Modern Creator
Why Modern Creator?
Nick Puru | AI Automation · YouTube

Why I Cancelled My $200 Claude Plan

An agency owner who spent over $10,000 on Claude this year breaks down the vendor-diversification setup he built to replace it.

Posted
2 days ago
Duration
Format
Talking Head
sincere
Views
5.6K
75 likes
Big Idea

The argument in one line.

A single AI vendor outage or price change can freeze an entire business, so the fix is splitting coding, planning, admin, and private-data work across separate models routed by one shared rules file.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…
  • You run an agency or business where AI tools touch client work, and a single vendor outage would stop billable work cold.
  • You've built a personal AI assistant or automation stack so deep into one platform that switching feels impossible.
  • You want a concrete example of a multi-model routing setup instead of a vague 'don't put all your eggs in one basket' warning.
  • You handle sensitive client data (contracts, financials, call transcripts) and want a model for keeping it off external APIs.
SKIP IF…
  • You use AI tools casually and have no dependency risk worth engineering around.
  • You're looking for a coding tutorial rather than a vendor-strategy and cost breakdown.
TL;DR

The full version, fast.

The creator spent over $10,000 on Claude in a year running three businesses, then watched Anthropic pull Fable and Mythos overnight under a government export order. That outage exposed how dependent his whole operation, client sites, an internal assistant, ad campaigns, was on one vendor. His fix splits the work into two layers: a brain layer (which model thinks) and a body layer (which tool executes). Fable stays as the planner/reviewer since it doesn't type or execute; GLM 5.2 does the bulk coding and writing; Kimi K3 acts as a second planner; DeepSeek V4 handles menial renaming and sorting; a locally hosted model (Qwen) never leaves his Mac for anything with a client's name on it. All of it routes through one shared rules file that every tool, Claude Code, Codex, GrokBot, reads before touching a job, so swapping any model is a one-line edit. He now pays for a ChatGPT plan, a GrokBot plan, Claude at $20/month instead of $200, one OpenRouter account, and nothing for the local model, aiming for the same output at lower cost and zero single-vendor risk.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →
Chapters

Where the time goes.

00:00 – 00:20

01 · I cancelled my $200 Claude plan

States the cost ($10K+/year, 3B tokens/month) and the headline decision: cancelling the $200 plan for the first time.

00:20 – 00:34

02 · The thought I couldn't shake

Names the recurring worry: total dependence on one AI vendor for everything.

00:34 – 01:13

03 · When Fable got pulled

Fable (the model line he'd moved his entire operation onto) was pulled after a government directive suspended access; his team sat blocked with no timeline.

01:13 – 01:48

04 · None of this is hypothetical

Lists concrete things that already happened: weekly limits changing, prepay requirements returning, big customers pushed to 2-3x pay-per-use pricing.

01:48 – 02:20

05 · The money doesn't add up

At 3 billion tokens/month against Anthropic's published rates, the real cost would run north of $10,000, yet he only paid $200 flat.

02:20 – 02:37

06 · What that price does to you

A flat price removes the incentive to ever question cost, so you stop asking what a task is worth.

02:37 – 03:25

07 · What I used it for

Client sites, landing pages, finance reports, Instagram systems, ad campaigns, outreach lists, research, DM replies, roughly $300-400K in returns traced back to this workflow.

03:25 – 04:22

08 · The assistant that made me notice

A personal assistant built into a folder with access to email, calendar, ClickUp, CRM, and Telegram, briefing him each morning, became the thing that made leaving feel impossible.

04:22 – 04:39

09 · Exactly what any company wants from you

Every routine and skill taught to the assistant made it stickier; switching vendors got more expensive every month he waited.

04:39 – 05:12

10 · What I use instead

Rejects the idea of one replacement tool; splits AI work into four separate jobs: the code, the plan, the admin, and the private files.

05:12 – 05:41

11 · Fable stays (as the manager)

Fable remains the planner and reviewer, still one of the most capable models, but never does the typing or execution.

05:41 – 06:17

12 · GLM, Kimi & DeepSeek

GLM 5.2 does the bulk of coding and writing; Kimi K3 is a second planner that reads problems differently; DeepSeek V4 handles simple, high-volume tasks cheaply.

06:17 – 06:54

13 · Local models for client data

A locally hosted model on his own Mac (LM Studio/Ollama) handles anything with a client's name on it, so sensitive data never reaches an external API.

06:54 – 08:12

14 · The bodies: Claude Code & Codex

Claude Code is where he does deep desk work (memory lives in the folder); Codex is the engineer that builds anything, taught once via a skill file, then repeats the job on command.

08:12 – 08:53

15 · GrokBot: the always-on desks

GrokBot runs a team of always-on bots in the cloud handling leads, DMs, and emails, keeps working when his laptop is closed, least friction, no config.

08:53 – 09:37

16 · My assistant: private work, local

The personal assistant is the opposite of GrokBot: fully local on his Mac, handling anything sensitive like client files, financials, and contracts.

09:37 – 10:13

17 · Routing: one file decides who does what

A shared rules file, read by every tool before it touches a job, is what makes the whole multi-vendor setup actually function once you're paying per token.

10:13 – 10:58

18 · The staff, one line each

Walks through the rules file line by line: Fable plans and reviews, GLM builds, GPT-5.6/Astra takes the hard builds, Kimi is the second opinion, DeepSeek gets the small stuff, Qwen gets anything with a client's name on it.

10:58 – 11:20

19 · The rules underneath

Client data always goes local with no exceptions; simple jobs never touch the planner; failed reviews escalate one level and stop to ask after two failures.

11:20 – 12:15

20 · The test: same job, two folders

Running the identical job in an empty folder versus one with the rules file in place: the empty folder burns several times the tokens because it has to guess, and guesses with the most expensive model every time.

12:15 – 13:15

21 · What I pay now

Breaks down the actual current bill: a ChatGPT plan (Codex, GPT-5.6, Astra), a GrokBot plan, Claude's $20 plan plus credits for Fable, one OpenRouter account for GLM/Kimi/DeepSeek, and Qwen running free on his own Mac.

13:15 – 14:35

22 · The takeaway: never one vendor

Closes on the core lesson: audit what stops working if one company changes a single setting or price, and build in at least one local or second-vendor fallback before you're forced to.

Atomic Insights

Lines worth screenshotting.

  • Spending $200/month on a flat-rate AI plan removes the incentive to ever question what a task actually costs, which is worse for your business than the bill itself.
  • A single-vendor outage that only lasts a few days can still freeze a team that built its whole workflow around that one tool.
  • The real lock-in isn't the subscription price, it's the accumulated skill files, routines, and assistant setup that make leaving expensive.
  • Splitting AI work into a brain (which model thinks) and a body (which tool executes) lets you swap either half independently.
  • A planner model that never writes or executes code (only plans and reviews) catches expensive mistakes before a cheaper model burns tokens building the wrong thing.
  • Running the same job through an empty rules file versus a populated one shows the empty one burns several times more tokens by guessing, and guesses with the most expensive model available.
  • Client-sensitive data (contracts, financials, call transcripts) should route only to a locally hosted model that never leaves the machine, never to an external API.
  • One shared rules file that every AI tool reads before starting a job turns a multi-vendor setup into a one-line edit instead of a rebuild.
  • Cheap, high-volume tasks like renaming, summarizing, and sorting belong on the cheapest available model, not the same model doing your hardest builds.
  • Running three businesses without a $200/month flat-rate plan, and paying per-task across five providers instead, is a real, testable alternative to vendor lock-in.
Takeaway

Split your AI stack before one vendor forces the split for you

WHAT TO LEARN

Routing tasks across a planner, a builder, a cheap bulk model, and a local model behind one shared rules file removes both the cost blindness of flat pricing and the risk of a single vendor freezing your business.

01I cancelled my $200 Claude plan
  • A flat-rate plan removes the incentive to ever ask what a task costs, which is a worse long-term problem than the bill itself.
03When Fable got pulled
  • A single-vendor outage doesn't have to be permanent to be damaging: even a temporary access loss can freeze a team mid-project with no way to estimate a return date.
04None of this is hypothetical
  • Test your real usage against a provider's published per-token rates occasionally, so you know whether a flat plan is actually subsidized or you're just not tracking spend.
09Exactly what any company wants from you
  • The deepest lock-in isn't the subscription, it's the accumulated skills, routines, and integrations built on top of one tool, so audit those before they become unswappable.
10What I use instead
  • Separate 'which model thinks' from 'which tool executes' so you can swap either layer independently instead of rebuilding your whole workflow to change one vendor.
11Fable stays (as the manager)
  • Route the highest-value planning and review work to your most capable model, but keep it out of typing and execution so it never burns expensive tokens on rote work.
12GLM, Kimi & DeepSeek
  • Give cheap, repetitive tasks (renaming, sorting, summarizing) to your cheapest available model rather than defaulting everything to your best one.
13Local models for client data
  • Anything with a client's name on it should run on a locally hosted model that never leaves your machine, with no exceptions written into the rules.
17Routing: one file decides who does what
  • Put your routing logic in one shared file that every tool reads before starting a task, so a vendor swap is a one-line edit instead of a rebuild.
20The test: same job, two folders
  • Test your setup by running the same job with and without your routing rules in place; a system with no rules guesses, and guesses expensive.
22The takeaway: never one vendor
  • Even if you keep your primary vendor, add at least one second vendor or one local fallback so a pricing change or outage doesn't stop everything at once.
Glossary

Terms worth knowing.

Rules file
A shared text file inside a project folder that every AI coding tool reads before starting a task, telling it which model or role should handle which kind of work.
Brain vs. body
The creator's split between which model does the thinking (planning, reviewing) and which tool actually executes the work (writing code, browsing, running routines).
OpenRouter
A service that gives one API key and one bill for access to many different AI models from different providers, instead of separate accounts for each.
Local model
An AI model run entirely on the user's own computer instead of over the internet, so data never leaves the machine.
Resources

Things they pointed at.

10:01toolOpenRouter
09:53toolLM Studio
09:53toolOllama
07:36toolGLM 5.2
07:50toolKimi K3
07:56toolDeepSeek V4
06:54toolClaude Code
07:05toolCodex
07:59toolOpenCode
08:20toolGrokBot
Quotables

Lines you could clip.

00:00
“I have spent more than $10,000 just on cloud this year between the subscriptions and the API.”
shocking dollar figure as cold open→ TikTok hook↗ Tweet quote
03:38
“What that price does to you is worse than the bill cost. It stops being a question.”
sharp reframe of flat pricing as a psychological trap→ IG reel cold open↗ Tweet quote
11:33
“Any of the empty ones burn several times the tokens because it just had to guess what I meant, and it guesses with the most expensive model every time.”
concrete proof point from his own test→ newsletter pull-quote↗ Tweet quote
13:51
“So this is my effort to do. Ask what stops if one company changes one setting or their pricing.”
closing thesis, quotable as a standalone warning→ newsletter pull-quote↗ Tweet quote
The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphoranalogy
I have spent more than $10 ,000 just on cloud this year between the subscriptions and the API. I was a top 0 .01 % cloud power user and over 3 billion tokens just last month alone. Just from me, my team runs our entire operation on this as well.
All three of my businesses, they run on this. 20 of my last 30 videos, they were about cloud. It is the best tool that I've ever paid for.
And last week, I canceled my $200 plan for the first time. So for some context, for the last few months, there's just been one thought that keeps coming back into my mind. head and that's that i'm so dependent on claude and i know exactly when this started so when fable launched i pretty much moved everything onto it every client system the team setup my own assistant all of it inside of maybe a week and it was that good and then the government they stepped in and fable was gone you all know about all that and that entire debacle and it was gone for everyone and i just sat there completely blank with a team that had just done all that work and just waiting on a company that couldn't tell me when it was coming back so the questions started to roll in like what if that happens again what if one setting changes and all the times that cloud is down like what the hell do we do what if the price goes up and there's just so many other concerns that i have with being locked to a single vendor and you should too and none of those are hypothetical anymore like the 20 plan that lost fable overnight this summer it's back now but only if you prepay for it on top i believe but the weekly limits they get cut next week their biggest customers are all
already on the pay per use at two to three times the old bill now all of it has already happened and I just kept paying you probably are too because when something works this well you stop asking questions about it you know what's the saying it's if it isn't broke do not fix it that's why i just decided to test things out find out for myself so that's why i'm going to be running all three businesses without the 200 plan see what breaks and see if i even miss it at all let's first start with money because that's really what made me look at this so having about three billion tokens in the last month if you take that number two anthropics own pricing page and if you do the math at the rates that they publish you land somewhere north of about ten thousand dollars now of course because i'm on the 200 plan i just paid that 200 so one of two things is true either the public prices made up or someone is just covering the gap to just keep us building on top of it and i think it's maybe some of each and i'm not complaining about this it's
literally a great product and i paid for it again tomorrow but what that price does to you is worse than the bill cost it stops being a question and the models they stop being a question as well so you open it you go you build whatever and i did that for about a year and i still ran out every single week and i fixed was just never to think about it my fix was to just wait for the reset so for me obviously i use this a lot i use it for client sites and landing pages the finance reports instagram systems ad campaigns outreach lists like the research behind all of these videos the team's dm replies literally everything in my company and to be fair the returns from all of those builds and marketing campaigns probably netted us somewhere between three to four hundred thousand dollars now all of those different builds and systems are obviously directly related to the business and it's the point of doing all of this and those are actually unheard of returns from you know 200 plan to you know equating to about 400 000 or whatever it is exactly of how much we have made
made from returns of running this business and by the way on September 16th I'm going to be running a free live webinar on how to actually grow yourself a profitable AI business as a beginner so I will be covering what to sell what's charged and how to get clients without a huge audience but anyways all of that stuff it went through the same place but then there was my own assistant that i was using as well now this assistant i built in a folder on my mac i just gave it access to all of my important things like my email my calendar click up my crm telegram and i just gave it a set of routines that actually runs my day so right now it just briefs me in the morning it preps me before all my calls it drafts what i would have drafted in my inbox and it was just the assistant that really made me notice everything and caught my attention because normally i would reach for claude on just reflex before the question of any other tool even came up but quite literally every single routine that i added every skill that i taught it it made it more useful as you would imagine and it just made leaving harder so it was very sticky and i never once chose that it just happened and you know one small convenience came by me at a time and by the time i actually looked at it switching to one other vendor is a huge pain in the ass and every month that i waited it cost me more and if you think about it that is exactly what any company wants from you and that's not sinister that's just a good business and it was working on me and probably you watching
this video so myself i was left with two problems where that deal goes next and how much i would actually be building on something that i don't control so now the most important piece of this video what do i now use instead and i'll be honest because i asked the wrong questions first i went looking for which one tool replaces cloud is it codex is it grokbot maybe something else that is just open source and to answer that there isn't one and there shouldn't be because why should one company supply every single brain inside of my business so the coding the plan the admin and the private files those are just four different jobs and there's no reason one company should win all four of that if you think about it so here's the split that i actually use now fable this is staying it is still on arguably the most capable model there is and i'm not going to pretend otherwise know maybe you can't make the argument of astra but but to that when gpt 5 .6 and astra did launch i built a whole application and so many other things with fable as the manager and astra and soul as the engineer that video is also on the channel and that split is pretty much my everyday setup now so fable plans fable reviews fable doesn't do the typing or the execution so we also have glm 5 .2 this is doing the bulk of the coding and the writing for me and my team it is
by far cheaper and most of the day's work does not need more than that anyways then we have kimmy k3 this is my second planner so it just reads a problem differently from fable and two plans of course beats one in this case and then deep seek v4 this takes all the simple stuff so any renaming any summarizing any sorting this is basically free and it is incredibly fast anything that's going to be related to client or sensitive information it's handled using locally hosted models all running on side of my Mac.
So the local AI, this is wired directly into the main setup. So any sensitive data stays local rather than being sensed to any external models. Now reaching them, this is the easy part.
So open router, this is just one key to one bill for practically every cloud model on that list. And LM Studio or Ollama, this runs the local ones on my Mac. And just keep in mind, there's actually gonna be a few other alternatives to this, like hosting these LLMs on another computer that's actually suited to handle these models.
So it needs to be pretty beefy. computer or you can just throw this on a vps like hostinger or whatever else there is and that's the pivotal thing here and i would say it's the whole point about all of this we're not going to be locked into a single vendor ever again alrighty now the brain this is only one half of this the other half is the body that it actually runs in and each body it has its own job so cloud code this is where i do all of the deep work at my desk so i drive it does the building and the folder this is the memory now codex this is the engineer so anything that has to be built it's going to be built and this year it actually got a browser on top which i made a whole video on and it's insanely powerful and one of the best tools or releases that i have ever used this year that's also on the channel if you guys want to check that out but the way that i use it it's very simple i teach it a job once it is in a skill file inside my folder so maybe things like running meta ads they're a good example of this where i wrote down how i want a campaign set up i do it just once literally only once and now when i say the word it goes and it does all the things for me and for the places with no api at all school perfect example of this my school community it opens the browser and works through
you know pretty much anything in the way that i would so responding to people or just flagging posts so that i can check them out get to them in a respectable amount of time so that's codex skills at my desk now grokbot another huge important harness that has just recently came out this has a browser as well but that's not really why it is in this video it's here because it has a team it runs on grok's own model you don't pick the brain and what you get is just a bot with job descriptions each on its own computer inside of the cloud handing work off to each other and it keeps going when my laptop is closed so they're always on desk it lives right there so maybe a front desk that's able to answer leads inside of a minute i have a lot of different functions for this like being able to review my crm and all the reporting from my appointment setters being able to respond to dms on instagram very fast respond to all of my emails notify me of anything important anything like that so it's legitimately the least friction to using ai and i don't really have to configure anything at all it just runs which is perfect my assistant this is the complete opposite so this is just living on my mac
to my folder and anything sensitive like client files or financials, contracts, call transcripts, all that stuff, it just goes through it and it never leaves my machine. And that's the line right there. So public routines, they usually run on GrokBot or any private work.
This is going to be running through the assistant completely local and safe. Now, if you don't want to pay for a body at all, OpenCode, it's open source, it's free and takes literally any of these models. So you can go with that approach, but CloudCode.
codex and open code they all read the same folder and grokbot it just keeps its job descriptions in its own cloud so a copy of each one lives in my folders as well now if i ever leave the jobs they come with me as well so i can use my phone and do work from anywhere all right now the moment that you do pay per token routing stops being optional and this is the part that proves the setup actually works so i don't want to slow this down but this is the block right here so this is going to be sitting inside of my rules file and every tool that i just showed you it reads it before it touches any job so cloud code or codex open code whatever it is i wrote down what does what and they do it exactly that way so Walk through it with me.
So the first line, this is the rule. Whoever is running the folder is the manager, and the manager never does the hands -on work. Then we have the staff.
This is just one line each. It's very simple. It's very concise.
It doesn't lead to any context bloat fable. This is planning and reviewing. GPT 5 .6, this does the building.
Now the hard builds, these now go to GPT -6 Astra, which landed in Codex just last week. And adding it was simply just one line here. So it's very easy to adjust.
and optimize for any future releases now glm 5 .2 this does any of the bulk writing and decoding and then kimmy k3 this is the second opinion on really anything worth real money and then we have deep seek v4 this just gets all the small stuff so all the menial repetitive things and then we have quen on my mac this gets anything with a client name on it again that's the sensitive private stuff that it don't want us to be showing then there's the rules underneath that so client data this goes local there's no exceptions with that simple jobs these never touch the planner so for review fails it just goes up one level and twice it stops and asks me now every single result it comes back with a root line on top saying who did it so i can actually see all the handoffs so it's some sort of audit log so here's the test if we have same job this is going to be running twice so once we're going to be running an empty folder once in mine with this block in place now any of the empty ones these burn several times the tokens because it just had to guess what i meant and it guesses with the most expensive model every time and this is what actually holds up the setup together so the folder this is going to be the box
and the models these are the staff and any one of them they can be swapped just by editing one line and nothing else in the business is moving from that now all of this it's possible inside of cloud code the whole time so the 200 plan it just made it pointless to even bother and continue paying that and i've turned this into a template as well so it's completely free inside of my school community link will be down below but you can just paste in your rules file put your own models on the line and it works in cloud code codex or open code all the same way also we're going to be diving very deep on how to actually make money with this stuff inside of our webinar make sure to check that out on september 16th if you are interested so with all of this the question just becomes like how far can this go can i run three businesses without the 200 plan and not miss it well Here's what I pay now.
I have a ChatGPT plan, and that gives me Codex with GPT 5 .6 and Astra. Then we have Grokbot on its own plan. Cloud on the $20 plan with credits on top for Fable.
When a job does call for it, maybe I'll upgrade to the $100 plan if it begs the need. And then we have one open router account for GLM, Kimmy, and DeepSeek, which bills me for exactly what I use. And then, of course, we have Quen on my Mac, which costs me exactly nothing because it's on my Mac.
local i'm not trying to build my own cloud here i just want the same output for the same money or less and i want to see where my folder is actually weak there's going to be a few different things that i'll be discovering and proving over the next few weeks because this is discovery like this is me testing and seeing and piloting like what works best what separation can i create from cloud where i'm going to have multiple vendors but still getting the best of pretty much every world because there's a lot of different options out there so i want to make sure that i'm exposing myself to the best of the best although in this space everything is just moving at such a rapid pace so maybe it just won't be worth it in the end but i'm still going to be testing it out myself and i encourage you to do that as well but i'll be sharing my results with all of you guys of course so the most important thing that i want you to take away from this is that you shouldn't be locked into a single vendor.
even if it's just having two different cloud providers, whether that's GPT and Cloud, or maybe even including some sort of local version inside of that, where you're going to be having Quen on your computer, or even having a VPS, whatever it may be, you should always have some sort of local instance of maybe using Quen or Kimmy, whatever else it may be, because you want to keep all of your stuff secure.
And that's really the only way that you know it's going to be secure. But ultimately, it's just very important that we're not locked into a single vendor. So this is my effort to do.
So anyways, guys, full setup the folder the routing file the manager prompt it's all going to be completely free inside of my community link down below in the description and again on september 16th next week i'm going to be doing a completely free live session on building a profitable ai business as a beginner so that link for there is going to be down below in the description so have a look at your own setup this week just ask what stops if one company changes one setting or their pricing because for me the answer was almost everything and i'd rather have found that out on my own terms but anyways i'll see you in the next one
The Hook

The bait, then the rug-pull.

He was a top 0.01% power user running three businesses on one AI platform, until a government export order pulled it overnight and forced the question he'd been avoiding: what happens when the vendor changes something and you have no say in it.

Frameworks

Named ideas worth stealing.

06:33concept

Brain vs. Body split

  1. The brain: which model plans/thinks (Fable, GLM, Kimi, DeepSeek, Qwen)
  2. The body: which tool executes (Claude Code, Codex, GrokBot, OpenClaw)

Separates the decision of which AI model does the thinking from which software tool actually carries out the work, so either layer can be swapped independently.

Steal forany workflow that risks becoming dependent on a single AI provider
10:01model

Shared rules file routing

  1. One rule at the top: the manager plans, never executes
  2. One line per model describing its exact job
  3. Escalation rule: failed review goes up one level, stops after two failures to ask the human
  4. Hard rule: client data always stays local, no exceptions

A single plain-text rules file, read by every coding tool before it starts a task, that assigns each model a specific job so the whole multi-vendor setup runs without per-task manual routing.

Steal forany team routing tasks across more than one LLM provider or coding agent
CTA Breakdown

How they asked for the click.

VERBAL ASK
03:16product
“on September 16th I'm going to be running a free live webinar on how to actually grow yourself a profitable AI business as a beginner”

Soft-pitched mid-video and repeated near the end, tied to the free webinar rather than a hard product sell; the full routing template itself is offered free inside his Skool community.

FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Storyboard

Visual structure at a glance.

the $10K claim
hookthe $10K claim00:00
Fable pulled
promiseFable pulled00:34
local vs cloud split
valuelocal vs cloud split06:33
the shared rules file
valuethe shared rules file10:01
what I pay now
ctawhat I pay now13:16
Frame Gallery

Visual moments.

the $10K claim00:01Frame at 00:07 from Why I Cancelled My $200 Claude Plan00:07Frame at 00:14 from Why I Cancelled My $200 Claude Plan00:14Frame at 00:22 from Why I Cancelled My $200 Claude Plan00:22Frame at 00:24 from Why I Cancelled My $200 Claude Plan00:24Frame at 00:32 from Why I Cancelled My $200 Claude Plan00:32
the $10K claim00:01Frame at 00:07 from Why I Cancelled My $200 Claude Plan00:07Frame at 00:14 from Why I Cancelled My $200 Claude Plan00:14Frame at 00:22 from Why I Cancelled My $200 Claude Plan00:22Frame at 00:24 from Why I Cancelled My $200 Claude Plan00:24Frame at 00:32 from Why I Cancelled My $200 Claude Plan00:32Fable pulled00:34Frame at 00:44 from Why I Cancelled My $200 Claude Plan00:44Frame at 00:58 from Why I Cancelled My $200 Claude Plan00:58Frame at 01:08 from Why I Cancelled My $200 Claude Plan01:08Frame at 01:18 from Why I Cancelled My $200 Claude Plan01:18Frame at 01:30 from Why I Cancelled My $200 Claude Plan01:30Frame at 01:42 from Why I Cancelled My $200 Claude Plan01:42Frame at 01:53 from Why I Cancelled My $200 Claude Plan01:53Frame at 02:04 from Why I Cancelled My $200 Claude Plan02:04Frame at 02:14 from Why I Cancelled My $200 Claude Plan02:14Frame at 02:22 from Why I Cancelled My $200 Claude Plan02:22Frame at 02:28 from Why I Cancelled My $200 Claude Plan02:28Frame at 02:34 from Why I Cancelled My $200 Claude Plan02:34Frame at 02:43 from Why I Cancelled My $200 Claude Plan02:43Frame at 02:54 from Why I Cancelled My $200 Claude Plan02:54Frame at 03:07 from Why I Cancelled My $200 Claude Plan03:07Frame at 03:18 from Why I Cancelled My $200 Claude Plan03:18Frame at 03:30 from Why I Cancelled My $200 Claude Plan03:30Frame at 03:46 from Why I Cancelled My $200 Claude Plan03:46Frame at 04:00 from Why I Cancelled My $200 Claude Plan04:00Frame at 04:14 from Why I Cancelled My $200 Claude Plan04:14Frame at 04:24 from Why I Cancelled My $200 Claude Plan04:24Frame at 04:30 from Why I Cancelled My $200 Claude Plan04:30Frame at 04:36 from Why I Cancelled My $200 Claude Plan04:36Frame at 04:40 from Why I Cancelled My $200 Claude Plan04:40Frame at 04:55 from Why I Cancelled My $200 Claude Plan04:55Frame at 05:06 from Why I Cancelled My $200 Claude Plan05:06Frame at 05:16 from Why I Cancelled My $200 Claude Plan05:16Frame at 05:26 from Why I Cancelled My $200 Claude Plan05:26Frame at 05:36 from Why I Cancelled My $200 Claude Plan05:36Frame at 05:47 from Why I Cancelled My $200 Claude Plan05:47Frame at 05:59 from Why I Cancelled My $200 Claude Plan05:59local vs cloud split06:11Frame at 06:23 from Why I Cancelled My $200 Claude Plan06:23Frame at 06:35 from Why I Cancelled My $200 Claude Plan06:35Frame at 06:47 from Why I Cancelled My $200 Claude Plan06:47Frame at 07:01 from Why I Cancelled My $200 Claude Plan07:01Frame at 07:17 from Why I Cancelled My $200 Claude Plan07:17Frame at 07:33 from Why I Cancelled My $200 Claude Plan07:33Frame at 07:48 from Why I Cancelled My $200 Claude Plan07:48Frame at 08:04 from Why I Cancelled My $200 Claude Plan08:04Frame at 08:17 from Why I Cancelled My $200 Claude Plan08:17Frame at 08:27 from Why I Cancelled My $200 Claude Plan08:27Frame at 08:37 from Why I Cancelled My $200 Claude Plan08:37Frame at 08:47 from Why I Cancelled My $200 Claude Plan08:47Frame at 08:58 from Why I Cancelled My $200 Claude Plan08:58Frame at 09:09 from Why I Cancelled My $200 Claude Plan09:09Frame at 09:20 from Why I Cancelled My $200 Claude Plan09:20the shared rules file09:31Frame at 09:37 from Why I Cancelled My $200 Claude Plan09:37Frame at 09:55 from Why I Cancelled My $200 Claude Plan09:55Frame at 10:07 from Why I Cancelled My $200 Claude Plan10:07Frame at 10:18 from Why I Cancelled My $200 Claude Plan10:18Frame at 10:29 from Why I Cancelled My $200 Claude Plan10:29Frame at 10:41 from Why I Cancelled My $200 Claude Plan10:41Frame at 10:52 from Why I Cancelled My $200 Claude Plan10:52Frame at 11:01 from Why I Cancelled My $200 Claude Plan11:01Frame at 11:09 from Why I Cancelled My $200 Claude Plan11:09Frame at 11:16 from Why I Cancelled My $200 Claude Plan11:16Frame at 11:25 from Why I Cancelled My $200 Claude Plan11:25Frame at 11:36 from Why I Cancelled My $200 Claude Plan11:36Frame at 11:47 from Why I Cancelled My $200 Claude Plan11:47Frame at 11:58 from Why I Cancelled My $200 Claude Plan11:58what I pay now12:07Frame at 12:15 from Why I Cancelled My $200 Claude Plan12:15Frame at 12:33 from Why I Cancelled My $200 Claude Plan12:33Frame at 12:46 from Why I Cancelled My $200 Claude Plan12:46Frame at 12:58 from Why I Cancelled My $200 Claude Plan12:58Frame at 13:10 from Why I Cancelled My $200 Claude Plan13:10Frame at 13:24 from Why I Cancelled My $200 Claude Plan13:24Frame at 13:40 from Why I Cancelled My $200 Claude Plan13:40Frame at 14:02 from Why I Cancelled My $200 Claude Plan14:02Frame at 14:05 from Why I Cancelled My $200 Claude Plan14:05Frame at 14:21 from Why I Cancelled My $200 Claude Plan14:21
One-click upgrade to your Google

Get more breakdowns in your search results

Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.

Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
Watch next

More from this channel + related breakdowns.

15:24
Jordan Urbs · Essay

Why I Cancelled My Claude Code Subscription

A $200/month Claude Max power user runs the real numbers on his own token usage, then rebuilds his workflow around open-weight models and tiered model routing instead of switching to a different single vendor.

September 6th
08:41
Chase AI · Tutorial

GPT-6 Astra Is Insane for Motion Design

A free GitHub skill turns a one-line prompt into a finished motion-graphics video by pairing GPT-6 Astra, Claude Code, or Codex with Higgsfield's Seedance 2.5.

September 8th