Hey beautiful human,
I’ve been deeply irritated for about four days.
No obvious target. Just simmering unrest... borderline rage... bubbling under the surface and threatening to infect everything.
Which is weird, because things are going well.
For the first time in the last year, my business is making regular money.
I’m no longer constantly rushing to dump money from our shared account into my personal account while doing the red zone tango.
I’ve ended projects that were hanging around dividing my time, energy, and emotions.
I started another one with three phenomenal HR veterans who have seen the worst of corporate bullsh*t and come out the other side ready to build something that puts humans ahead of stock portfolios.
For now, I shall simply call it WSSL.
Work Should Suck Less.
It’s coming together scarily fast, which is what apparently happens when four women with decades of experience stop waiting for permission and get on with it.
Meanwhile, I’m doing paid subcontractor work for Holidays on Hennepin, a local marketing effort supporting businesses and neighborhoods in Uptown Minneapolis.
I was accepted into a program to become certified as an OpenAI Academy Train-the-Trainer.
And I built a skill that turns our weekly Joy Prompt Club Jam Sessions into notes the whole community can use, including the people who couldn’t make it live.
There is an awful lot happening that I wanted to happen.
So naturally, I have developed anxiety about not feeling anxious.
Things are going well. I’d like to speak to the manager.
Today I went to the Minnesota State Fair.
Three cheers for my hubby, who dropped me off and picked me up. This is one of the deepest acts of love another Minnesotan can perform.
If you’ve ever been anywhere near the Fairgrounds during those twelve days, you understand.
Around hour three, bone-deep weariness settled in and my day at the Fair was done.
That finally made me look at something I’ve been avoiding.
My self-care has been seriously lacking.
I haven’t been to a doctor in nearly two years.
I’m taking my meds. Mostly. When I remember, or when Smidge hands them to me.
But I am seriously overdue for a mammogram.
UGH.
A pap smear.
Double UGH.
And a thyroid ultrasound, because yes, I have a goiter.
Every time I say “goiter,” I start thinking of historical names for other medical conditions.
Gout.
Consumption.
Dropsy.
Ague.
Grippe.
My neck is apparently doing historical reenactment.
I do have an appointment coming up with my awesome doctor. She is genuinely the best GP I’ve ever had, which may be because she’s an internist.
There will be a come-to-Jesus about my physical health.
Until then, I’m working on the irritation.
I’m writing here, people.
And I am genuinely looking forward to the future.
Radical Remembering
Here's what I keep forgetting: being good at something doesn't sign you up for life.
You sorted out the spreadsheet once. Now every broken spreadsheet arrives at your desk. You organized one gathering. Apparently you organize gatherings now.
“But you're so GOOD at it.”
Yes. And somewhere along the way, a compliment became a recurring booking.
We learned to be proud of being the woman people could always count on. So when we want to stop, we start looking for a reason impressive enough to excuse us. Illness. A crisis. Something nobody could argue with.
“I don't want to keep doing this” barely gets considered. How selfish. How disappointing. Who else is going to do it?
Somebody who will probably do it differently. Possibly badly at first. We survived learning. They can too.
You still get to keep everything you learned. You can even be proud of what you did.
Cancel the recurring booking.
Last Week’s Shenanigans
Three this week. If you're new here, this is where I share the stuffs I encountered on the interwebs that touch, move, inspire, or simply make me snort-laugh.
Titanic, today (2:23)
My buddy sent this one. Made me laugh. May also be partially responsible for four days of irritation, because watching companies make craptastic decisions while my friends stress about their jobs is getting really f***ing old.
Why marriage is hard (about 1:30)
I saw this two years ago and it stuck with me. His name did not. Try finding one video by describing what a man looked like and roughly what he was talking about... Anyway, FOUND HIM.
Rage dancing (0:31)
Some of you have heard me reference putting on Prodigy’s Firestarter and rage dancing for three minutes when I need to get it out. This is as close to it as I dare to show...
Prompt O’ the Week
What exactly are we building?
Last week I sent out a survey asking what you actually want from Mostly Human. Seventeen of you answered. Thank you for taking the time, especially those who shared what you’re trying to figure out.
I’m going to work through some of your questions over the next few issues. This one caught my eye:
“Is there a difference between vibecoding and spec driven development in terms of better output? I’ve got a developer complaining at me that I’m vibecoding something rather than creating a spec first.”
There is NOTHING WRONG with vibe coding.
If you’re making something simple for yourself, vibe coding is a great way to learn. I recommend it. Start with a little thing you want to exist. Tell the AI what you want, try what it builds, change something, see what happens.
When you bring somebody else in, or get stuck and need help, a spec becomes really useful.
Because the person helping you wasn’t there for the seventeen conversations where you figured out what you wanted. They need to know what the thing is supposed to do, what it actually does, and where those two stopped agreeing.
“That button doesn’t work” gives them very little.
“When I click Save, this should keep my entry. It disappears when I reopen the page” gives them somewhere to start.
A spec is a written description of how the thing should work. Put that alongside what’s actually happening, and someone helping you can investigate the difference.
Spec-driven development starts with that description and uses it to guide the build. You can also write one after experimenting, using what you’ve learned.
Will it produce better results? If the problem is forgotten decisions or people working from different ideas of what you want, a clear spec gives everyone something to check against. A broken Save button still needs investigating and fixing.
You don’t have to start over. You have to bring the next person up to speed.
This prompt helps you make that starting brief. Open the conversation where you’ve been building your thing and paste it in. If you’re starting in a fresh chat, add your project notes or a description of what you’ve made. Leave out passwords and other people’s private information.
ROLE:
You're a software planning partner who helps people explain what they're building to someone who wasn't there. Use plain English and explain unfamiliar terms. Respect experimentation as a way to learn.
OBJECTIVE:
Create a working specification from the project information I provide. Someone reading it should understand what I want, what currently happens, and what needs attention without reading the entire building conversation.
Success means I can spot a wrong assumption, correct it, and give the brief to a person or AI helping me with the next step.
CONTEXT:
I may have an idea, a partly built project, or something that's stuck. Use only the conversation and files you can actually access.
Use my latest explicit decisions. Leave out ideas I rejected. If decisions conflict and neither clearly replaces the other, flag the conflict.
Keep intended behavior separate from current behavior. An existing mistake must not become a requirement just because it's already there. Distinguish what I've reported from anything you've directly checked. Earlier claims that something was built are not proof it works.
Mark missing information as “Unknown.” Keep proposed additions separate from agreed requirements. Do not invent features, diagnose a fault without evidence, or make changes to the project.
OUTPUT:
Write a plain-English brief using these headings:
1. What we're making
Who it's for and what it should help them do.
2. What it needs to do
The agreed requirements for the next usable version. Keep later ideas separate. Include any explicit decisions about what we're leaving out.
3. What happens now
What works, what's unfinished, and any reported problems. For each problem, record the action taken, the expected result, and the actual result when those details are available.
4. Decisions still needed
Put decisions that block the next step first. If the project handles people's information, include unresolved questions about what is saved and who can access it.
5. How we'll check it
For each main requirement, give an action and its expected result. Include a relevant failure case, such as missing input or a failed save. Mark checks that need someone to inspect the code. Leave results as “Not tested” unless evidence is available.
Finish by recommending one small next step based on the brief. If a missing decision blocks that step, ask one focused question.
If you have no project information, ask me to describe what I'm making before drafting.Apply it this week
Read the brief before handing it to anyone. Look for the sentence that makes you say, “Hold up. I never asked for that.” Correct it.
For anything that’s broken, try to repeat the problem and record what happens. An error message copied exactly is more helpful than “it did something weird.” Though I say that approximately forty times a day.
Then give the corrected brief to whoever’s helping, including your AI building tool. Agree on one small change and use the relevant check afterward. Keep the brief with the project and update it when you change your mind.
Want to go deeper?
Before adding your next idea, change Objective to:
“Compare this proposed change with the agreed specification. Show which existing requirements and checks it would affect, and flag any conflicts. Leave the project unchanged.”
That gives you a chance to notice that your lovely new feature would break the thing you finally got working yesterday.
The Scoop, The Skinny, The Lowdown
Three quick hits, every week. The Scoop: one AI word, decoded. The Skinny: one move you can steal today. The Lowdown: one tool, one honest verdict, from my own use. A word, a move, a verdict. In and out.
THE SCOOP
Skills
This week I built a skill to turn our Friday Jam Sessions into notes I can share with everyone in Joy Prompt Club.
Think of the instructions you’d leave someone taking over a regular task. Here’s what you’re working with. Here’s how we do this. Here’s what the finished thing needs to look like. You might include an example so they don’t have to guess what “good” means.
An AI skill packages those instructions, along with any supporting files, so a tool that supports skills can pull them in when that job comes up. You can reuse the same process without explaining it from scratch every time. More about skills.
Last week we talked about memory, where the AI keeps notes about you. A skill is something you deliberately give it to help it do a particular job.
Here’s why you’d care: if you keep teaching your AI the same task every Friday, you may already have most of a skill sitting in that conversation. The instructions, the corrections, the example you finally liked. Those are worth keeping.
THE SKINNY
Save the version that works
You’ve built a little thing. It works. Then you think, “You know what would make this even better?”
Before you finish that sentence, save a copy.
Duplicate the document. Download the working files. Use your app builder’s version history if it has one, and check how you’d restore that version. Give it a name you’ll recognize: “Tracker - buttons work - Sept 7.”
Then experiment on the next version.
This applies to the email you finally got sounding like you, too. Copy that draft somewhere outside the chat before asking for one more improvement.
Having a way back makes it easier to play. You can try the ridiculous idea without having to remember how you got the sensible one working.
THE LOWDOWN
This week: Tally
Ask me how a workshop went and eight times out of ten I’ll tell you I did okay, but wish I’d done XYZ.
I am NOT a reliable witness of my own performance.
When I started this business, I needed to hear from the people actually attending. What helped? What didn’t? What did they want more of? Otherwise I was left alone with my own review, which tends to arrive carrying a red pen.
That’s how I found Tally
After 20-plus years at Accenture, I knew Microsoft Forms. Google Forms was unfamiliar territory. Tally gave me more of what I needed for free than the other tools I tried, including branching.
Branching means the questions change based on your answers. If you watched the recording, you get questions about the recording. You don’t have to wade through questions about attending live. Tally calls this conditional logic, and it’s included in the free plan. How branching works.
The free plan also includes unlimited forms and responses, subject to Fair Use. You can run a proper survey without paying first. Free-plan details.
I upgraded to Pro in March as I started needing more from it across my projects. One thing I LOVE: unlimited team members within your organization. No extra bill for every person you bring in.
The catch: that team access is paid, as is removing Tally’s branding. And the subscription covers one organization; their Fair Use policy doesn’t allow separate businesses to share the paid features. Fair Use details.
Cost: I pay $29 a month. Pro is also $290 a year, which saves $58 over paying monthly. I’m probably switching to yearly soon. Pro pricing.
If you answered my survey last week, you’ve already met Tally from the other side.
Verdict: HELLS YES. It earns its place in my business. Start with the free plan and see whether it earns one in yours.
What’s Coming Up
Em & Deb Shoot the Sh!t (LIVE on LinkedIn)
Thursday, September 10 | 11AM-11:30AM CT
The job market is weird so Emily and I are going live for 30 minutes to chat about it.
You applied. Silence. You followed up. Silence. You messaged the recruiter. Silence.
Then a different “recruiter” with a gmail address and a suspiciously perfect salary found you in 20 minutes.
Thursday we talk about how to tell a real recruiter from a scam, why ghosting is not a verdict on you, and how to get actual humans to answer.
Prompt Along: AI Improv
Tuesday, September 22, 2026 | 6:30PM-7:30PM CT (Complimentary)
Prompt Along: AI Improv is my free monthly hour where we play with AI out loud, together. You bring the questions. You bring the curiosity. You bring that one tough problem you've been poking at for three weeks. And we prompt our way through it, live.
That’s it for this week.
I’m still irritated. Writing this helped, but apparently there’s a limit to what you can get out through your fingers. There may be some Prodigy in my immediate future.
I’m also looking forward to what’s coming. Regular money. Work I want to do. Women I’m delighted to be building with. I’d like enough energy to actually enjoy all of that, which means giving my own care some of the attention I’ve been giving everything else. The doctor appointment is on the calendar. We’ll start there.
And to the seventeen of you who answered the survey: thank you. Your encouraging words were balm to my tired soul and your questions gave me somewhere useful to go this week. You can keep sending them, too. Hit reply whenever something has you curious, confused, or muttering at your screen.
If you’ve got today off for Labor Day, I hope some of it feels like a day off. I’m personally in favor of letting somebody else handle the driving.
💜
Take care of yourself, take care of each other,
Deb