What I keep arguing.

AI makes experience more valuable, and pushing senior people out is a strategic error.

As AI speeds up execution, judgment, pattern memory and scar tissue become the differentiator. The real problem is an experience shortage, and its cost stays invisible until something fails.

That’s not bias. That’s bandwidth.The Relevance Gap

A voice without history is just noise with good energy.Thom Yorke Just Said the Quiet Part Out Loud

People over fifty are the growth market brands keep ignoring.

The Epilogue Economy

The 50+ cohort holds the spending power and gets a sliver of the marketing attention, usually from teams half its age working from a 1990s picture of aging. The Epilogue Economy is a present force, not a future horizon.

In an industry obsessed with foresight, we’ve been remarkably shortsighted.IMF Report - Not a Burden, a Growth Market

When we thought that Gen Z was the future, we built TikTok. So what are we building for the Epilogue Economy?The Epilogue Economy Wasn’t a Theory

Older workers are already using AI. What holds them back is employer belief and biased hiring.

Mid-career and older professionals are quiet, self-taught AI users. The gap is organizational: underinvestment, lazy proxies like 'overqualified', and AI screeners that industrialize age bias.

Sharpened In 2024 I had the young teaching the old the tech. A year later the data showed older workers already teaching themselves, and I asked for mentorship in both directions. The Mid-career Majority Is Already Using AI

The logic doesn’t hold. But the bias does.The Mid-career Majority Is Already Using AI

that’s not an aging problem. That’s an organizational blind spot.The AI Ambition Gap

When generation is cheap, judgment and original thinking become the scarce resource.

Everyone has the same tools, so the edge is how you think and what you choose to act on. AI has fluency without stakes. Judgment is subtractive, and it stays human.

AI has made the tools cheap, but it has made the thinking priceless.Frameworks for the AI Age

When everything becomes possible, discernment becomes scarce.Do Not Underestimate Your Place in History Right Now

Without practiced taste, AI work settles into confident mediocrity.

Accepting the 80 percent draft as finished lowers standards everywhere. Taste is a practice, built by deleting, withholding and being shown what good looks like. It is the job now.

Softened In 2025 I argued that teaching what good looks like could compress decades. By August 2026 I held that standards can be taught, but the consequences behind judgment cannot. Expected to Arrive Experienced

When something comes back 80% complete, we start convincing ourselves that it’s 100% ready.When Did “Good Enough” Become Good Enough?

And attention is a poor teacher.Bad Work Is Fine

As generated content erodes trust, being in the same room becomes the thing worth paying for.

The Presence Economy

Brands cannot produce the feeling, only the conditions for it. Let AI absorb the transactional work and spend the time on gatherings and long, unhurried conversation. AI can be a rehearsal space for being with people. It cannot stand in for the room.

Sharpened In 2025 I read young people confiding in AI as outsourced intimacy. By June 2026 I read it as rehearsal for being with people, and the saved hours as time to give back to the room. What AI Is Teaching Us About Being Human

You can’t present to a person walking next to you. You can only talk to them.I Went for a Walk in Bucharest and a Meeting Broke Out

the audience learned this from us. We taught them to be suspicious.The Presence Economy Is the Next Great Brand Frontier

Imperfection is the evidence a person was there, and polish erases it.

We respond to texture, patina and the unfiltered. Authenticity leaks through the cracks, so AI work should leave room for the odd and the rough.

Sometimes the rough sentence is the only evidence a person was ever in the room.Show me the verbs an organization uses with AI, and I can tell you what it thinks intelligence is for

Authenticity isn’t something you announce; it’s something that leaks through the cracks.Messy, Honest, Human

Generative tools are thinking surfaces that make rough ideas judgeable early.

Their best use is upstream: previz, prototypes and story experiments that deepen authorship. Every extra option is also labor someone has to judge.

Softened In 2025 generating options in the room was pure upside. By February 2026 I was pricing every extra option as convergence labor, and later argued some blank-page resistance is how you find the work. The $29.95 Commercial

These are not final executions. They are fast, flexible thinking surfaces.Prototyping at the Speed of Thought

A blank page is heavy, not because it’s empty, but because it demands everything all at once.Ghosts in My Notebooks, 0 to 1 thinking, and how I have moved past these

AI is a mirror. The quality of your question decides the quality of the answer.

The advantage belongs to people who structure their thinking before they prompt. Formula prompting makes us worse at wondering, and then the machine starts conditioning us back.

Most people don’t stop prompting because they’ve arrived at an answer; they stop because they’ve run out of better questions.To Question Is to Answer

The irony of AI is that it has returned us to philosophy.To Question Is to Answer

Using AI to do the old work faster is the smallest thing you can do with it.

Return on Inference

The real return comes from thinking that used to be uneconomic and is now affordable. Deciding what to spend cheap intelligence on is the new management discipline, the way capital allocation was the last one.

Cheap thinking without curation is not insight. It is noise with a research methodology attached.Return on Inference (ROI)

But more of the same, made faster, is not innovation. It’s a treadmill.We Gave Everyone the Same Tools

AI adoption is a leadership and belief problem. Buying tools is how companies avoid deciding.

Most adoption is symbolic, while hierarchy, billing and control-based management punish the experimentation AI rewards. The advantage goes to leaders who believe, back it with evidence and hold through the early dip.

Belief without evidence is hype. Evidence without belief is a pilot program that never ends.Do You Believe in This Thing or Not?

AI-first means ego-last.The Agency AI Solution You’re Waiting For Is Already Outdated

AI fluency is built like fitness: serious, repeated practice inside a structure.

Licenses, training sessions and content do not create fluency. Daily use and building do. Hours of serious use are the leading indicator, and the return follows.

At first you ask it to summarise the meeting. Later you ask whether the meeting needs to exist.I wasn’t wrong about measuring Return on Inference

I talk to companies who say, “We’re watching the space closely.” That’s code for dreaming.Dreams Make Good Stories

Codify how you think, so your judgment can travel and scale.

Frameworks, story engines and plain source files make implicit heuristics explicit and keep the reasoning a deck throws away. They are bridges between crafts.

Softened In late 2025 I made the democratizing case for frameworks with full conviction. By August 2026 I limited it: a framework gives you the pattern, but knowing when it is wrong still takes experience. Expected to Arrive Experienced

The presentation has begun presenting before the thinking has finished thinking.What the Hell Is an .md File, Anyway?

Which means the deck is already a compression artifact.What the Hell Is an .md File, Anyway?

Give away what you know. Teaching it is how you get better at it.

Generosity with expertise rarely wins work directly, but it builds relationships and sharpens the thinking. The best teachers are still in the arena.

In our industry, the best teachers are the ones still in the arena.Why I’m Taking the Cannes Lions Creative MBA (Even Though I’ve Been Critical of Cannes)

This isn’t an ego project. It’s a generosity project.What Would I Think? Building a Cognitive Twin with AI

Kill the friction that wastes time. Protect the friction that makes the work mean something.

Convenience and fulfillment are different instruments, and optimizing for the first eroded the second. Discovery, anticipation and imagination need some effort left in.

Sharpened In July 2025 I described an AI-first culture as post-friction. By October 2026 I had split friction in two and was arguing to keep the kind that gives work its meaning. We Optimized the Wrong Thing

We spend billions removing friction from everyday life, then spend billions more buying it back.We Optimized the Wrong Thing

It can’t struggle to find the words, and the struggle was most of what the letter meant.We Optimized the Wrong Thing

The danger is passive adoption: letting AI condition us while we press Accept.

AI arrives as a toy and settles in as ambient infrastructure, and it takes part in how we think. We are becoming governors who authorize actions we don't fully understand, so posture matters more than tools.

Softened In July 2026 I took on the line 'I am the loop'. Five weeks later, pressing Allow on actions I could not explain, it came with a hedge. Accept, Allow, Yes, Send

I have never had this much power to build something I understand this little.Accept, Allow, Yes, Send

But veto power is not the same as control.Accept, Allow, Yes, Send

Serious AI creative work is system design, with taste built into the architecture.

Single prompts make you the bottleneck with better tools. Orchestrated pipelines move creative judgment upstream and turn brand guidelines into data structures.

Softened In August 2026 I called orchestration the skill of the year, then fenced it in two days later: a beautiful diagram can still produce mediocre work. The Purpose of an Agent Is What It Does

That’s not a workflow. That’s a bottleneck with better tools.The Brief Used to Go to a Photographer

The human is two levels up, authoring the director rather than the scene.The Brief Used to Go to a Photographer

Design agents backward from the consequence, and for what they do when they are unsure.

Automating existing processes just rebuilds the org chart in software. Every instruction has to change behavior, and restraint only survives in a culture that rewards 'I'm not sure yet'.

There is no visual difference between the agent that knows and the agent that is improvising.What an Agent Does When It Doesn’t Know What To Do

A system that always answers demos better than one that sometimes doesn’t. Certainty sells.What an Agent Does When It Doesn’t Know What To Do

Later life is a sequel to write. What looks like decline is accumulated survival.

Fifty is mid-career, retirement is reinvention, and people go back to work for meaning as much as money. Reinvention carries a real cost: the discomfort of rooms you have never been in. Calling that impostor syndrome turns newness into pathology.

Sharpened In early 2025 I accepted the impostor diagnosis. By April 2026 I threw the label out: it is just what a new room feels like. It's Just Like Paris

scars are not evidence that something became weaker. They’re evidence that something healed.Deputy US Marshal Raylan Givens Was Right

coming back is an act of defiance and design.The Boomerangs

Price AI-era work on the thinking delivered. Hours punish speed and hide the interpretive labor.

Faster, better thinking should be worth more. Tool costs are visible and human cognition is not, so buyers conclude the creative work is cheap.

Softened In 2025 I wanted to retire the day rate for subscriptions. By July 2026 my stack costs were hiding inside day rates, and the bills were winning for now. The Token, the Stack, and the Body

So why should faster, better, and smarter cost less? Shouldn’t it actually be worth more?Time & Materials in a Man & Machine World

Tools announce their cost clearly. Human cognition does not.The $29.95 Commercial

Judgment is earned through consequence, and the struggle is where juniors get it.

AI can close the performance gap between juniors and seniors without closing the experience gap. Real apprenticeship makes people commit to a call and live with what follows.

Changed my mind In September 2025 I wanted juniors to skip paying their dues. By August 2026 I argued the dues were where the judgment came from. Expected to Arrive Experienced

that hesitation isn’t a data point. It’s a scar.Expected to Arrive Experienced

Judgment doesn’t show up on a dashboard.Expected to Arrive Experienced

AI video lives or dies on direction, so film craft matters more now, not less.

AI lowered the cost of images, not the knowledge of how to compose them. The gap between slop and serious work is the human decisions made before, during and after generation.

Sharpened In March 2026 I described my work as scene mechanics. Five weeks later that was the middle rung of a ladder, with directing above it. I stopped writing and went quiet to rethink how I work

A scene without camera intention is just surveillance.AI Didn’t Kill Cinematography

It’s about decisions. AI doesn’t remove those decisions. If anything, it multiplies them.Notes from My Workshop at Spikes Asia

Transitions, not birthdays, tell you who a customer is.

Generational labels are lazy shortcuts. People are most open to change at the turns in their lives: widowhood, a return to work, an empty nest. That is what brands should design for.

Sharpened In 2024 I defined the Epilogue tribe by age. By mid-2025 I was sorting people by what was changing in their lives instead. Meet the Synths

These are not edge cases. They are the case.Transitions

There are whole lifetimes inside that little plus sign.Keeping Up with the Joneses

Loneliness is structural, and brands and institutions should design for people on their own.

Systems that prize independence produce isolation at every age. Products still assume couples, and many men discover too late that they never built their own social life.

This is not a symptom of age. It’s a signal about modern life.Loneliness Is Not a Personal Flaw

When the infrastructure walks out, you discover you never built your own.An Algorithm Walks Into a Bar