She told the story about how she would walk through graveyards and notice the little dash between the dates.

She then said (paraphrasing) “our lives are summarized by that one little dash between the dates.”

I’ll never forget that.

A few weeks ago, I was walking through the Edmond, OK, cemetery north of Danforth. I noticed dozens of graves that ended within the World War I years. I tried to imagine the years of their birth and the lives they lived. I imagine how many of their lives were abruptly cut short to ensure people like me continued to live in a free country.

But damn… still? An entire life is one dash? It seems almost disrespectful to the dead to not have more to say on each stone.

Joddy

I probably shouldn’t write this while it’s all so raw.

However, I’m not sure it’s wise to wait either.

I promise to write a more thoughtful piece that does her justice.

For those who don’t know, my mom died last Thursday. She was 73.

She was also my hero growing up.

She also had many, many demons. I wrote about this before here and here (thank you, Internet Archive).

However, she was also gritty, stubborn, and determined. She managed to make the best life she could for her three boys.

She also lived a full life, impacting thousands upon thousands of people in the community. She was my mom, sure. However, she was like a mother to so many of the students that she onboarded into the school. 30 years and 3,000 kids (and their parents) later… she continued the legacy that my grandfather started at the school.

Don’t Be a Dash, Be The Asterisk

My mom wasn’t a dash.

She was an asterisk *

She was a person the sat at the intersection and intersected with everyone. She touched so many lives (of those living and dead).

I’m already getting texts and emails. The things she did for people. The way she made them feel. The way she helped out. What she meant to them. I cry with each one.

This is nothing new to me. When she retired, her wall was full of cards and posters. As her son, it was overwhelming and inspiring. Could I imagine a life where I impacted that many people?! It would be a tall order. I’m still nowhere close, but I’m trying.

Time to Heal

There’s more to come, but just know that I may go dark for a little bit. I’ve been expecting this since her first suicide hospitalization a decade ago. Still, despite already grieving for so long, it still hurts… a lot. The more the love, the more the grief. It’s beautifully painful, and now I need to process that for a bit.

Much love and appreciation to you all.

If you can, hug a loved one. Call your mom (and dad) if they are still alive.

And just once… walk around the graveyard… and refuse to accept being a single fucking dash. Be an asterisk.

RIP, Mom

Quick context before the post: I get a fair number of inbound questions from founders and tech leaders (DMs, emails, over coffee, etc). A lot of them have answers that are worth sharing to others.

So I’m starting a new section of this newsletter called (maybe called Inbound?). When a question lands that I think other operators are quietly wrestling with, I’ll sanitize it (no names, no identifying details, no industry details that could give someone away) and post my answer here.

If you want to opt in, great! They’ll come through this same feed. If you’d prefer to receive only my regular weekly content, I’ll send instructions next time on how to manage your settings.

This is the first one. I hope you like it.

The Background

A founder is facing VERY expensive costs from Anthropic while undergoing an SOC 2 audit. I noted that open-weight AI models like Deepseek are showing a similar performance, but at 25x less cost per token.

Unfortunately, it’s a Chinese model hosted on Chinese servers, which led to the question…

The Question

Paraphrased and Sanitized…

“How do you feel about hooking your code to a Chinese-hosted model? I’m going through SOC 2 compliance and multiple enterprise audits, and this scares me.”

The Short Answer

DO NOT USE HOSTED SERVICES FROM CHINA!

Sending your data to China is always a no-no IMHO.

The Real Answer

It’s a great question, and the fear is real. However, there are multiple issues at play. Let’s break it down.

1. The model itself: its weights, its training data, its origin

2. The harness: the IDE plugin or agent that actually runs it

3. The context: what data your tooling is allowed to send

4. Where inference runs: your machine, a Western cloud, or a server in China

Let’s take them one at a time.

Layer 1: The model

When people say “DeepSeek is a Chinese model,” they mean the weights were trained in China. But weights are static artifacts. They’re matrices of numbers. They sit on a disk like a PDF. They can’t phone home. They can’t steal anything (at least directly). You can hash them, audit them, run them air-gapped on a laptop in a Faraday cage if you really want to.

Auditors don’t care that DeepSeek was trained in Hangzhou any more than they care that Llama was trained on Meta’s TPUs. SOC 2 focuses on where customer data flows when it’s used, not on where the model came from.

There are real concerns about Chinese-origin models (e.g. training bias, censorship of certain outputs, license terms, etc), but those are output quality and IP questions, not SOC 2 questions. Different column on the risk matrix.

Now, Chinese-origin models could be trained to secretly inject insecure code in discrete ways. Here is where you might want to be careful and include code reviews by other models.

Concern level for compliance: ~zero.

Concern level for security: ~medium.

Layer 2: The harness

This is where most of the fear should actually be redirected.

The harness is the tool that runs the model and bridges it to your environment (Cursor, Claude Code, etc.). It’s the thing that reads your filesystem, indexes your repo, decides what to forward to the model, and (in agentic modes) potentially executes shell commands on your behalf.

For security purposes, this is your real attack surface. Things to consider

• What permissions does it have on your machine?

• Which 3rd-party MCP servers can it write to?

• How are your team members utilizing it?

• Is this an open-source harness with lots of eyeballs on it?

Given the harness is where the model connects to your machine and performs actions, this is where a permissive configuration OR usage policy can expose risk… particularly if the Chinese model has been trained to inject insecure code.

Concern level for security and compliance: high.

Layer 3: The context

Here’s the underrated one. What data is your tooling allowed to send?

If your IDE plugin sends only the file you’re editing, the risk is contained. If it sends your full repo, your .env file, and your customer data sitting in test fixtures, then that’s the leak vector, regardless of where the model runs or who trained it.

The model could be Gemma 4 hosted by your sweet old grandmother, and you’d still fail audit if you piped customer PII into it without governance.

What auditors want to see:

• Deny lists configured (.cursorignore, .aiderignore, Claude Code settings.local.json)

• Repo segmentation: customer data doesn’t live in the same repo as code your AI tools touch

• Prompt-level redaction where appropriate.

• Audit logs of what got sent.

This is the layer most people don’t think about until an auditor asks. Get it right, and a lot of the model-of-the-week anxiety goes away, because the risk is bounded by what you let leave the boundary in the first place.

Concern level for compliance: high. This is where the audit lives.

Concern level for security: high. This is why you want the right model for the right context

Layer 4: Where inference runs

Now we’re at the layer where the original fear was actually justified, but only for one specific topology.

Three sub-cases:

DeepSeek’s API hosted in China: (e.g., calling deepseek.com directly). Data leaves your boundary and lands on servers under PRC jurisdiction. Chinese data laws require companies to provide data to the government on request. For SOC 2, this is a hard problem; it’s a subprocessor disclosure issue, and most enterprise customers will reject it in their vendor questionnaires. This is what should scare you.

DeepSeek’s weights running on a Western host: AWS Bedrock, Together.ai, Fireworks, Groq, etc. Data leaves your boundary, but goes to a Western SOC 2’d vendor under their controls and DPA. Same posture as using Llama via Bedrock. The model came from China; the inference is governed by a Western provider. Tractable. Routine vendor risk management.

Local inference: the weights run on your own GPU via Ollama, vLLM, llama.cpp. Data never leaves your boundary. The model’s origin becomes irrelevant. Cleanest answer for compliance. Not always the cleanest answer for cost or capability, but compliance-wise, hard to beat.

The principle: a model’s “nationality” only matters when the model is also the host. Decouple the model from the inference provider, and the China question stops being load-bearing.

So what would I actually do?

If you asked me, here’s where I’d land:

Ideally, use non-Chinese open-weight models (Gemma, Mistril)

HOWEVER, if you must use Deepseek, then…

1. Hosting: Use it locally or on AWS Bedrock.

2. Harness: Use best-in-class harnesses like Claude Code or OpenCode

3. Context: Don’t use it everywhere. Use it strategically for certain use cases and flows.

4. Check-it: Always have another model run QA and security reviews on generated code.

The fear in the original question is real, but it’s mis-targeted. The right fear isn’t “Chinese model in IDE”. It’s “ungoverned context flow to any model, anywhere, hosted by anyone.”

Fix that, and the model-of-the-week question stops mattering. You stop relitigating compliance every time a new provider ships a frontier model. You’re not auditing geopolitics; you’re auditing your own data boundary and your vendors’ controls, which is what SOC 2 was always about.

The Reply

I already gave a long answer above. In the future, I may share a sanitized version of the response that I emailed the founder.

If this hit something for you

This is the kind of answer the Inbound section is built for… questions that come up privately, but where the answer benefits a wider room than just one DM thread. If you’ve got something on your mind, reply or send it over. I’ll sanitize and answer here if I think other people are quietly working through the same thing.

Until next question…

• Many people have mile-long TODO lists.

• This results in analysis paralysis, procrastination

• Without a plan, prioritization is difficult.

• Product manager frameworks have already solved this.

• Here’s one I use for myself and teams I coach.

Quickstart

Scoring System

Process:

1. Create a spreadsheet with each TODO item.

2. Rate each item over each dimension (go fast, don’t overthink)

3. Add up the total score.

4. Sort by highest score

5. Pick 1-2 items in the top 5 and go.

Alternative Process:

1. Send the prioritization skill to your AI chatbot.

2. Work through an existing TODO list interactively.

3. Review the ranked results.

4. Pick 1-3 items in the top 5 and go.

When Things Show Up In 3’s

A teacher of mine told me to pay special attention to things that show up in your life in 3’s.

In the last month, how to prioritize has come up 8 times with the following archetypes:

• 2 Startup founders

• 2 SMB CEOs

• 2 coaching clients

• 2 CTO colleagues

So clearly, people are struggling to figure out where to spend their time and energy.

Hell, I’ve had to re-run these exercises for myself on any given day! I have 30+ TODOs and 8-10 hours in my workday. It’s not going to all fit, so I have to be aggressive about sorting my daily targets to make sure the most important things get done.

Sounds Nice, But How?

The fix is simple, but don’t let its simplicity fool you.

Most people (including me) struggle to mentally sort all of this on the fly across multiple issues. By getting it out of your head and quantifying, you can help yourself zero in on the items with the biggest impact.

So get started by doing the following:

• Get a spreadsheet.

• Column 1: TODO label

• Column 2: What is done?

• Columns 3-7: Sorting across the following dimensions.

Effort is inverted on purpose. Cheaper is better, so 5 is cheap, and 1 is expensive. Impact, Reach, and Recurring follow intuition; the higher is better.

You’ll know two key things about scoring.

• Integers only. No 3.5s. The moment you split the difference, you’ve reintroduced the illusion of precision that the coarseness was designed to strip away. You can’t hide behind 4.3 versus 4.4. You have to decide: is this a 4 or a 5? That forced choice is the point.

• Orders of Magnitude. Points matter because they represent big jumps. For something to go from a 4 to a 5 on effort, you have to 10x the speed. For something to go from 4 to 5 on recurring, you have to 10x the frequency you benefit from it.

This makes 5 very hard, and 4s the highly desirable floor.

You can tolerate 3s, but only if the other categories are 4s and 5s.

1s and 2s should be avoided whenever possible. Or you should redefine the task to get a better score. If something of incredible value (5) takes a quarter (1), perhaps you find a way to get great impact (4) but in a week (3). A score of 7 beats a 6, so you should try to find those optimizations.

Don’t Let the Math and Numbers Scare You

This sounds like a lot of work, but it isn’t.

I’ve worked with teams of >10 people, and we’ve been able to rip through and score >20 items in less than an hour.

That’s way better than running with scissors and wasting days or weeks of time on a low-value item.

And if you want, you can use AI to do the scoring for you. Just have it interview on each item, and it’ll assess and spit out the scores.

An AI-era footnote

One axis has been quietly shifting on everyone: Effort.

What was a “1 week build” (score 3) in 2023 is often a “1 day build” (score 4) in 2026. AI tooling has compressed the Effort axis hard, and your rankings will systematically undervalue small, fast, high-leverage work until you recalibrate the legend.

The other three axes are stable. Only Effort drifts. Recalibrate quarterly.

Try It as a Claude Code Skill

I built this as an interactive Claude Code skill. It picks dimensions with you, calibrates the legend against your domain, scores each item, and produces a ranked table you can paste straight into a doc.

Install once:

/plugin marketplace add rickmanelius/skills

/plugin install save-your-startup

Then run it on whatever’s on your desk:

/sys:prioritize 8 features for our Q3 roadmap

It works equally well on startup features, AI experiments, side projects, or life decisions. Pick the hard list you’re avoiding right now and run it through.

I loved working through hard problems with ambitious founders trying to get their startups to the next level.

Unfortunately, life got busy, and I had to start turning down opportunities to join various programs. This was one of the inspirations behind writing my book, Save Your Startup. I wanted to take the best advice and war stories that I had shared over the years and turn them into actionable playbooks.

It’s been super helpful to me. I can now tell people to check out playbook X in chapter Y because it’s exactly what they need for their current problem/opportunity.

Unfortunately, that’s still a lot of work for the founder. And nothing is as powerful as a dialogue with the author.

Enter, AI Skills

I’ve now converted some of the top playbooks from Save Your Startup into Claude Code Skills (available here).

What the heck does that mean?

It means the exercises in my book are now interactive.

Before: you read the section and follow the steps.

Now: you trigger the skill and enter an interactive dialogue (powered by AI).

It’s quite profound. I’ve been using Tim Ferriss’ “Fear Setting” and Brene Brown’s “Painting Done” exercises for years, but it’s sometimes hard to do them without an external sparring partner. Now Claude adopts a clone of me (in my style, tone, etc) and coaches me through it.

This skill pack can do the same for you.

And if you’re an author with a book, you can convert it into interactive content too.

Laurel, my 5-year-old, fits this profile perfectly.

• Why do you need to hold both hands on a steering wheel?

• Why would a policeman pull you over?

• Why do adults pay taxes?

Ok ok. The last one I’m actually quite proud of... my little future libertarian!

But after a while, it’s too much. I wanted a solution that would allow her to endlessly pursue her curiosities with an endlessly patient partner.

And I had one non-negotiable: NO SCREENS.

Enter: Curiosity Claw.

I’ve been using an AI app called Monologue for some time. You hold a button, speak into it, and it transcribes from audio to text. It’s awesome.

What I wanted was something that could dialogue. Hold a button, ask the question, and have the AI respond back (with audio).

Fortunately, OpenClaw is a capable AI platform where I could create a simple prompt, and it built this for me in under 10 minutes. I was able to select a teacher’s voice and tune the answers to be age-appropriate (5-7 years old).

Laurel loved it! She submitted 5 questions in her first session and kept asking follow-up questions. Now she wants me to bring it home so she can use it in the playroom.

Work-in-progress

I may decide to spin this up into an OpenClaw skill that other parents can download, but for now, it’s a fun experiment in using AI for learning, education, and exploration.

And who knows... maybe you’ll be able to buy one on Amazon in the near future!

He would give you the right answer, but it would sting.

Kevin had no formal education (didn’t even graduate high school), yet he had a reputation for solving some of the most complex, gnarly software challenges. It was why he was always bubbling to the top of any organization he joined.

What was his secret? Many things. Tenacity. Curiosity. A willingness to ask dumb questions.

And reading.

I know you’re thinking to yourself. Reading? Isn’t that something everyone learns in elementary school?

Yes, but adults are lazy.

I’m lazy.

Hell, I used to read books from cover to cover. Now? If I see that an article has a 5-minute read time, I bookmark it rather than start it.

I’m proud of you for making it this far!

AI Amplifies The Problem.

We already live in information overload.

AI just makes it 2x to 10x worse.

Now people are generating 5,000-word essays in seconds, and readers are using AI to compress them back down to a few bullet points.

We skim just to survive.

Actual, word-by-word reading is a lost art.

And it’s more important than ever.

Kevin, Meet Steve Yegge

On the Pragmatic Engineer podcast, Steve said exactly what Kevin harped on 10 years ago.

…something that’s contentious, but it’s just the reality of the world. Most people can’t read. I’ve ruined much of my work in my life. I’ve just completely gone down the wrong path by overestimating people’s ability to read. And I think that reading is, if anything, getting harder to come by as a skill these days.
And this is the situation that we’re in right now is that Claude Code makes you read a lot. So I think we’re in a weird limbo for the rest of this year. Until the UIs arrive that are good enough for everybody who can’t read. Everybody who can’t read is going to be a severe disadvantage.”

So there you have it.

The best AI outputs come from well-crafted, accurate prompts and requests.

Unfortunately, most people in AI are trying to move so fast that they don’t even attempt to read.

They just YOLO and respond “Looks good! Do it!”

I mean, I know I have.

Slowing Down to Speed Up

The problem is obvious, but few people will actually fix it.

Therein lies the opportunity to stand out.

If you’re one of the few who actually read and think deeply about what AI outputs, you can learn, refine, and improve much faster than those who just accept and move on.

Learning to read slowly and thoroughly may be the new superpower in an era where most people can’t be bothered.

I’m going to lean into it.

• I played with Apple IIe’s in elementary school.

• I hacked my Packard Bell Pentium 60Mhz in middle school.

• I worked at a computer repairs store in high school.

• I went to MIT twice (undergrad and grad).

• I’ve been building tech startups for the past 18 years.

And yet I’ve seen and felt the pull that many others are feeling... the return to analog.

Here in Oklahoma, a few family friends have returned to the 90s. They ditched their smartphones for a single landline. They have a single, shared family computer. They’ve removed all TVs from their homes. Their lives are almost entirely screen-free, and they don’t appear to have any regrets.

I feel the pull, too.

During my work day, I’m YOLO’ing into the bleeding edge of AI, building multi-agent systems to scaffold software that would have taken me days/weeks before. It’s surreal and addicting.

After hours? I’m with my kid on a bike, cruising around the neighborhood, helping her find friends to play with outside.

My work is digital. My life is analog.

I hear rumors of people leaning hard into this concept. It’s almost like a version 2 of Amish living. But instead of living like it’s the 1700s, they want to build communities that feel like the 1970s, 80s, or 90s. A time-warp, if you will, where you can still visit the world we live in today, but your new normal is all-in on analog.

I’m going to do a little more digging and research on these topics and share if I find anything of interest. I’m intrigued, but it feels infeasible (maybe?). But part of my longing for this type of simplicity, at least after hours.

“Violent Self-discipline.”

Not a wish. Not a hope. But an unwavering attack on keeping your commitments at all costs. No excuses, and no tolerances for even one missed day.

It’s a high-bar, but it reminds me of Jack Canfield’s policy: 100% is a breeze, 99% is a bitch. The second you add even a little wiggle room, it’s a decision every time. If it’s a non-negotiable, zero mental energy is expended because the decision is already.

415 Days of Discipline

In 2024, I shared how I joined a 101 day boot camp called Heroic Elite. The premise was simple. Fill out a daily Carpe Diem page that defines everything from your big goals to your daily targets. Review your results at the middle and end of the day. Never skip a day.

At the end of 101 days, I decided… why stop? The entire point was to build momentum and the daily discipline was helping me stay accountable. So I kept going.

I’m now 415 days in.

I only missed two days. The first missed day was due to illness. The second was because I was so busy I jumped into a day of fire fighting and just never got back to it.

But I never let myself miss two in a row, and I lit a fire under my ass to make it almost impossible to miss another one.

Was I perfect in hitting my daily targets? Hell no! There were some days where work and family life was so chaotic that I only hit 1-2 out of 6-9 targets. That’s hard feedback to stare at, and it only provided more fuel to double down the next day.

To a lot of people, this level of focus seems over the top. However, I’ve had phases in my life where years blinked by without much to show for it, and that led to lots of regret.

Life is short, and I want to fit a lot of life into it. And the only way to make that happen is violent self-discipline.

• Eat nutrient-dense food

• Exercise regularly

• Get consistent sleep

• Limit junk food

It’s not rocket science (and we all know that).

And yet, people will place all kinds of unnecessary dependencies upon themselves before they get started.

• Endless research

• Obsessing about optimal plans

• Buying the right clothes and equipment

• Setting goals and affirmations

Let me be clear! These are helpful in supplementing your results, sustaining motivation, and holding yourself accountable. However, I’ll watch people spend 10-20 hours preparing, and then spend 10 minutes on a treadmill.

They know what to do, but they pretend they don’t. They pretend they need a perfect plan before taking action.

And it’s not just exercise.

The AI Learning Hampster Wheel

If you’re on LinkedIn or X, you’re probably getting buried with AI tutorials, demos, courses, and the like. The FOMO is intense, and some of the smartest people I know are crying uncle because they can’t keep up.

They are also missing the point.

AI is meant to help you get things done. It’s not meant to be an all-consuming 2nd job to keep up with.

My belief? Most people can stop trying to learn new AI tools and techniques. Why? Because they aren’t even applying what they know!

Learning as Procrastination

Look, I have a habit of overlearning.

I have a PhD I’ve never used, because I thought I needed it to be the best at my profession.

In reality, I could have joined a company and gone further faster by learning and applying on the job.

Many people fall into the learning trap. They convince themselves they need just a few more facts.

However, some of the best learning happens through direct experience and feedback. Your code isn’t working? Great! Now you have a real problem with real motivation to fix it.

Forgetting What You Used to Know.

I loved a story that Rich Schefren told on a recent webinar.

He was feeling stuck in his business and decided to look at his bookshelf for guidance. He then noticed a copy of his book he wrote 10+ years prior. When he opened up and started skimming, he realized that the answers he was looking for were right there.

He ended up pulling his book into ChatGPT, asking questions, and got exactly what he was looking for. He was astonished. The answers were already inside of him; it had just been so long that he felt like it was written by another person.

This happens to me all the time.

I’ll go back to my emails or to my blog to write about something, only to find out I already did 10-15 years ago. And after I re-read, I often gain new insights from my past self.

You Know What To Do

We live in a world of infinite intelligence. AI + the open web + your personal experience + people you know. The answers are all out there (or already inside you). You just need to ask, but you also need to get off your ass (I’m saying this for me too).

Stop pretending you have some insurmountable amount of knowledge you need to acquire. Yes, there are certain professions that require it (medical doctor). However, 90% of the time you already know enough to get started in the right direction… and you can make adjustments as you go.

I’m rooting for you.

I’m also rooting for myself.

Because I’m sure I’ll need to re-read this every so often as a reminder.

This is a shout-out to Jake. Beyond being a best friend of 44 years, he’s a perfect example of the archetype of this article. He was the guy who wouldn’t take no for an answer when trying to figure out what do on a Saturday night. Where is everyone? I don’t know… let’s find them. Let’s go cruising.

Mind you, this was in the glorious, tech-free 90s. You couldn’t cheat by checking social media or texting the group chat. Hell, even parents were clueless if you called to ask whether their kids were at.

So, how did you find friends? The old-school way. You drove around until you saw where the cars were in the driveway. And when you saw them, you didn’t chicken out and keep driving. You marched up to the door, knocked, and tried to get in on whatever was going down.

Old School Tactics for Generation Alpha

Kids these days have work harder to make friends.

Everything is a scheduled play date.

Parents move kids from house to house vs. letting them free-range.

It’s better in Oklahoma than in most areas. Parents are negligent, but they also aren’t helicoptering 24×7. The result is that there are safer neighborhoods where kids play in their yard and (gasp) sometimes go 2-5 houses down the road to play with other kids in their yards.

This is what I’m trying to instill in my youngest. She’s still too young for me to at least not be in eye sight for… reasons. However, every weekend I remind her that friends are where the bikes are.

So we go cruising around the neighborhood until we see a few bikes plopped in the front lawn, and then I encourage her to see if she can see or hear them. If we can’t, we knock on the door.

In the past 3-4 months, she’s made 3 new friends this way.

And there’s another pocket of kids across the neighborhood that she’s gaining more confidence with. You know, potential friends material.

Adulting

There are articles and studies on how it’s become more difficult for adults to make and sustain friendships. Add this AI bot apocalypse, and it’s just going to get worse.

But the strategy is the same.

Future friends are where the cars are. Where the bikes are. Where people are already gathering for an event or a recurring activity. You just gotta put down the phone, go cruising, and keep your eyes peeled.

And don’t be afraid to knock.

“Who’s Building This?”

Short answer: “Why not you?”

Seriously.

If you’ve been on Twitter for more than 5 years, you’ve seen this question all the time. Person X had a new idea and wanted to see if it already exists. They’d post a quick summary of the idea/wish and ask, “Who’s building it?”

Given how crazy fast AI is these days, this is the wrong question to ask.

Or better yet, it’s the wrong person to ask it to.

A day ago, a colleague asked, “All apps for X, Y, or Z suck. Do you know of anyone building one for your local machine?”

Instead of Googling, I prompted Claude Code.

30 minutes later, I had a working demo.

Was it perfect, no? But it was 80% there and personalized to me.

The issue? I speculate that we as humans still haven’t internalized this reality. We’ve seen and experienced this, and yet our reflex is still to ask and Google vs just prompt and build.

When Delegation is Slower Than Doing

Another example. A client of mine was having some communication issues on QA’ing a specific feature. The backend team wasn’t clear on what the front-end was asking for. Rather than another 1-hour meeting, I created an app in 30 minutes that pulled the backend data and displayed what was working and what wasn’t.

Instead of asking “who on the team can build this?” I just did it. It was faster to build than communicate it.

This is trend will only amplify and accelerate. Soon, it’ll be a normal behavior for everyone outside the software industry.

Who’s building a workout track app for pregnant moms? You can.

Who’s building a family vacation planning app? You can.

It’s just a matter of trying.

Prompt More.

Make it your first instinct vs your last resort.

The best part of business is manifesting an idea.

Seeing a problem, then fixing it for yourself and others.

The worst part of business is trying to herd cats…

…Claude Code has brought back my fire.

I felt that fire last Thursday night.

Why? I had been sitting on it for 2 years. It was something small, but very important to me. Finally, I had enough. After I put my youngest to bed, I fired up Claude Code with Opus 4.5 and went at it.

45 minutes later… it worked!

I felt elation! Joy!

This isn’t about the thing I built (it’s a Claude slash command called “Markdown Provenance” if you care).

No. This is about the sheer rush of empowerment running my soul because of the knowing (really, really knowing) that this could be my new normal.

Overcoming Learned Helplessness

Years ago, I wrote about the pain of “Creative Constipation.”

If you’re a builder, you know exactly what I’m talking about.

You have hundreds or thousands of ideas… but you lack the time, capacity, or expertise to manifest them into reality.

Like Andrew, I had come off a long few years of depression.

My career path started as a builder (developer, architect, product), but then later evolved into management, leadership, and eventually became a CEO at my last company.

As I moved from contributor to executive leader, I could feel my skills as a builder atrophy.

Whenever I wanted to see something built, I was increasingly reliant on others to see and manifest my vision.

Sometimes they nailed it, often they couldn’t quite get there.

And what about my side projects?

On the job, my engineers were there to help. But what about passion projects and side quests where I needed to scratch an itch? Those hit a brick wall. My skills had waned so much that even spinning small proofs of concept became frustrating.

Enter Claude Code

A lot of AI is overhyped. I get that.

But it can also do a lot of incredible things in minutes. Things that would have taken some of my best engineers days or weeks in the past.

Case in point, the Markdown Provenance tool I shipped.

I don’t know the exact numbers, but a previous version of that tool took a few days of development time to ship a few years ago. And it took a few iterations of back-and-forth to tweak some features and functionality.

However, with Claude Code, I was able to give 5-6 fast rounds of feedback, and it was all buttoned up in one 45-min session.

Joy.

I’m not kidding.

Sheer joy.

As an entrepreneur, founder, builder, etc., you are always coming up with new ideas while simultaneously living within the constraints of reality.

Claude Code has been the closest thing to having a magic wand. It’s not, but I feel more powerful and empowered every time I use it.

And this, as we are always reminded, is the worst AI will ever be. It’ll continue to get better, faster, and more powerful for years/decades to come.

Now onto the other 100 ideas that have been stuck in my psyche.

No more creative constipation. It’s time to build.

My doctor thought he was clever when he wrote the diagnosis on the wall—FOS. At first, I thought it was a word instead of an acronym, so I looked confused when I asked. “What is foss? I’ve never heard of that.” He smirked. “You are full of shit.”

He then reviewed my X-ray to show me just how bad it was. Apparently, a fraternity diet (filled with 2nd and 3rd helpings per dinner) combined with a massively stressful academic workload wasn’t a winning combination. The diagnosis seemed obvious in retrospect. However, it was tricky to identify as it happened because the pain slowly built up over 30-days and wasn’t localized to just my stomach. It radiated across most of my lower torso, and I stubbornly ignored the symptoms until I had no choice but to go to urgent care. But hey, at least my doctor got a good laugh when the fear on my face dissipated. Well, dissipated before I turned red with embarrassment that I wasted his time on something so trivial.

What’s important here is not my one-time bout with constipation (although as I get older, I’m OK with being the occasional butt of a good joke). Instead, it serves as a physical representation of a more insidious emotional and mental version, which has affected me throughout most of the stressful dumpster fire of 2020: creative constipation.

From an Inability to Shit to an Inability to Ship

I apologize for being crude, but the comparison of being unable to shit and unable to ship (e.g., turning ideas into reality) is an appropriate one.

In the book War of Art, Steven Pressfield describes the artists ongoing battle with resistance. From an outsider perspective, nothing appears to be wrong. If you took an X-ray or measured the person’s vitals, they would all check out normal. At worst, you might see a slightly elevated heart rate or blood pressure.

However, if you were able to perform an X-ray equivalent of their inner world, you would see something remarkably different. Instead of seeing a physical blockage of food being converted to fuel but not being expelled from the body, you might see another form of pile-up. You would see the person’s hopes, dreams, wishes, and ambitions all stuck and gridlocked. Visions of the girl they never asked out, the business they never launched, the city they never moved to, or the book they didn’t have the guts to write.

True, you would still see the symptoms of creative constipation that radiate throughout them. The constant tone of frustration in their voice. The frequent use of words like “someday” when discussing their ideas. How they often sigh or look down towards the ground. How you can almost feel their self-esteem depressed into themselves.

The New FOS Diagnosis

Creative constipation is a different form of full of shit (FOS). Only this time, the shit part can take many forms. Here are two that I find useful in my own life:

1. Full of self-doubt

2. Fully overloaded schedule

The first version is the obvious one. Many people (myself included) have to get over the fear of exposing their creative outlets to the world. What will people think? Will they like it? If they don’t, will they like me? The lucky few among us won’t give a shit about what others think. We are a tribal species wired for connection. The things we create open us up to the possibility of rejection. This can be scary and painful. As a result, some of us would rather metaphorically eat our feelings or simply hold back from expressing ourselves.

The second version (fully overloaded scheduled) is the one I’ve been battling for years. I once joked that I’m the victim of my own ambition. However, this is weak attempt attempts at reframing a vice to make it sound like a virtue. Also, labeling oneself as a victim is always disempowering because it places the blame on external factors rather than internal factors within one’s control.

If I was honest with you, a fully overloaded schedule is another form of self-doubt in disguise. I don’t always have enough conviction on what things I should say yes to vs. no, so I end up committing to too much. Had I had more confidence and conviction, I could probably eliminate more than enough time to make the space for my creative endeavors.

Finding Relief

2020 has been a jam-packed year. The pandemic. Moving to Oklahoma. Running a startup. Welcoming our 2nd child into the world. Switching our 1st child into homeschooling. Oh, and finding time for my marriage and myself. It’s a lot to fit into each day.

Still, these are no excuses to allow creative constipation to continue until it becomes spiritual distress. We all have responsibilities, yes. But we also need to take care of health. We need to do the things that allow us to show up as the best version of ourselves so we can meet these responsibilities.

For me, that has required doubling down on making a small amount of space each day that’s just for me. It’s hard. I can always find reasons that I can’t because of emergency X, Y, or Z. And yet, I have to do it. I know the pain of being backed up for so long as well as the relief and joy when I finally get to write that poem or article or whatever.

This article is one of many attempts to remedy the situation. After all, I’m writing this at 6 AM on a Sunday morning. However, it also can’t just be a one and done exercise. Just like you can’t build strength by going to the gym once (no matter how much you lift in any one session), the only long-term solution for creative constipation is to consistently ship. To express one’s ideas and get them out into the world, freeing up space for new ideas to come in and repeat the process.

With over 500+ blog article titles in the queue, I have a long way to go before I’m ever close to running out of ideas. Now’s the time to start chipping away at them.

What about you? Where are your ideas backed up and starting to cause you some pain? How can you get started to release the flood gates?

How’d he do it?

He vibe-coded an app with Claude Code that let him upload a credit card statement and instantly surface his top annual spends (without writing a single line of code himself).

Even better, Frank shared the code on Twitter. Others followed. They uploaded their own credit card receipts. Many shared their results, showing they, too, saved $5,000-$7,000 a year by downloading Frank’s source code from GitHub and running it themselves.

That’s the power of prompting.

AI Isn’t Plug-and-Play

Jason Lemkin of SaaStr recently shared on Lenny’s podcast that AI agents (20 of them) have replaced his 10-person business development team.

But he also made an important point: most companies can’t do this.

Why? Because they think they can just buy AI and drop it in.

In reality, it takes work. You prompt, fail, prompt again, iterate, tune, test, and repeat. Over and over.

The payoff is massive. Jason now has superpowers as a non-technical founder, managing agents in parallel, and a massive advantage over 99% of competitors who won’t get their hands dirty and can’t outsource their way into this capability.

Lesson? Prompt More.

From “Getting Off Zero” to Getting Good

I previously wrote The Prompt Gap to inspire folks to get off zero... literally just opening ChatGPT and doing something. Saying something. Experiencing one thing.

That’s table stakes now.

It’s not enough to try AI once. You need reps. Hundreds of them.

There’s a huge difference between learning about AI and knowing it through experience. It’s the difference between reading about sex and actually having sex. They are not even close.

Humans often suffer from a failure of imagination when it comes to AI. You don’t know what’s possible until you try. Until you experience that moment where something you assumed was impossible suddenly happens in seconds. Something that would’ve taken days or weeks before. Something you thought was outside your capability as a non-developer.

But you won’t get that insight until you try.

Prompt More.

My First AI Chief of Staff

Yesterday, I created my first AI Chief of Staff agent.

It’s rough. It’s rusty. But the idea is simple: I have more ideas than I can possibly execute in a day. That’s stressful. Pre-AI... I had lists upon lists of ideas I wanted to try, but never had the time.

This time is different.

I want to throw things over the wall... and now I can!

The agent captures ideas, processes them, clarifies them, does research, comes back with options—and sometimes even executes part of the work.

That alone is a pressure release valve.

Glamping (and Killing the Friction)

At a recent family dinner, friends mentioned they’d been camping. My girls have never gone, and they got excited about the idea.

We’ve been talking about new and novel experiences, so we thought, why not?

But who has time to plan an entire trip?

Claude did. It generated full itineraries in nearby states (costs, timelines, options). Suddenly, I had a clear picture of what was possible. Before, I probably wouldn’t have started because the research overhead felt too high.

Now I can do it while walking the dog, send it off, come back, and keep the idea alive.

Now I’m inspired to actually pull it off.

This Article (and Friction Removal)

No, this article wasn’t written by AI, but AI helped a lot.

I struggle to find uninterrupted time to sit down and write. Today I used Monologue to dictate while on my morning walk. It captured long-winded thoughts (I talk a lot), but made the hard part (starting) trivial.

Once captured, consolidating, refining, editing, and posting become easy.

Just like Frank sharing his app and others saving money, my hope is that by posting this, I encourage you to prompt more.

Reps > Results

@zmanian (a colleague I respect) has a simple goal: Prompt More.

And he takes it to a beautiful extreme.

“I did a month of work in the bathtub right now”

He’s constantly trying to push and break Claude Code.

He’s constantly trying to max out his token usage.

Does the code always matter? Maybe not. Some of it may suck. Some of it may be useless.

But he’s getting reps. He’s having aha moments. Eureka moments... literally in the bathtub like the original Eureka!

And he’s not alone... I’m doom-scrolling less, prompting more, everywhere.

It’s addictive, but in an empowering way.

10,000 Prompts

Malcolm Gladwell coined the “10,000 hours to achieve mastery” meme.

In the age of AI, it’s 10,000 prompts.

Most will be mediocre. Thousands will be bad. But over time, you’ll develop intuition, reflexes, taste, judgment, and leverage. You’ll become faster, sharper, and harder to replace.

I’ve written about the 1% rule for AI:

• 1% will build something extraordinary

• 9% will iterate on what others share.

• 90% will watch from the sidelines (just reading).

Because AI is such a force multiplier, that 90% will be left behind.

Want to be employable? Impactful?

Prompt More.

A Prompt A Day

Frank didn’t just strike gold on day one.

He’s been playing a game. Every day, he tries to find something that Claude’s Opus 4.5 model can’t do.

The money-saving app was on day 24.

You can have your own luck, but you gotta start somewhere.

Make it a ritual. Make it a game. Make it a challenge.

“There’s no way AI can do that!”

Well, try! Test. Experiment.

Prove me wrong (but try at least a dozen reps before you send me comments).

Prompt More

Embrace the mantra.

And just do it.

You’ll thank me later.

2025 Results

Last year, I picked the following words (w/self scoring in parentheses)

1. AI (A-)

2. Micro Moments (A)

3. Phoenix (B)

In short, not bad!

On the AI side, I had the privilege of joining Every as an AI Fellow. There, I conducted over 60 customer interviews, built a dozen or so training sessions, and really pushed the boundaries of my experience writing prompts for real-world use cases.

On the personal side, it was a breakthrough year for my own AI usage. I’m now an active Cloud Code user, and I have come to embrace AI as an essential tool in my workflows. I estimate I am 2-2.5x more productive than I was a year ago. It’s hard to believe, but I can look at my time logs and prove it.

There is still plenty of room to grow here, but I feel I’m well on my way.

In terms of micro moments, I feel this is where I knocked it out of the park. My photo app now contains lots of new, core memories with the family… things that we will talk about for years to come. I also made a point of leaning into this with friends, co-workers, and the community. I’d highly recommend it to anyone who feels they are sleepwalking through life.

As for Phoenix? 2025 was a brutal year. I faced many difficult challenges head-on. The amount of stress was hard to process at times, but I kept showing up. Was I perfect? Nope. My coping skills were not always the best, but I eventually got my head above water and feel confident going into 2026.

So overall an A-

And that is a win in my book.

Now looking forward to 2026.

Interruption Judo

This one makes sense in my head, but it might be tricky to explain. Let’s see how I do.

Ever since becoming a father of two kids, life has been full of lots of interruptions. I have two very active and talkative children. As a result, there are many interruptions (sometimes a half dozen in a minute). For someone who likes to do deep work, this is challenging and frustrating.

At some point, I’ve developed a bit of learned helplessness, where I stop starting things because I know I’m going to get interrupted any minute. This leads to self-defeat and the thought, “Why bother starting?” Unfortunately, this leads to not starting at all, which is even worse.

I did a bit of a Scott Adams reframe and said, “Well, how could I change my approach to this?” Similar to Elizabeth Gilbert’s advice in her book Big Magic, I had to figure out how to make progress in small chunks vs long stretches of time.

The idea of Judo came to mind. In Judo, you take the energy of the opponent, and you convert it into your own attack. I realized I could still make progress by using AI to offload requests and have the results come back to me minutes later, maybe once the interruption is over, with the results.

Bingo!

This has the added benefit of giving me lots of practice to further increase my reps with AI on the go.

We’ll see if it works out, but I’m going to give it a try.

Kill criteria

I love Annie Duke’s book, Quit, particularly its concept of kill criteria. As a poker player, you literally need to know when to hold them, when to fold them. Unfortunately, many people hold a bad hand far too long.

The metaphor applies to all aspects of life. As someone who is highly loyal and typically joins a company, knowing that it takes at least three to four years to really build momentum and sustainability, I tend to stay in things far too long that I should, even when the warning signs show up within one to two months.

The result: too much effort in a bad bet.

The idea of kill criteria is to set up benchmarks up front that give you accountable checkpoints. That way, instead of just doubling down over and over and over again ad infinitum, you face yourself with accountability face-to-face early on.

I’ve been known to bet on people and things for far too long when I should walk away much faster. This is probably why I faced burnout in the past. I could have saved energy, time, and stress by just walking away sooner, but I didn’t.

So this is a reminder to always set criteria up for every new project or idea and be able to exit them faster, versus trying to keep too many plates spinning.

At the end of 2025, I had over 160 projects active. Most of these were ideas and hobbies I started and never stopped. The result was just too much stress, too many things to track. I eventually pared the list down to about 75 projects, most of them work-related. Even that’s large, but it’s much more manageable. A lot of those open loops have now been killed.

10x > 2x

This is similar to the title of a book by Benjamin Hardy. And it’s an important mindset shift for someone who likes to muscle his way through problems.

Most of us try to do more by just throwing more hours and effort at a problem. To have a true transformational moment, a 10X requires a new strategy, a new way of thinking, and a new approach.

So this is a reminder to me not to just keep throwing hours and effort at things, but to reevaluate.

Whereas Kill Criteria asks me if I’m playing the right game, 10x over 2x asks me, am I playing the game right?

Onward

I don’t suggest that you follow my words for this year. They are very personal to me, reflecting my usual blind spots and current goals. However, rather than New Year’s resolutions, I highly suggest the three-word paradigm because it gives you very easy-to-remember high-level targets instead of trying to remember dozens of wordy goals.

I don’t always do them perfectly, but the daily reminder keeps me focused on making forward progress.

Simple! Men sometimes do stupid things to prove stupid points.

Ok ok. The real reason?

A colleague of mine posted how he did 5,000 pullups in one month, and my jealousy got the best of me (yes, it wasn’t my finest moment).

My internal monologue went something like this:

Before we go any further, I want to share a snippet from a famous article by Dan Sivers: Relax for the Same Result.

Like me, Dan used the Boston Esplanade as a cycle track to prove how fast he could go. One day, he biked it slowly... only to make the shocking discovery.

This is a key insight when it comes to physical exercise: the amount of perceived effort doesn’t linearly translate to achieved output.

In fact, sometimes it can lead to the exact opposite outcome.

Before trying to do 5,000 pushups in a month, I was consistently able to do 40-45 pushups in a row (my all-time best at 51). I tricked my brain into thinking... if I do 5,000/month, I should be able to at least get to 70 or 80! I’ll be soooo close to my goal!

Spoiler alert. It didn’t work.

In fact, I got WORSE.

After celebrating my accomplishment, I took 3 days off for the muscle soreness and tightness. Then I made my first attempt to achieve a new pushup PR.

Andnnd failed.

36 pushups before my arms quit.

How is this even possible? Well, before you say I should know better, I will say I absolutely did. Decades of workout wisdom will tell you that strength training requires recovery periods for muscles to rebuild (and that’s where a lot of the strength gains happen).

Here’s The Real Problem

I sometimes live in that smart-enough-to-be-dangerous part of the IQ bell curve. You know the type? The person who thinks they can outwit the rules of nature just because they are good at math and science.

Well, now I know for a fact that nature didn’t make me the exception to the rule. That, in fact, sometimes (as Derek so beautifully described), we can just relax and get the same damn result (maybe even better).

This doesn’t mean don’t push yourself! The reason HIIT and anaerobic strength training work is that you push your muscles to the point of failure, then let them repair. However, that last part is the key. You have to go into it with the right plan, or you’ll undermine your results.

As a guy, I can’t promise I won’t do other stupid things to prove I’m fast or strong, but hopefully this lesson has sunk in for good.

Before you leave! Some reminders…

• Yes, the world is full of bad ideas.

• Still, someone, somewhere will try it anyway (just because they can).

• Someday, someone you know will consider seriously this option.

• I want you to be equipped with the knowledge to pushback strongly.

A final disclaimer. I can empathize with the instinct. I know people (sane, high-IQ colleagues) that have tried to create virtual avatars of their dead parents to help with the grieving process. I imagine parents grieving the loss of a child will be even more inclined to fill a hole in their heart with something that keeps them connected.

At best, this will be a stop-gap solution that requires real healing. And I want to highlight the many, many deficiencies that an AI child will never be able to overcome.

First, the X post that sparked this article.

https://x.com/zxlava/status/1991869478976770260

I’m aware this sounds totally insane rn but I think AI children will be a much bigger deal than AI girlfriends

And my flippant response https://x.com/rickmanelius/status/1992692933896851742

Why stop at AI children? Wait until miniature AI child robots exist. And they are marketed/sold to parents who lost their children due to unforeseen tragedies. (and this will happen)

I’m actually proud of my attempt using AI to create the Tesla baby. I mean, it’s kinda cute… right?

But let’s start with giving the benefit of the doubt.

The Inevitability

Fact: AI is already being used as a substitute for human therapists and friendships.

This is not surprising. ChatGPT has gotten so good that interactions feel like a human is really there.

There are apps already out there (e.g. Friend) that let people have a persistent AI friend at the touch of a button (pendant). And finally, there are people willing to go as far as leaving their marriage if forced between a choice of their spouse vs their AI companion.

So this isn’t a theoretical possibility. We’re already here.

The Obvious Problem

You can’t replace flesh and blood with an agent (even one that’s a robot with a cute outfit). But that’s just the surface. The deeper issue is how many aspects of human evolution and transformation that an AI child simply cannot address.

The Physiological Transformation It Can’t Trigger

“Mom brain” is real. A mother’s brain literally restructures during pregnancy. Gray matter changes that predict maternal attachment. Hormonal cascades of oxytocin, estrogen, and progesterone rewire neural pathways in ways no screen interaction can trigger.

“Dad brain” is real too. Testosterone drops, brain structure changes, permanent neurobiological shifts occur. These aren’t metaphors. They’re measurable, lasting alterations to who you are at a cellular level.

And physical touch with your child releases bonding chemicals that shape both of you. An LLM cannot be held at 3am. It cannot trigger the biological responses that come from skin-to-skin contact with a crying infant. Nor will a crying LLM spike a mom’s maternal instinct to rush-to-protect it.

The Psychological Rite of Passage It Can’t Replicate

I learned a few words doing the research for this article. Here’s one: Matrescence. It’s defined as the psychological transition to motherhood, and it is a developmental stage as significant as adolescence. You become a fundamentally different person, not a better-connected version of your old self.

The difficulty is the transformation of taking care of a new, helpless infant. Sleepless nights, fears, and failures force growth that comfort cannot produce. You don’t become a parent by feeling parental emotions. You become one by being broken open and rebuilt through necessity.

There’s also the permanence of the decision. An AI Robot has a 30-day return policy. Kids are not like that. You’re all in the stakes of walking away are immense. Human identity reconstruction requires these irreversible decisions. But with an AI child, you have all sorts of outs when the going gets tough.

You can turn off an AI child. You can reset it. You can walk away. You cannot un-become a parent. That permanence is what creates the stakes, and the stakes are what force the growth.

When your child’s wellbeing genuinely depends on you becoming better, you become better. Simulated stakes produce simulated growth.

The Extended Family Dynamics It Can’t Create

A new child doesn’t just change one relationship, it changes them all.

When you have a child, your parents become grandparents. Their identity shifts too. Your siblings become aunts and uncles. New roles and relationships emerge across your entire family network, not because anyone chose them, but because a new person now exists who relates to everyone differently.

Intergenerational negotiations begin. Navigating different parenting styles, managing boundaries with in-laws, mediating between generations. All of these are experiences that do not happen otherwise.

Your place in the family lineage changes. You’re no longer just someone’s child. You’re a bridge between generations, carrying responsibility in both directions.

An AI child creates none of these dynamics. Your mother doesn’t become a grandmother. Your brother doesn’t become an uncle. The relational web that real parenthood generates simply doesn’t exist.

The Community Integration It Can’t Generate

Real parenthood plugs you into community structures. You join new social networks: schools, pediatricians, parent groups. Your status in community shifts as new responsibilities emerge. Think of the number of people your parents came into contact with solely because you existed.

The “village” activates around you. Support networks form (whether intentional or not): other parents who understand, neighbors who watch out, teachers who invest in your child’s future. I hate the quip “it takes a village”, but it’s the truth. A human being’s life is shaped by everyone around them.

An AI relationship cannot access any of this. There’s no PTA meeting for digital offspring. No playdates that become adult friendships. No shared stake with other families in the neighborhood school.

The Bidirectional Transformation It Can’t Produce

Real children develop in response to your parenting. Your choices shape an actual person: their confidence, their fears, their capacity for love. This isn’t simulation. It’s consequence with direct, real-time feedback. You become different, which causes them to become different, and on and on it goes.

Children surprise you in ways that force adaptation. Just last night I was putting my 5-year old to bed, and she had a vomit explosion. It took the entire family to rally together and scramble to fix the situation. Kids are full of daily surprises (big and small) that force you to think on your feet and play the hand your dealt.

My Closing Arguments

Many will still be tempted to try and build and use AI-children.

They should not be judged or faulted, but if they go into it without understanding these limitations, they are setting themselves up with false hope.

The emotional response may be real. Comfort has value, and I won’t tell anyone their coping mechanism is wrong.

But for those imagining AI children as a substitute path to parenthood’s rewards (e.g. the growth, the meaning, the place in something larger) will miss almost everything that makes parenthood transformative.

Parenthood is just a 1×1 experience to serve our emotional needs. It’s a multi-dimensional transformation that operates through embodied experience, hormonal changes, relational webs, and irreversible stakes. In short, it’s hard, but it’s supposed to be that way.

An AI child might offer comfort. It cannot offer transformation. And transformation is what parenthood actually is.

PS

I’ve been sitting on this article for two weeks and was about to give up on it. It lacks my normal amount of time and attention to get the spelling, grammar, and editing improved. And I chose to rely on Claude to get 2-3 of the sections written because I was having writers block. Future articles will be less sloppy, but I felt this one was important enough to ship in a timely fashion. Thank you.

I was suffering from the curse of knowledge.

Having been in this space for so long, I’ve normalized away the magic. When ChatGPT 3.5 first came out and people were asking to do silly things like create Eminem rap songs in Old English, I was enthralled. AI finally hit an inflection point from theory to reality and my worldview flipped in an instant.

Then, I took it for granted.

As an AI trainer, I now focus on edge and corner cases. The magic moments still happen, but I expect them. It’s like watching a magician but knowing the trick. It kills the magic.

Still, every now and then I get to watch people experience their AI eureka moment, and it brings me right back.

Multiple people on the Zoom call shared, shaking their heads, saying they couldn’t believe they’d build something in minutes that would have taken weeks or months before.

Yes, these are just vibe-coded prototypes and not production-ready, but the empowerment these people felt to participate was a life-changing moment for many.

This summer, I remember sitting with my wife as she used ChatGPT for the first time. She resisted it until she needed it to solve a gnarly problem she was facing as a school teacher: how do I modify a set curriculum to a grade-appropriate difficulty from pre-K to 3rd grade?

This was a major undertaking, but as ChatGPT generated a comprehensive plan in under a minute, complete with changes, justifications, and cliff notes, her eyes widened.

Ultimately, she ended up going a different direction, but the experience changed her perception of AI’s potential.

I live for these moments when running trainings at Every. Last week, we had someone unlock a major pain point costing team members hours daily organizing schedules. Seeing their smiles and joy was soul-filling.

I hope we never lose the sense of wonder and magic about how AI can empower us to do more with less.

Yes, a non-trivial number of parents are abusing the system by purchasing nothing but junk food. And while I would love to argue that empty calories are better than no calories, historical evidence from post-WWII Japan shows a marked rise in average heights when food supplies and quality were restored.

Fortunately, this shutdown will eventually end because of increased pressure on lawmakers to stop playing this stupid game of chicken. However, issues of child hunger will persist.

So what does this have to do with school lunch debt?

I never heard of such a thing until my colleague Tyrone Ross started posting about it on Twitter.

This particular post was gut-wrenching because one child was being threatened with being taken and placed into foster care if his parents couldn’t clear the debt.

https://x.com/TR401/status/1154139983529172992

Free and Reduced Lunch

When I was in school, kids that couldn’t afford lunch were placed in the “free lunch” category. My brothers and I were frequently in this category. While it was embarrassing to talk about this with friends in line, at least we didn’t have to skip lunch.

Apparently, this isn’t universally true. There are some places where kids must pay money for each meal. Sounds reasonable, but what if parents won’t (or can’t) send money with their kids? Well, they go into debt.

Kids don’t understand that debt is bad. They need to eat, and the school just runs up the tab.

Eventually, this tab gets to $10 or $100 and a threatening letter goes to the parents.

Unfortunately, this isn’t uncommon. Something like 75% of schools have some form of school debt, with some totals being as high as $20,000 over hundreds of kids.

This is insane, but not surprising.

My Mom

Despite being on free lunch, it turns out that my mom was able to help other kids that didn’t qualify, but still showed up to school empty handed.

Everyone knew that my mom had a big heart. So it turns out, kids that weren’t able to get the lunch lady to budge, they would ask her and she’d help them out.

Tyrone Ross does something similar. In the past, he has stopped by local schools and paid off the school debt for multiple students.

I think this is why I felt so inspired to donate heavily to local food banks during the COVID pandemic, and why I try to be as generous as possible to them in the winter months where being cold and hungry is the worst.

While SNAP pausing will be very disruptive, food insecurity will continue to be a major issue for many. If you feel so inspired, donate anything you can to a local food bank or call a school and clear some lunch debt for a kid in need.

Getting to the punchline, it goes like this.

If peers perceive you as smart, senior, or experienced, your use of AI is perceived to be a good thing! You are being proactive by outsourcing administrative and other lower order work so you can work on higher order, strategic items. Good for you!

If you’re perceived as junior or inexperienced, your use of AI is perceived by a bad thing! Clearly you can’t hang with people with credentials, certifications, and tenure, so you’re using AI to shore up your deficiencies. You’re a poser using AI to fake who you are not. Shame on you!

I’m kinda/sortof embellishing a bit, but not by much. We see AI as either a way for smart people to get more leverage as well as a way for ignorant people to fake it.

This is an important cognitive bias to be aware of, because it can amplify mistrust to people based on age, sex, role, experience, etc. We should all be using AI to amplify our abilities while not holding it against anybody.

At the very least, we should be aware how we’re not applying our perceptions evenly, and that can lead to unfair biases.