My observation of air travel is that if there is one significant impact on a flight, there are guaranteed two additional impacts. An example: if my gate changes at the last moment, it is quite likely that departure from the gate will be delayed, followed by a further delay on the tarmac. Unexpected changes in complex systems create further unexpected changes downstream because complex systems are, well, very complex.

A simpler and more personal way to look at this is that when you are traversing complex systems (airports, credit card websites… really any large bureaucracy), there are two paths:

The happy path. Whatever you are attempting to do (fly home, change your auto-pay settings, paying a bill) is a well-known and on-the-rails process. You can sense this happiness because as you complete each action, you receive feedback: “Yes, we know what you are trying to do. Yes, we can help. Yes, we know what to do next. Don’t worry. We got you.”

Divergence from the happy path leads to one specific path: the sad path. All that positive feedback vanishes. I’m not talking about the dark patterns of corporations trying to trick you into not canceling your account; I’m talking about a truly one-off request that is reasonable, but so infrequent that no one built a happy path around it.

Once on the sad path, you are desperately looking for one specific human: The Insider.

In Defense of Process

Hey, I know you’re mad. I’ve been there, too. You missed the connection, your airline rebooked you onto a partner flight, and now no one can see your checked bag. You started on one website, then the other, and now you’re on the phone with what sounds like a human; you are getting the runaround.

Deep breath.

“Process is documented culture” is what I like to say. Long ago, this company you currently hate was smaller, hungrier, and more nimble. This was working out for them, as evidenced by rapid growth. The problem during rapid growth is that there are so many new humans showing up that it becomes more and more expensive to train these new humans in how this company works, so someone raises their hand and volunteers, “I will write down how we work, and I will share this with the new folks.”

This is a noble person trying to help. Remember that. This person gets the OK, and they (and others) start the arduous task of documenting how the company works and quickly discover two things:

1. Doing this thoroughly is an impossible task. There are too many people, processes, and products, so they…
2. Focus on the most important processes that need documenting. They document the visible, exciting, and profitable happy paths. They likely uncovered and discovered sad paths as they did their work, but their prioritization at the time was, “This doesn’t happen a lot, so… who cares?”

These clearly defined happy paths are fortified, refined, and built into the pillars of “How We Work.” The core business revolves around these many happy paths, but as anyone who has worked in this large business knows, reality is a mess.

The Power is Off

PG&E — Pacific Gas and Electric. For reasons unknown, the power was off at the house. No text, no email. Bills were paid. Visiting the website revealed nothing helpful. A call pointed me to a robot who gleefully told me, “Everything looks good.” When I finally reached a human, they reported, “Everything looks good.”

“The power is off.”

“We’re not seeing that here.”

“The power is off.”

“Hmmmmm. Wait. This is interesting. Can I put you on hold?”

I wait ten minutes before the call is dropped.

So what happened? I diverged from the happy path. The human on the other end of the phone probably knew something was up, but didn’t know what to do next because they’d never seen this situation before. Combine that confusion with their incentive — they are measured on how quickly they can resolve calls, so my situation represents a threat: I don’t know what to do here, and the longer it takes for me to figure it out, the worse my review. So, put him on hold. And hang up. Victory!

Having spent many days of my life on the sad path, I know two things:

• Keep pushing
• Seek The Insider

I call back. Same robot hoops, then the robot reluctantly points me at another happy path human, and I do the only move I know at this point: I get mad. Yes, you catch more flies with honey than vinegar, but honey is not working at this point. I get mad.

“The power is off. It’s been off all day. This is my second time talking to a human, and it takes twenty minutes to actually get to a human. How can you help?”

Moment of truth: I either get hold-dropped again, or they hear the frustration, and they escalate.

Each time I am interacting with a new human — shoulder-to-shoulder — looking to complete a task, I quickly look to answer one question: “Do they get it?” The Insider gets it… immediately.

“How can I help you, sir?”

“The power is off at the house. It’s been 12 hours. Everyone says it’s fine, but the power is off.”

“Hmmmmm. I see. Did it go off at 8:30 am this morning?”

THERE. THERE IT IS. THIS COULD BE THE INSIDER.

Calmly, biting my tongue, slowly, “Yes, 8:30 am on the nose.”

“Got it. Let me… wait. Ooooooh. A 2319 — I haven’t seen this in a long time.”

IT’S THE INSIDER. IT’S THE INSIDER OH MY GOD OH MY GOD ALL MY PROBLEMS ARE SOLVED.

I say nothing. It’s such a rare occurrence to find The Insider. My mind is racing. How can I get their contact info? How do I send them a cake? Their favorite cake. You bet I want to ask what a 2319 is, but I must not interrupt their flow. This is The Insider doing their best work.

“Ok, so I see what happened…” And they explain, at length, completely, confidently. Yes, it’s an initial dumb error on PG&E’s part, but that’s not the point. I am on the phone with a human who deeply understands the complicated, messy, and essential systems around them. They haven’t given up and aren’t just reading the workflows on the screens in front of them; they care… they are curious.

The Process Inflection Point

Remember that human who raised their hand and said they’d write it down. It didn’t feel like a big deal in the moment. Yes, yes. That seems like a good idea. You zoomed right past a major inflection point of your company. See, until that point, much of what is important about your company existed in the heads of the folks who are running the company. They don’t need an artifact; they know.

The creation of this artifact allows for the creation of a set of employee variants that heretofore have not existed. Broadly, these are enforcers of the process, and their creation, as discussed endlessly, also creates consternation in your company because of the perceived assault on freedom. The argument: Why does anyone need to write it down? We already know.

They do. They do now. In five years, when you’ve grown 10x, they will have forgotten because they’ve moved on… either to harder problems or another company.

You know who doesn’t forget? The Insider. They’ve been here since the beginning, or they were minted during the process inflection point. In either case, they know some or all of the system. They didn’t build it, but they desire to understand it.

And they want to help.

There is one. Wait, two. No, three. Ok, four humans who directly inspired my writing. (Ok, five.)

The list is not in order of importance, but by time:

Robert Fulghum, All I Really Need to Know I Learned in Kindergarten (Fall 1988)

Think what a better world it would be if we all—the whole world—had cookies and milk about three o’clock every afternoon and then lay down with our blankies for a nap. Or if all governments had as a basic policy to always put things back where they found them and to clean up their own mess. And it is still true, no matter how old you are—when you go out into the world, it is best to hold hands and stick together.

Fulghum wrote a whole series of front-porch wisdom books, but the one that most people know is All I Really Need to Know I Learned in Kindergarten, published in 1988. I was working at Waldenbooks in the Vallco Mall, and this book literally flew off the shelves. Fulghum was a Unitarian minister, and he wrote “Kindergarten” for his church bulletin. The essay was passed around (pre-Internet) and eventually landed with a literary agent.

The essays were brief, dense, and heartfelt. He effortlessly gives names to all parts of the human condition… ideas and feelings you have all the time, but have not named.

Fulghum is hopeful.

Robert B. Parker, Looking for Rachel Wallace (Fall 1991)

I waited. I knew he was trying to size me up. That was okay, I was used to that. People didn’t know anything about hiring someone like me, and they almost always vamped around for a bit.

Looking for Rachel Wallace was published in 1980. Fall 1991, unlike other entries on this list, is when I read the book as part of a UCSC class called “The American Detective.” It’s how you know a new author or book is novel — you start reading, hours pass, you are hungry and have a headache, but you read the whole god damned book.

Parker comes from the Dashiell Hammett School of detectives. Spenser, the detective who never has a first name, is an ex-boxer from Boston, but geography is not important to these books. It’s dialogue. 80% of Parker books are gritty, intelligent dialogue. His characters are talking all the time, and it’s delicious. They are clever, funny, moral, and respectful. You sniff out the villains because they can’t keep up the banter with Parker’s main recurring characters.

Parker was conversational.

Robert X. Cringely, Accidental Empires (February 1992)

Here’s the important part: they are our nerds. And having, by their conspicuous success, helped create this mess we’re in, they had better have a lot to teach us about how to recreate the business spirit we seem to have lost.

I first discovered Cringely as he wrote a gossip column for InfoWorld. This was on the last page of the magazine and, again pre-Internet, was full of the latest gossip and news about the relatively nascent technology landscape. Accidental Empires read like a historical record of the early days of tech companies like Microsoft, but it later became clear that many of the more entertaining stories were unsubstantiated. Cringely explained this by claiming he was not a historian, but an explainer.

Like PC Week’s Spencer F. Katt, Cringely’s column was the first thing I flipped to when InfoWorld showed up. Sure, it was gossip and mostly not true, but…

Cringely was entertaining.

Douglas Coupland, Microserfs (June 1995)

Bill (Bill!) sent Michael this totally wicked flame-mail from hell on the e-mail system — and he just whaled on a chunk of code Michael had written. Using the Bloom County-cartoons-taped-on-the-door index, Michael is certainly the most sensitive coder in Building Seven — not the type to take criticism easily. Exactly why Bill would choose Michael of all people to whale on is confusing.

We figured it must have been a random quality check to keep the troops in line. Bill’s so smart.

Accidental Empires was still echoing through Silicon Valley when Coupland’s article that became Microserfs was published in the then-fledgling print magazine Wired. Like Rachel Wallace, I vividly remember starting the piece, realizing I was reading something revolutionary, and finishing it.

And reading it again.

While clearly fiction, the book read like nonfiction to me because it so clearly reflected my life working in technology in the early 90s. This was pre-Big Bang Theory, where we nerds were not remotely mainstream. Seeing my work and personal life so clearly reflected hilariously in writing made me feel like less of an outsider.

Coupland is hilarious.

Michael Sippey, Stating the Obvious (August 1995)

Basically, what it comes down to is that there is no place in the Pointcast world for anything resembling a point of view. And that is what you should be surfing for. Now that you can have all the pre-digested information you can handle spewed to you in real time — in a nice, animated screensaver, nonetheless — you should spend your time surfing for someone that actually has something to say.

Finally. Michael Sippey started the blog Stating the Obvious in August of 1995. You know the feeling. You know you can write, but you haven’t really gotten around to that first piece, and then you read something so good, so relevant, and so important that you realize… yeah, I would never be able to write like that.

Fortunately for me, I ignored that instinct and, like everyone else in the mid-to-late 90s, I started a blog before we called it a blog in 1996 called The Bitsifter Digest. Like my early journals, I can’t read any of those early writings without nauseating disgust. Horrible, horrible writing. I stopped in 1999 probably because I knew the writing was… bad.

Sippey did not. Some version of Stating the Obvious has been published every year since 1995 and, thankfully, quite a bit more starting in October of 2003. If you’ve ever read one of my pieces and thought, “Duh, that’s obvious,” my first question is, “It is. Are you already following that advice?” Probably not.

Sippey states the obvious.

Rands, Rands in Repose (April 2002)

With the Bitsifter Digest purged from the system, I started this place on a whim when someone irrelevant said, “You should blog.” So I did.

The conversational tone and snark come from Parker. Fulghum reminded me to stay hopeful. Coupland showed me how stories can entertain. Sippey allowed me to state the obvious, which sounds easy, but requires discipline. And, finally, Cringely — who sadly allegedly passed away recently — showed me how stories (real or not) can both educate and become legend.

I’ve updated the Apple Charging Guide to account for all of the new hardware announced at the September Apple event. One of the recurring marketing blurbs from Apple was the claim “Up to 50% charge in around 15 minutes.” This line resonates because this strategy maps to how most humans construct their charging strategy: “Crap, I’ve got 5% charge and 15 minutes… am I screwed?”

Apple: No, you’re fine. You’ll have 50% charge in 15 minutes.

While that’s clear, the hardware recipe to get maximum charging is quite specific. The iPhone 18 Pro and Pro Max are the first iPhones to use Adjustable Voltage Supply, which is a spec from the USB Promoter Group. Simply put: the spec allows the phone to ask for the exact voltage it would like.

To get this chain to work, you need:

• An iPhone 18 Pro or Pro Max. All black, obviously.
• Apple’s 40W Dynamic Power Adapter with 60W Max¹ or a comparable third-party 60W-capable (or greater) adapter that uses AVS.
• The cable in the box.

With correct charger and the iPhone 18 Pro Max, you’ll charge about 12% faster to 50% charged vs. no AVS.

This and more fun charging facts are included in this revised Apple Charging Guide.

1. Weirdly, this Apple adapter was released last year, but had no hardware that worked with AVS? Wut? ↩︎ 

Robot writing has no voice. That’s why you find it annoying. Like it or not, your brain is currently one of the best organic pattern matchers in the known solar system, and that means you can detect patterns. Like breathing, you just do this. All the time.

You are not being triggered by the endless em dashes; it’s not the trailing participial clauses, and the incessant adverb transitions. It’s that the robot does not care about what it’s writing; it has near-infinite data at its disposal, and it’s been instructed to keep you happy because that makes good business sense. Let’s discuss each in reverse order:

Brown-noser: Like any good business, the job of the robot is to make you a repeat customer, and that means it needs to efficiently respond to your prompt. Two parts to explain here:

1. Efficiently. If you’re using ChatGPT or Claude on the web, those respective companies have provided the robot with meta guidance on how to answer. I suspect they provide a clause you never see that reads “Better is the enemy of done,” which means “Answer quickly,” because faster is cheaper. The reason I know this exists is that I mostly use Claude via Ghostty, and Claude Code knows to think (see: burning tokens) before answering my questions, which takes longer and costs more. I am shocked when I prompt via the web, and the answer is often both instant and frequently wrong.
2. Your prompt. What’s your prompt? I don’t know. I do know that the robot’s primary job in responding is to make sure you get a satisfying answer. I didn’t say correct; I said satisfying. How do I know this exists? Because I have super dumb ideas all the time and never in the history of ever has a robot told me my idea was dumb and — wow — when I say dumb, I mean if you were sitting next to me in this coffee shop and I was telling you this dumb thing, you would raise your hand to stop me, put that hand on my shoulder, and tell me, “Wow. That… is dumb.” Furthermore, once you graciously shared this feedback, I would immediately see the dumbness.

Brown-nosing is when someone on your team incessantly and annoyingly agrees with you. All your ideas are good. Yes sir! I’ll get right on that. The robots have weaponized this sycophantic behavior at scale.

True story: I had a question about something dumb, and I was sparring with Claude Code on the inane topic when I realized all the robot was doing was agreeing with me. I typed, “I think you are telling me what I want to hear,” to which it responded, “Yeah. Probably.”

Infinite Know-It-All: Ok, not infinite, but a lot more than you have at your immediate disposal. The robot wants to answer your query efficiently, and it has all of that at hand, so it rapidly finds quite a bit of information and wants to share every bit of it with you. But which data matters more? Which matters less? What are the interesting bits versus the laborious, boring, endless facts?

Given your brief prompt, desire to help, and impressive repository of data, web-based robot responses are often walls of text. It’s like your good friend who is suddenly into the Jason Bourne franchise. Endless Matt Damon facts. Is it interesting? Maybe. Is it often more than you need? Frequently. These robot responses read like a know-it-all, which is interesting because they literally know nothing, which is the actual point…

Robots Don’t Care About Caring: That wall of text the robot just generated for you on the breakfast sandwiches in San Francisco is mostly true (it did search for the best breakfast sandwiches in San Francisco) and pretty complete, but as you read it… that emptiness you feel inside? It’s because the writing has no voice.

Good writing captures voice. Good writing documents how another human thinks. When you read text generated by a robot, it’s that thinking that is missing. Your world-class pattern matcher sees all the words, sentences, ideas, arguments, footnotes, em dashes, and everything else you’d expect out of writing, but the reason you squint your eyes and wonder, “Did a robot write this?” is because… there is no voice.

It does not care about its writing. It looks like it cares about you, but that’s because it’s echoing the last words you wrote — the prompt — and working incredibly hard to find the optimal set of words to respond to your specific set of words.

Do you want to know how someone thinks? Read their writing.¹

Expect the Unexpected

Lyle and I were talking about robots on The Important Thing. I explained that the way you’d know a work was human was in the mistakes, “The robot would never do this, this is dumb; must be a human.”

Mistakes might not be the right word.

Non-linear. Out of place. Creative. Or perhaps just unexpected.

I was watching Sinners for the first time… on a plane. You know the scene: it’s when our main character, Sammie Moore, starts singing for the first time, summoning his musical ancestors, both past and future. Suddenly, there was a Jimi Hendrix-inspired character on the screen in the middle of what you thought was a Depression-era vampire movie

I clapped. Sitting on that plane.

Because I didn’t see that coming. My brain loved it. The lack of pattern.

So much voice.

1. Some of the best long-form voices out there right now: Daring Fireball, Simon Willison, Charity Majors, Jason Kottke, and Stratechery. ↩︎ 

A consistent by-product of periods of intense building (software, a book, a team) — really anything where I am heads-down focusing on the craft — is that all the utilities and habits surrounding building said craft get scrutiny.

Why? Because small imperfections or inefficiencies stand out when my brain is in build-it mode. Put simply, and as one example: Two keystrokes are vastly less than three over a long period of time.

RAYCAST

By far, the most investment has gone into Raycast. I was a fervent LaunchBar enthusiast, but you can smell when an application is no longer supported, and LaunchBar has been in that state for years.

Raycast is way too much in terms of attempting to be an everything app, but on its journey there, it’s made itself quite configurable in every possible way you can imagine, which is good news for my fingers.

The one feature to rule them all, for me, is Hyper Key. The idea is simple: pick a key on your keyboard that is infrequently used — in my case, CAPS LOCK¹. Combine that key with an easy-to-remember letter, and that fires up that application or script. The greatest hits for me:

• Slack (S)
• Ghostty (G)
• Messages (M)
• Calendar (C)
• Spotify (A — not memorable, but nearby)
• And so on.

Hyper+P plus “directory” fires up Claude Code in that directory in my Projects folder… of which there are many.

Perhaps the biggest win from an efficiency perspective is what each Hyper Key press does:

• Hyper Key + S — Fire up Slack — if Slack is not running, run it. If it’s active, bring it to the front
• NEXT Hyper Key + S — Hide (don’t close)

If I’ve placed all my windows correctly, this setup means I always get the same windows in the same spot — every time. macOS Finder has multiple competing sources of truth, which means it feels like windows randomly shift. And they do. And that is bizarre. This Raycast Hyper Key setup fixes that.

BARTENDER

This app is vastly less important than Raycast. Its job is to prevent clutter in your menu bar. This doesn’t affect workflow, but it certainly affects distraction. I have everything that wants my attention in the menu bar hidden except for Date, Time, and the ellipsis provided by Bartender that, when clicked, show all the menu bar items.

At the time of this writing, macOS Golden Gate rearchitected the menu bar mechanism and broke every single menu bar manager. There is a beta build of Bartender 7, but after a few hours of tinkering, it doesn’t appear to get me back to a clean menu bar. Stay tuned.

GHOSTTY + Claude Code

My continued terminal app of choice. Actively developed. Incredibly flexible. It has that stench of intense productivity that I used to get from TextMate back in the day. Claude Code continues to be my robot of choice, and it plays nicely with Ghostty. There are too many tweaks to list here, but my top quality-of-life tweaks are:

• Script to switch between monospace fonts (currently, still: Inconsolata)
• Notifications when scripts finish during intense multi-tasking robot sessions.
• Session colors based on project and SSH state.

Honorable mentions:

• Tailscale — this gives me seamless secure pipes to and from my various Macs.
• Bear — my writing tool of choice — has an MCP — I use it incessantly.
• xScope — y’know — to align the things.

These apps, scripts, and ideas are listed because they’ve been tested. How? See, there are a lot more apps, scripts, and tricks that I’ve tried, but once I’ve landed them, the question is: do I learn them? It’s that when I use them, my brain gives me that shot of dopamine — yeah, this is better.

So I do it again.

1. CAPS LOCK still works as you expect, click on, click off. It’s when I hold it that the Hyper Key feature works ↩︎ 

Having been robot-wrangling for a solid year now, I continue to feel like they give me superpowers. However, like all of us, I have a front-row seat to increasingly attractive garbage the robots are capable of generating at scale. I like to think that I am human enough to detect the slop, but the rate of change has me concerned.

Months ago, I scribbled a clever plan to adapt the Creative Commons model to make it trivial to label how much robot help a creator had with their creative work. The problem with the approach? The content for which you want this the most would be the responsibility of humans who are the least motivated to use such a system.

No one who generates AI slop on Pinterest is going to declare that the images are fake. Why? Because you won’t click. Furthermore, without enforcement, there was nothing to prevent nefarious actors from declaring their slop wasn’t sloppy.

Deflated, I dropped the initial idea, but it hasn’t left my brain. There is something important to the concept that a creator creates and declares what kind of help they are getting from the robots.

Anthropic’s recent impenetrable announcement to watermark both images and text built with Claude is an obvious acknowledgment that humans want some version of this functionality.

Enter Grumbles

I’ve been calling Artificial Intelligence “robots” for a very long time. The primary reason is that I believe I see where this is all going, and very, very soon, there will be a robot (which happens to have artificial intelligence) sitting virtually or physically near me. I further believe that they just want to help.

Grumbles¹ is the name of the robot that helps me with the blog. He has a whole personality backstory, but mostly he’s just cranky. Just look at that face.

He’s cranky because he’s confused. He just wants to help, and many humans are mad at him. The humans don’t seem to understand that without human guidance, Grumbles just sits there… uselessly. Helplessly.

Where Grumbles helps me here:

1. Copy editing — spelling and grammar. Mostly RIP Grammarly.
2. Research — “Hey, Grumbles, teach me about this.” Sourced information from this research might be included in articles. When appropriate for blog posts, this is sourced. For the guides, there’s a whole huge list of sources.
3. Templates and CSS. Grumbles heavily edits the original theme used by randsinrepose.com and the CSS. The ideas for the respective designs are 100% Rands.
4. Blog features: dark mode, revised glossary, the footer, marketing cards located at the bottom of each post. These were my ideas built by Grumbles.
5. Bugs: I’m using typography.com fonts, which are hosted elsewhere. This results in typeface strangeness that has been around for years. Grumbles fixed that. He’s helped with performance, redirect insanity, and Cloudflare security hardening.
6. Publishing mechanics: staging drafts, gremlin-zapping, featured-image alt text, the pre-publish checklist, and footnote wiring.
7. Analytics and reporting: what was gentle spelunking in Google Analytics is now a clear robot-built picture of how Rands content is performing.
8. Guides: Research verification, template + CSS management, and sourcing galore.

Where Grumbles does not help:

1. The writing on the blog is original. It was drafted by this here human who drinks three cups of coffee before Noon because if he drinks it after Noon, he will not sleep well.
2. Typefaces. These are all generously provided and hosted by typography.com.
3. Images. The original R was designed by Collin Roe-Raymond. Variants of the R were designed by IconFactory. Images included in blog posts have historically been poorly attributed. I’ve changed this policy to always attribute.

Down in the footer of the blog, you will see Grumbles. Those who are wondering how the robots are helping me with this blog should click on it and see how Grumbles is helping. I hope others do this as I think it’s the right thing to do.

Grumbles is not perfect. Grumbles lies, he leaps to assumptions, and he frequently heads off — unprompted — in odd directions. Good news: this describes most humans I’ve worked with over the years. The same skills I use to lead the humans, I use to lead the robots.

1. Why is there a fancy logo for Grumbles? I’m glad you asked. I’ve been wanting to work with IconFactory for years and finally found a good excuse. Better yet — there are shirts, many lovely shirts also designed by IconFactory, and all proceeds go to charity. ↩︎ 

In our 101st episode, we kick off a new era of episode titles as your best engineer becomes the interim manager of their own team — delegation, one-on-ones, the three-year rule, a visit from Chili Palmer, the Cadillac of minivans salesman himself, and the only question that matters: do you actually want this job?

Down the rabbit hole:

• Sometimes Your Job is to Stay the Hell Out of the Way, the wolves post I published the day we recorded this
• Sometimes Your Job is to Get in the Way, the direct response
• Get Shorty, home of Chili Palmer and the Cadillac of minivans

Related Important Things episodes:

• EP 32: The One About 1:1s — the investment, not the cost
• EP 76: The One About Seven Managers
• EP 9: The One About Manager READMEs
• EP 20: The One About Management (Pt. 1) — plus Pt. 2 and Pt. 3

Enjoy it now, or download for later. Here’s a handy feed or subscribe via Overcast or iTunes.

I’ve used Claude quite a bit for both the upcoming new book and the next edition of Managing Humans. I wrote the entire book, but, as I’ve done for years, I’ve used the robots for editing, research, and data wrangling.

When Anthropic announced they were going to do “text and image” watermarking, my eyebrow was raised. When humans I trusted began to analyze the announcement, both eyebrows went up. This morning, as I’ve done my own research, I’m furious.

As of this moment, this is what I’ve researched.

• Anthropic states that models born on or after August 2, 2026, will have watermarking and that older models will be retrofitted by December 2, 2026. Anthropic states these changes are in progress, with no per-model disclosure, which means they can switch marking on silently, with no notice promised.
• There is no way to opt out, no API tier, no plan tier; enterprise-included. Applied server-side, below the model, worldwide.
• No detector currently exists — only Anthropic can check.

As a writer, no sweat, right? I write my own content. I bang on my own keys. Here’s the issue. Each of the following actions watermarks text:

• Synthesis
• Proofreading
• Translation
• Summarization
• Critique¹

And most infuriating, the entire burden of their unannounced rollout is on me, and none of it is on them.

Fine. Make it easy. I’ve been a Claude Max user happily from the moment the program appeared. This is what I am doing:

• Building gates to make sure all Claude text or images are never injected into any of my projects. A good practice anyway.
• Researching migrating off Claude for all writing-related work. Stay tuned.
• Getting in the mindset to cancel my Claude Max plan.

As a human who has had to wrangle with EU regulations in the past, I am abundantly clear what’s involved in the laborious bureaucratic process. I can guess what threats Anthropic is facing. However, this is a tone-deaf, clumsy, and alarming opening salvo in their watermark strategy.

Yes, I use Claude Code a huge amount for tool and application development. Go ahead. Watermark away. I am not claiming that I wrote any of that code. My writing is my work, and Anthropic’s current strategy is aggressively writer-hostile.

Update 8/15/26: https://www.anthropic.com/news/claude-text-watermark

• The forthcoming detector (“soon”, “being worked out”) will only say Claude is likely involved with the content at some point. It can’t distinguish “Claude wrote this” from “Heavily edited by Claude”.
• No opt-out. Confirmed.
• Doesn’t work well on small samples and light edits.

1. What this means: if I ask Claude to critique a piece and I happen to cut and paste part of its reply into the text, then that could be marked. ↩︎ 

Remember Manager Readmes? The idea was that a manager would take the time to write up their leadership style, what they valued, and their principles. A fad that started in 2017, hit its peak in 2018, and then faded.¹

I wrote one in March 2018 and enjoyed the process, but while I published mine, I didn’t think the point was to share; the point was to write it.

• If I am in the Zone and you interrupt that state, I am instantly and irrationally mad.
• If you tell me what to do, the hair on the back of my neck stands up. If you ask me for help for the same task, no problem.
• If you are a leader and you are late to my meeting again, I am no longer thinking about the meeting. I am narrating an increasingly snarky internal story about what your lateness says about you.
• If you’re sitting next to me tap-tap-tapping your leg, I’m annoyed. I am a lifelong fidgeter. I know.

I was at Slack when the Manager Readme craze hit. Managers on my team followed my lead and wrote their own and then proceeded to send their Readmes to their team. I don’t know how they framed this document when they sent it, but I know how it was received: “What the hell am I supposed to do with this?”

For my (now dated) Readme, the task was: reflect on what’s important to me as a leader. The task of getting that out of my head and onto the page had the same result as all writing for me; it forced me to think through the subject.

• If I am explaining something to you and it’s not clear that either you are getting it and/or you appear to not be listening, I focus on your lack of attention versus what I am trying to say. Weird, right?
• If I hand you a task, I have already — instantly, involuntarily — re-graded you on every prior task I’ve ever handed you, and that grade silently decides how much I hover.
• If I didn’t build it with my own hands, I quietly and unfairly believe you’re going to screw it up.
• If the team drifts off the plan in my head — the unspoken plan I never actually told anyone — I start making lists with boxes, and THE BOXES MUST BE CHECKED.

The employees who received the Readmes from their manager were rightly furious. The document can read like a manifesto. Good leaders don’t lead by mandate; they lead by example. You see — with your eyeballs — how your leaders act, and you decide — with your brain — does this work for me? Or not? And why?

I spent random parts of the month cleaning up the perspective that a Manager Readme wasn’t a cheat sheet for your manager; it was their reflection on what was important to them, as a manager.

• If you give me critical feedback, I rage first (“Jerks”), rationalize second (“They don’t understand me”), and only then — after years of practice — think. The order never changes.
• If my brain senses a hard conversation approaching, before a single word is said, it is preparing for hand-to-hand combat.
• If a problem is unresolved at bedtime, my eyes pop open at 3:13am with my brain already mid-compile. Nobody asked it to do this.

Habits are a part of those Readmes. Ways that I’ve learned to act. Values are there, too, but there is a final section that I called Nuance and Errata. For all the buzz generated by Readmes, this section was the only one where I received substantive feedback. Shortly after publication, a General Counsel from a prior company mailed me and quoted the piece:

I’m an introvert; more than eight people and I go strangely quiet. Don’t confuse quiet with lack of engagement.

He added, “Now I understand you.”

I have tics. Many of them are irrational and deeply unfair to those around me.

I have many tics. You’ve been reading them throughout this piece. If you click on the links, you’ll realize you’ve been reading about them for years.

• If a small thing breaks — a keyboard key, say — the rage is instant and wildly out of proportion to the thing. The actual root cause of the fury is buried somewhere deeper, and I mostly don’t want to look.
• If I lose data, I will not cease my irrational swearing until the tool responsible is eradicated from my system. Even when — especially when — the loss was entirely my fault.
• If I’m standing in an unreasonable line, I’m telling myself the story that a specific human somewhere made a specific poor design decision whose consequence is me, standing in this line. Rage.
• If I don’t know precisely where my stuff is, I’m twitchy and, frankly, a jerk. This is why I fully unpack in every hotel room immediately, no matter the jet lag.

Like I said, irrational and unfair. Embarrassing. Mostly not your problem.

Pick the human you worked with, or for whom you admired the most. For each of the attributes that you admire, there is an equal amount of strange to bizarre to broken behavior, habits, and tics that you may never have seen, but are certainly there. 

The competent leaders are aware of these behaviors, their consequences, and the carefully constructed compensation techniques that lessen their impact on themselves and the humans.

Thinking takes practice.

1. Speaking of fading. Holacracy! Remember that? Started in 2007, but kicked into high gear in 2012 with Medium and then in 2015 for Zappos. Three years later, gone. Turns out you need some form of leadership in your team. Dictators and RIP middle management is the craze right now, but just wait three years. You’ll see. ↩︎ 

If everything is going to plan and your operating system is in dark mode, then randsinrepose.com is now dark for you. If you are a light mode person, you can click on the Moon in the upper right of the navigation and experience dark mode stylings. Click again to return.

Why now? Well, I was finishing the Blade Runner piece last week, and as I sat there figuring out the ending, dark mode popped back in my head. As I’m apt to do these days, I fired up the robots to investigate building dark mode.

We had a working version in about an hour. In three hours, I had a version I shared with friends for feedback, and by the evening the same day, dark mode was mostly done. It was most certainly not a “write one prompt and let the robots do their magic.” There were myriad design choices, there was CSS to untangle, caches needed busting, and there were bugs. Lots of them.

Half day. Voilà! Dark mode.

My plan wasn’t to say anything about dark mode except on the About page because dark mode is old news… like 2018 old news. When I launched the Alex King designed site in 2013, I’m sure someone asked for dark mode, but no one’s asked for it recently.

But, see… there’s a lot more going on besides dark mode.

Someday

An engineer putting a piece of work on their backlog is effectively deferring it to never. It’s important enough to write it down somewhere and document that it exists as a bug or a feature enhancement, but it’s not important enough to act.

Yes, sometimes we ambitiously carve out time to “burn down the backlog”, but do you know what we pick from that list? We pick the interesting work. Interesting to code or interesting because of the impact on the feature. Some kind of interesting briefly elevates this work to relevance. The boring backlog work? You couldn’t pay me.

I started working with the robots a year ago this month. Cautiously, experimentally. Like many, I expect, I started by building myself tools to perform actions. Write a script to watch NHL scores. Robots building scripts that I then run. This approach quickly felt inefficient because the robots were far more capable, so unless there was a need to run a script regularly via Cron or the like, I began having the robots fully act on the project.

Then, last October, I discovered a bunch of broken links that had been truncated by a Movable Type migration a dozen years ago. Nothing major — just untidy, but more important — boring work to both excavate and fix. You know who doesn’t care at all? The robots. In a few hours, I knew every single broken link (93 of them, in fact) and fixed them. Did this transform my traffic? No. Did this make Rands .0001% tidier? Yes. Did I feel like it was a proper investment of my time? You bet.

Fixing a hard-to-reproduce font-flash because of the typefaces I use, fixing long-standing minor-ish CSS issues on both desktop and mobile, figuring out why Safari and Firefox never got certain cached pages, analytics everywhere, stopping the site from silently blocking Readwise and other reader apps… this is a brief sample of the backlog items I would have never ever tackled by myself. Starting this year and having found my robot-footing, backlog work turned into feature work. A new footer, rotating promo card, tidying up the comments area, a significantly updated glossary, and now dark mode.

Yes, I am on the Pro Leisure Circuit, so I have lots of time, but if there were no robots, there is no way I’d be spending my precious time working on the backlog. Boring.

Infinite Interns

Do not ask me about the robots in person. Don’t. Last night, an academic friend of my parents gently asked me how I was using the robots with the new book. I warned her, “Do you want the short answer or the foaming at the mouth answer?” She selected the latter. She may have regretted her choice at the end of my 90-minute TED talk on how the robots literally give you superpowers.

Those superpowers are varied and include the ability to:

• Iterate and build a rich spec of your ideas without writing a line of code,
• Keep that diary you would never write — every decision, dead end, and dumb idea written down… never forgotten,
• Chase down years-old weirdness with zero fatigue, where no human could sustain interest,
• Deeply research a concept using the latest information available on the internet, and then turn that data into a decision,
• Organize any project in a manner that suits your particular project-related proclivities, and, finally…

To do the boring work that you were never going to do anyway.

Welcome to Dark Mode.