna be able to run them for cheaper. And I don't exactly know which way these closed source models are gonna go as far as, like, pricing, but still, what I know right now is we are getting this stuff for very cheap, especially in the next one month for Codex and two months for Cloud Code, so just use them both. I don't really know who's gonna come out on top of this war. I don't know. It might not even be OpenAI or Anthropic.

05:42But what I do know is that I'm going to start building my habits. I'm gonna build my patterns. I'm gonna build up a bunch of different projects that can move from Codex to Clawd Code to Hermes to Open Claw to whatever's next. And the ability to be able to have a project that is so flexible that if one day Anthropic just disappeared, like literally out of thin air just disappeared, if you could move everything over to a different tool in the matter of like an hour and the rest of your week is completely unchanged,

06:07that's a win. And you need to make sure you're setting yourself up in something like that so that you're not becoming extremely dependent on one thing. Now this is just where I landed after I kind of sat with these two tweets and really thought about what's going on here. I would love to hear in the comments if you guys have differing opinions, and I think that there's definitely room for some really great discussions around this. I'm not out here saying that everything that I just told you guys is a 100% fact and a 100% true besides, like, those actual metrics that I called out. You know, a lot of this is just my opinions and and what I feel about the space. So please let me know if you guys disagree and why. Anyways, the closing frame that I had here was you're not paying $200 a month for AI. You're paying $200 for twelve to twenty four month exception from real prices. Now that's me just kind of throwing something out there. I don't know if that's true. Right? But, like, that's just where I feel we might be heading towards. Right? Like that's not me saying that I know the strategy and I know what's gonna happen, but it's something to think about. So anyways, I just wanted to come in here and talk about that real quick because

07:00CTAI could almost feel some people looking at these two tweets and just getting a little bit more of that overwhelm or more of that panic. And instead of approaching it as, like, oh my gosh. Like, what do I do? These these tools are going at each other. Just, like, think about it more like, this is actually a blessing. I get to try more and more, and I get to see what I like better for basically the same price. I don't have to pay any more to do this. Now I guess I don't exactly know when they say we're giving companies a free month of codex usage. I don't exactly know, like, what qualifies for a company to get a free month of codex usage, but still. So anyways, that's gonna do it for this one. I know it was quick, but hopefully it made you think a little bit. And if you guys enjoyed and learned something new, please give it a like. Anyways, I appreciate you guys making it the end, and I'll see you in the next one. Thanks, guys.

§ · For Joe

Steal the free sample framing.

Own your stack playbook

The $200/month is not a subscription fee — it is a 12–24 month window to build skill and portfolio before real prices kick in.

The 'free sample phase' concept is a ready-made hook for any 'own your stack' or 'stop renting' video.
The four-phase subsidy cycle (land grab → habit → competition thins → reset) is a reusable visual framework for any SaaS pricing commentary.
The six-section deck format (dark bg, stat cards, numbered sections) is clean and executable — lower production cost than B-roll, higher info density.
The three-reads structure (bullish / bearish / where I land) is a clean opinion-video template that avoids wishy-washy fence-sitting.
Build the 'move in an hour' benchmark into every stack decision: if switching tools takes more than an hour, you are dependent, not efficient.
Newsjack the morning something breaks — this video was live within hours of the Ramp article dropping.

§ · For You

What the AI price war means for you.

If you use Claude Code, Codex, or any AI tool

Current AI pricing is artificially cheap by design — the window to lock in skills and build at low cost is open right now, but it will not stay this way.

Use both Claude Code and Codex aggressively while the promotions run (through July 13 for Claude Code).
Build your projects so they are not hard-wired to one provider — if you could switch tools in an hour, you are protected when prices reset.
The Twitter war between Altman and Anthropic is competition working in your favor. Do not panic — this is a blessing.
Open source models are improving fast and getting cheaper to run locally — this is your long-term hedge against price resets.
Treat your $200/month as buying 12–24 months of below-market access, not a permanent price point you can plan a business around.

§ · Frame Gallery

Visual moments.

Ramp chart open

Frame at 00:08 from Anthropic Just Dethroned OpenAI. Here's What Happens Next.

Sam Altman tweet

Frame at 00:20 from Anthropic Just Dethroned OpenAI. Here's What Happens Next.

Frame at 00:26 from Anthropic Just Dethroned OpenAI. Here's What Happens Next.

Claude Code 50% tweet

§ · Watch next

More from this channel + related dossiers.

Nate Herk | AI Automation · Listicle

I Tried 100+ Claude Code Skills. These 6 Are The Best

A 13-minute walkthrough of the six Claude Code skills Nate Herk says are the only ones businesses will actually pay for — plus how to package them into an AI-automation offer.

Nate Herk | AI Automation · Tutorial

Higgsfield Just Turned Claude Into a Creative Agency

Nate Herk wires the Higgsfield CLI into Claude and Claude Code, then runs a one-prompt creative agency that researches a brand, generates the catalog, fills a Google Sheet of ad variants, and ships hyper-motion video on a routine.

Nate Herk | AI Automation · Tutorial

This is the Most Powerful Tool to Give to Claude Code

A 14-minute tour of Printing Press — a CLI factory that turns any site (even ones without an API) into a token-efficient command-line tool your agent can call.

aiwithbrandon · Tutorial

Google Gemini CLI Is FREE & Crazy Powerful: Real World Coding Test & First Impressions

Brandon Hancock spends 35 minutes putting Gemini CLI through three live tests — a one-line styling fix, a full memory-feature build, and a from-scratch landing page — and lands on a single rule: Gemini CLI thrives with context, dies without it.

Simon Scrapes · Tutorial

Claude Code has a new UI (pair it with Claude OS)

A 5-minute walkthrough of Anthropic's native Agent View TUI and how it slots into a folder-based Agentic Operating System.

Ray Fernando · Tutorial

Two MCPs That Save 97% of Your Context Window

An ex-Apple engineer benchmarks ref.tools and Exa AI against Cursor on a live Tailwind v4 refactor — and Claude Code wins at 2,800 tokens vs 98,000.

November 23rd 2025