For the writer who's also the whole masthead.
The Writing Takes Two Hours. Everything Around It Devours The Other Thirty-Eight.
The source hunt, the fact check, the links, the cross-posting, the reader questions you meant to answer. Hand it to a small team of AI assistants. Finished work waiting for you, nothing waiting on you.
The Half Of Your Week That Never Made The Calendar
Your content calendar says what goes out this week. An essay Tuesday. The newsletter Wednesday. A video Thursday. Three shorts cut out of it Friday.
That's not the list making you tired.
THIS WEEK
Content calendar / the rest of it
- Tue Essay ships
- Find the study you saved in February. Check the figure. Pull the two quotes. Confirm the link still resolves.
- Wed Newsletter
- Last week's replies, unread. Two reader questions worth answering in the issue.
- Thu Video
- Pull the transcript. Cut the three clips. Write the description. Finish the thumbnail.
- Fri Three shorts
- Decide which of Tuesday's sections carries a short. Check what already went out on each channel.
- Someday Course lesson
- The comment from March that should have been a lesson. The affiliate link that broke. Four saved sources you haven't opened.
The left column has a day. The right column has you.
Nothing on the right has a deadline, which is the only reason it survives. It never knocks. It just gets longer while the left column keeps its dates.
And it grows in a way that catches people out. Add a channel and you add its format, its checks, its comments. Add an offer and you add its links, its questions, its follow-up. The left column grows because growing was the plan. The right column grows because the left one did.
So you let some of it slide, and you should, because the next deadline is real.
The bill just arrives later. A source you won't find again. A strong piece posted once and never used. A reader question that would have been the best thing you published this month. An hour before a release, spent on something you could have handled on Tuesday.
It's Friday afternoon. You shipped everything you said you would.
You couldn't honestly say you wrote much of it.
Almost None Of That Needed Your Judgment
Look at the right column again, one line at a time, and ask which of them needed your taste.
A few of them do. Deciding what the essay is actually about. Choosing which reader question is worth a whole issue. Judging whether a piece is good. Those are the reason you do this.
The rest is a pattern you could describe to somebody else in two minutes:
- Check these sources and bring back anything that meets my criteria.
- Tell me what went live this week and what's missing a link, a description or an asset.
- Collect the questions readers are actually asking and group the repeats.
- Pull the transcript, find the clip candidates, tell me why each one.
- Assemble the source packet for Tuesday's piece and flag anything that doesn't check out.
- When a routine check finds nothing, say nothing.
Every one of those has an input, a result you can look at, and a point where you decide what happens next.
That's the whole opening. One of those jobs gets an owner, you inspect what comes back, you correct it, and you stop carrying the reminder.
The specificity is what makes it work. "Help me with my content" gets you a vague helper to manage, which is a sixth job. "Check these eleven sources every Monday and bring me anything that meets these three criteria" gets you something you can grade. Did it look in the right places? Did it catch the thing you'd have caught? Can you fix the brief and get a better result next Monday?
You can answer those. That's the difference.
You've Already Paid For Three Things That Should Have Fixed This
You have probably paid for all three. None of them failed because you used them wrong.
The chat window. It's a very good writer and a very good thinker, and it has never once opened your analytics, your CMS or your saved sources. You type, it types back, and then you go and do the thing. Every step after the conversation happens because you did it.
The automation platform. A canvas, a trigger, a filter, a webhook. It's half built and still open in a tab. It broke the week a site changed its layout, and fixing it meant learning the tool again from the start.
Hiring somebody. You got as far as the job description and stopped, because it's not a job. It's thirty unrelated small things, most of which nobody could learn without sitting next to you while you did them, and teaching all of that costs more than doing it yourself forever.
Three budgets, three different years, one reason.
Not one of them could take a job off you and keep it. You were still the thing that had to remember it, start it, and carry it.
It Turns Up With Its Own Laptop
So here's what's actually different, and it's not a better prompt.
That's literal. It has a computer of its own, with a browser, a filesystem and a terminal, and that computer isn't yours. The closest thing to it is a contractor who shows up with their own kit.
It signs into the places your work lives, once, and stays signed in. Where there's a proper connection to a tool it uses that. Where there's not, which is most of the interesting places, it clicks through the site the way a person would.
Mine is called Karina. I handed it a screening announcement for a film I wanted to see and asked it to sort out tickets. It went to the cinema I always use, chose the seats I always choose, tried my discount codes, found they didn't apply to that screening, and stopped at the payment screen for my password and my approval.
That last part isn't a limitation. That's the design.
It works while your laptop is shut.
You Brief It. You Don't Build It.
There's no canvas and no node graph and nothing to abandon halfway through.
You brief it the way you'd brief a person, in a text thread, from your phone if you like. You show it a job once. What it learned gets written down. It holds that job from then on and doesn't come back next week asking how.
The Four Things That Make This Different From A Chatbot
The name on the box matters less than you think, and I'd rather you knew what to look for than what to buy.
Four things. An identity of its own, so nothing it does is ever done as you. The ability to sign into the places your work lives and stay signed in. The ability to hold a job after being told once. And a hard stop at anything it can't take back.
Nothing could do all four two years ago. That's the entire reason this wasn't worth setting up until recently, and the entire reason it's now.
The product that does all four best today is called Grok Bot, made by xAI, and that's what the course builds on. Something will do it better eventually, and when that happens what you wrote down moves across, because it was never the tool.
Nothing here asks you to stop using whatever assistant you already pay for.
The Second One Arrives When The First One Is Full
You don't build a team. You end up with one.
When a single assistant is carrying more than a single assistant should, a job leaves it and a specialist takes that job over. Jobs earn assistants. Naming five of them on day one is the most common way to waste money on this, and the course tells you not to, repeatedly.
Monday, 8:40 AM, And One File Is Waiting
Six weeks from now. You sit down, and one file is waiting.
MONDAY, 8:40 AM
Ready for you
Tuesday's essay
Source packet, six items, two flagged. The February study checks out. The figure changed in the June revision and your draft has the old one.
Wednesday's newsletter
Eleven reader questions from the last fourteen days, grouped. Three of them are the same question.
Thursday's video
Transcript pulled. Three clip candidates with timestamps and the reason for each.
Needs you. Two things.
- 1. The affiliate link in the March lesson is dead. Replacement found, different commission terms. Your call.
- 2. A reader asked something I couldn't answer from anything you have published. Worth an issue, or worth a reply?
Nothing to report
Shorts channel: no new comments since Friday.
Four saved sources: still nothing worth your time.
Read it again and notice what's not in it.
It's not asking you to approve three things that were already right. It's not padding itself out with the two jobs where nothing happened. And the thing it caught in the February study is the thing that would have gone out in your name on Tuesday.
The two decisions waiting for you're real decisions. Both take ninety seconds.
That's the product. Not a faster week. A week where the only things that reach you're the ones that needed you.
I Built This To Get My Own Week Back
I'm Nick Quick. I publish Co-Write with AI, a newsletter about working with AI without adding to the flood of sludge everybody else is producing. If you're already on that list, hello again. If you're not, here's the short version: I've been doing internet marketing for over twenty years, and I work alone.
Alone in the literal sense. I'm the entire staff of everything I run. The newsletter, a marketing studio, a CRM I'm building for Substack writers, and several other things sitting at various stages of finished. Nobody hands me anything and nobody takes anything off my plate.
This is the first thing I've built that isn't about writing.
I didn't set out to build a course. I set out to get my own week back.
I also wasted a lot of money learning how. I built a team too fast, handed the assistants overlapping jobs, trained them carelessly, and burned through my usage allowance doing it. On the cheaper plans I ran out in days. Most of what this course tells you not to do is something I did first.
What I kept running into was that almost everything published about this product was wrong in the same direction. It was written by people who had read the documentation, not by people who had run the thing for a month and watched it fail.
The most widely repeated build in this entire category works differently from the way everybody describes it. Including the way the first version of this course described it, before I checked. That correction went into the lesson rather than getting quietly patched.
Here's how it was built.
The Failures Are In The Course On Purpose
Most of every lesson is the build: what to set up, what to type, what to check before you trust it.
Where a build has a way of going wrong, the lesson shows it going wrong. These are the traps I hit first, and each one has a fix that takes seconds once you know it and finds you at the worst possible moment if you don't.
Six of them, and what stops each:
- Yesterday's numbers, served to you as today's, with total confidence. Why it looks entirely fine on a phone at 6:45 in the morning, the one clause that stops it, and why most people delete that clause because it reads like boilerplate.
- The feedback log that filled up with votes before a single human being had opened the email. Whose votes those actually were, and why every click-based number you have ever been shown has some of this in it.
- Why the test button in this product isn't a dry run. What the documentation actually says it does, and what to use as a sandbox instead.
- An assistant deleted without its scheduled jobs copied out first. Everything it owned went with it, immediately, with no undo, and thirty seconds of copying would have saved all of it.
- The one permission you can leave off and still have everything look like it worked, right up until the thing you needed it for quietly refuses and gives you no reason why.
- The logging setup that cheerfully reports "ok" every day while writing nothing at all, because a free tier ran out mid-month.
Every Fact In Here Is Dated, Because Grok Bot Won't Sit Still
Grok Bot changes fast. So every factual claim in the course sits in a box with the date it was checked and a link to check it again, and there's a short lesson on doing that yourself. You'll never be following an instruction that went stale without knowing it.
There's also a lesson on what other tools do better than this one, with the reasons. I'd rather you read that from me than find it out in month three.
You're Not Going To End Up Babysitting It
This is the objection I'd have, and it's the right one to have.
A thing that needs checking every morning is worse than doing the job yourself. Now you have the job and a supervisor's shift.
So here's the bar, and it works because it's a number you can check.
It starts with no authority at all. The first job runs with you watching. It finds, it drafts, and it shows you what it saw before it shows you what it wrote. You correct it. That's week one and there's no skipping it.
Then it has to earn the next rung.
Nothing runs when you're not there until it has been right five times out of five, scored on five axes, across three separate days, with low repair time, and saved as a procedure it can run cold with no re-explaining.
Four questions. You answer them off a file you have been filling in since week one, not off a feeling.
Evidence gates autonomy. Not enthusiasm, and not two weeks of not looking, which is time dressed up as diligence.
Then it does the job. It runs on Monday whether or not you remembered it was Monday.
And running when you're not there never means running unread. Anything that sends, spends, publishes or deletes stops and asks you first, permanently, and the course doesn't offer you a way around it.
It Signs As Itself, And It Spends From A Card With $83 On It
The other reasonable objection, and it's the most winnable one.
It sends from its own address, not yours. It gets an email address of its own on day one and it signs as itself. Nothing ever posts, publishes or writes as you.
And if you want it to, it spends. Not from your account. From a card you load with an amount you choose, on the kinds of purchase you named.
Mine holds $83. That covers a domain renewal and a stock photo and not a great deal else, which is exactly the point. The worst possible week is a number I picked in advance.
Plenty of people run this whole system without ever turning spending on, and that's a perfectly good way to run it.
Two Questions That Sort Every Job You'd Hand Over
Most courses tell you the line is whether something can be undone. That's the wrong line and it will send you the wrong way on both sides of it.
A purchase can usually be refunded. A post can't be unread.
So the real test is two questions, asked about every job before you hand it over.
How bad is the worst version of this going wrong? Not the likely version. The worst one. If the answer is a number you already decided on, that job can run on its own. If you can't put a number on it, it can't.
And did somebody else see it before you did? This is the one people miss. Anything that reaches another person is a different category, permanently, because you can't recall what somebody has already read.
Money is recoverable. Attention isn't.
Which gives you three kinds of job, and you'll sort your own week into them in about ten minutes.
Bounded and private. A renewal on a card that holds $83. A source packet assembled. A transcript pulled. A file written. Worst case is small, you can fix it in ten minutes, nobody else was involved. These run on their own once they have earned it.
Bounded and seen. A reply in a thread somebody else started. A note to a collaborator you already work with. Real people, but people you have a relationship with, about matters of fact. Higher bar to get here, and worth getting to.
Neither. Anything published. A first message to somebody who has never heard of you. Anything where you can't say in advance what the worst version costs. These stop and ask, every time, forever.
One Of Mine, And What It Does Every Morning
Here's one of mine, in full, because an example beats a promise.
It's a personal news briefing. It looks for material my usual sources and algorithms miss, using my stated interests to decide what might be worth my attention. A research routine brings back candidates. A second pass assembles a short issue, which arrives by email and lands in a web archive with its own feed.
Every item carries a way to say keep or pass, and to say why. Those answers feed the next research brief, so it gets better at finding what I actually want. On a weak day, nothing goes out.
That took design, training and review. The point of showing it's not the polish. It's that it's five separate jobs: finding overlooked material, judging what deserves attention, assembling an issue, distributing it, and learning from the response. Each one can be described, tested and corrected on its own. None of them required handing over a publication.
Your first job will be smaller than that, and it should be. A source check before your planning session. An inventory of what shipped where. A brief of the questions readers keep asking.
The course starts with the work you already repeat, rather than asking you to copy somebody else's finished system.
Six Moves, Starting With One Job From Your Own Week
Six modules. Each one is a move you make, not a topic you read about.
One. The interview. You work out what you're actually hiring, how to check anything you're told about it, and which of your jobs goes first. You finish with a full setup package written for your week specifically, before you have installed anything.
Two. Day one. Your first assistant gets its own identity, its own accounts and its own logins, and does one real job start to finish. One sitting.
Three. The first spin-off. A job leaves your first assistant and a specialist takes it over. The team starts here, in your first weekend, not at the end of the course.
Four. Standing jobs. The scored gate, the schedule, and the silence rule, which is the difference between a system you read every morning and one you stopped opening by week four.
Five. Growing it. A recurring need earns its own specialist. Extra tooling arrives only when the team has outgrown what the product does alone, and not one week earlier.
Six. Running it. The record of what actually happened, a knowledge base you and the team both write to, keeping the running cost sane, and letting go of the assistants that stopped being worth their keep.
What Lands In Your Account
- The Setup Wizard. One prompt you paste into whatever chatbot you already use. It interviews you for about fifteen minutes and writes your entire first week of setup before you have installed a single thing. This is the asset that makes day one a sitting instead of a Saturday.
- Six modules, twenty-two written lessons, built to be read and acted on rather than watched.
- Fourteen toolkit assets, the Setup Wizard among them.
- Fifteen formatted documents, ready to print or keep open beside you.
- Eighteen workflow maps.
- Thirty-three worked cases, every one sourced and dated.
- Every factual claim in a dated box with its source, and a five minute lesson on re-checking it yourself when this page is a year old.
A Few Of The Things You'll Know By Module Six
What It Has To Prove Before It Runs Without You
- The four questions to answer before any job runs when you're not there, and why three scored days beat two unscored weeks.
- The five axes you score a run on, and the fifth one that quietly decides all of it.
- The four-rung ladder, and why almost everybody who reports that this doesn't work started at rung four and never mentions it.
- The reason a schedule is the riskiest thing you'll ever hand over: it multiplies whatever it already does, including the wrong thing, at the same time every morning.
- What "runs without you" is allowed to mean, and the four verbs that stay behind your approval permanently, on every job, forever.
The Small Clauses That Do All The Work
- The five things a scheduled instruction has to name, and the two almost everybody leaves out.
- The silence rule. One line that decides whether you have a working system in six months or a daily report you stopped opening in week four.
- "Never invent a past interaction." Why a confidently invented history is worse than having none, and exactly how you find out.
- The small file that stops a daily digest handing you the same three items every morning until you quietly stop opening it.
- Why the pass that reads your reactions has to run before the pass that produces the work rather than after it. One scheduling decision, and it's the whole difference between something that learns and something that keeps a diary.
- The single unbounded sentence that costs more than anything else you'll type, and the boundary that turns it into a fraction of itself while producing a better answer.
- Why the measure isn't how much it produced, it's how much of it you used, and the ninety seconds a week that gets you that number.
Two Ways In, And The Cheaper One Gets You More
Co-Write with AI annual. $90 a year.
- Bot Zero in full, the written course and the toolkit, plus the video lessons when they land.
- The paid weekly issue. One usable system in each one, for getting more out of your own publishing operation.
- Co-Create, a monthly video call for paid subscribers. I bring a short lesson and the rest goes where the room takes it.
- A complete system drop every quarter while you're subscribed. Bot Zero is the Q4 2026 drop.
Bot Zero on its own. $149.
- The same written course and toolkit, and the same video lessons when they land.
- Yours permanently, with no subscription.
One thing worth knowing before you choose. Bot Zero is the system included with the annual through the end of 2026. From the first quarter of 2027, joining the annual gets you that quarter's drop instead. Bot Zero stays on sale on its own after that, so nothing disappears. But if you want it as part of the annual, this quarter is when that's true.
And the standalone price goes to $249 when the video lessons are finished. Everybody who buys before then gets them free. I'm not putting a date on that, because I'd rather finish them properly than hit a date I announced on a sales page. There's no countdown here for the same reason. The price moves when the work is done, not when a clock runs out.
Thirty Days, And You Don't Have To Explain Yourself
Read it, run it, and if it's not what you wanted, reply to any email I've ever sent you and say so.
No form. No questions about what you tried. No exit interview.
Thirty days is enough to do Module One and a weekend, which is enough to know whether the rest is for you.
Questions You'd Be Right To Ask
Do I need to know how to code?
No. There's no code in this course. One module touches a code host, and it's written for people who have never seen one, which is the only reason that module exists.
Do I have to stop using ChatGPT, Claude, or whatever I already pay for?
No, and the course specifically tells you not to. The heavy thinking, the long writing, the deep analysis, all of that belongs in the tool you already like. What this adds is the part those tools can't do, which is going and doing it.
Do I need anything else to follow along?
Yes, and you can try it before you pay for anything. Grok Bot access comes with a paid Cursor plan, which starts at $20 a month, or with a SuperGrok plan, and it isn't included in either price on this page. There's also a free trial. It's a usage credit rather than a set number of days, with a seven day window on top, so how far it goes depends on how much work you give it.
Lesson 1.1 shows you how to check what your own plan includes and where your usage is going before you commit to anything, because these pages change and have disagreed with each other before. Nobody can tell you what your allowance will hold, including me, because xAI doesn't publish the figure.
Will an assistant publish or message my audience without me?
No. Anything that reaches a person who's not you sits behind your approval, permanently. You decide what gets prepared, what gets reviewed, and what goes out under your name.
Is this a Substack course?
No. You pick a repeatable job from your own publishing calendar. Research, production checks, source packets, distribution prep and reader feedback matter whether you publish in a newsletter, on a site, in video, or across all of them.
What's a Co-Create session?
A monthly video call for paid subscribers. I bring a short lesson, then it opens up and we work through what people are actually building.
What if I join the annual after Bot Zero stops being the included drop?
You get that quarter's system instead, and you can buy Bot Zero on its own whenever you like.
When do the videos land?
They're being made now and there's no date on this page. Buy before they're finished and they arrive in your account free. The written course and toolkit are what you can use today.
Refunds?
Thirty days, no questions asked.
Four Reasons Not To Buy This
Setting it up is slower than doing the work. Teaching a job costs more than doing the job, and the first time through you're watching every step and rewriting the instruction. There's no version of this where that's not true, and your first week is an investment that doesn't feel like one while you're in it.
Review never goes to zero. It drops a lot, and on jobs that have earned it you stop reading every line. It never reaches nothing, and any course telling you otherwise is selling you something it can't deliver.
Nobody can tell you what your usage allowance will hold, including me, because xAI doesn't publish the figure on any plan. The course teaches you how to work inside it and how to see where the ceiling is. It can't tell you where yours is.
And if your week is mostly original judgment, this is the wrong purchase. If you publish rarely, if every piece is a different kind of job, if almost nothing you do repeats, then there's very little here for a system to take. Look at the right-hand column of your own week before you buy. If you can't name a job you repeat and could grade from a clear result, start there instead.
This Skill Isn't Going To Stop Being Useful
You're not going to spend a weekend learning a tool.
You're going to spend it learning how to hand work to something that moves faster than you can watch, which is a problem that will keep arriving, in more places, for the rest of your working life.
The team happens to be made of software. That turns out to be the least interesting thing about it.
The writing takes two hours. Give it the week instead.
Crafted with love (and AI),
Nick "The Whole Masthead" Quick
P.S. Bot Zero is $149 on its own and $249 when the video lessons land, and everybody who buys before then gets them free. It's also included with the Co-Write with AI annual at $90 a year, along with the paid weekly issue, the monthly Co-Create call, and a complete system every quarter you stay.
P.P.S. Bot Zero is the included system through the end of 2026. From the first quarter of 2027, the annual includes that quarter's drop instead. Bot Zero stays on sale on its own, so nothing disappears, but this quarter is the only one where the annual is the cheaper way to get it.
P.P.P.S. And if your week is mostly one-off judgment rather than the same jobs arriving over and over, don't buy this. That's the fourth reason above and it's the one I mean most. This is for the person whose right-hand column is longer than their left one.