AI = Almost Intelligence
The Intelligence We Settled For
I’m starting to think the most revealing thing about artificial intelligence has been sitting there in the initials the whole time.
AI.
Almost Intelligence.
Not useless, which would be an easy dodge, and frankly, a stupid one. I mean, the stuff is truly mind blowing. It can write, summarize, mimic, translate, brainstorm, analyze, generate, revise, and respond with a level of fluency that would have sounded like science fiction not that long ago. Pretending it’s merely a toy is just another form of denial.
But pretending it’s the same thing as intelligence may be the more dangerous denial.
Because “almost” is not nothing. Almost can be powerful. Almost can be persuasive. Almost can save your day, rescue your deadline, polish your mess, and give you the strange, intoxicating feeling that you just became more capable than you were five minutes ago.
That’s why this is so tricky.
If AI were obviously fake, clumsy, or ridiculous, we wouldn’t be having this conversation. Nobody loses themselves to something they can easily dismiss.
The danger is not that AI fails so badly we reject it. The danger is that it succeeds just enough that we start lowering the threshold for what we’re willing to call thinking, writing, creating, deciding, and even being ourselves.
That’s the territory I think we’re in now.
Not the age of fake intelligence.
The age of almost.
Let’s Look It Up
Merriam-Webster defines almost as “very nearly but not exactly or entirely.” That definition feels almost too perfect for what we’re living through. Because the danger of AI is not that it misses by a mile. The danger is that it comes close enough to make the miss feel irrelevant.
It very nearly understands. It very nearly sounds human. It very nearly captures your voice. It very nearly produces insight. It very nearly carries conviction. And because the result is so close, we begin treating the gap as technical rather than existential.
But “not exactly or entirely” is not a footnote. It’s the fault line. That missing remainder is where lived experience, judgment and authorship call home. It’s where the part of you that cannot be scraped, averaged, predicted, or politely optimized still lives.
The Seduction of Almost
Almost has a very particular psychological power. It doesn’t announce itself as counterfeit. It arrives as relief.
You’re tired, so it writes the email. You’re unsure, so it gives you ten angles. You’re stuck, so it finds the structure. You’re bored with the blank page, so it gives you something to react to. You’re under pressure, so it produces language faster than your own nervous system can catch up.
And because the output is often useful, the transaction feels innocent.
That’s the part people keep missing. The danger usually doesn’t arrive as some grand surrender of human identity. It arrives as a tiny bargain made under ordinary pressure. A sentence here. A subject line there. A post you didn’t feel like writing. A reply you didn’t want to think through. A difficult idea you let the machine soften because you were too tired to carry the tension yourself.
None of this feels dramatic in the moment.
It feels efficient, like a strange cross between cutting edge and keeping up with everyone else.
But if you repeat that bargain long enough, something begins to shift. Not all at once or in some cartoonish “the robots took over” way. More like a slow change in creative posture.
You stop approaching your own mind as the source and start approaching it as quality control. You stop listening for the thing only you would have said and start scanning for the version that sounds acceptable enough to send.
That’s not a small change in workflow we’re talking about. That’s a change in authorship. And once that genie’s out of the bottle it’s almost impossible to put it back in.
The Almost-You Problem
The most interesting question is no longer whether AI can sound human. Of course it can sound human. That question is already stale.
The more important question is whether humans are slowly becoming comfortable sounding less fully like themselves.
That’s where “Almost Intelligence” turns into something more personal. Because the machine doesn’t just create almost-content. It can help create an almost-you.
The almost-you is very appealing. It is cleaner than you. Faster than you. Less moody. Less hesitant. Less likely to ramble down a weird side alley. Less likely to leave a sentence rough because the roughness carries something true.
The almost-you has better transitions, fewer jagged edges, a more consistent tone, and a strange ability to make every thought sound like it belongs in a professional newsletter.
And for a while, that feels like an upgrade.
Your emails get smoother. Your posts get clearer. Your ideas get packaged faster. Your rough drafts look less embarrassing. Your output increases. From the outside, it may even look like you’ve found your rhythm. Like you’re always in the zone.
And hey, that’s a very seductive feeling.
But something quieter may be happening underneath.
The real you starts getting trained out of the process.
Not erased, exactly. That’s too dramatic. More like gently demoted.
Your instincts become raw material. Your taste becomes a filter. Your judgment becomes a final pass. Your lived experience becomes something you feed into the machine so it can return a more polished imitation of what you might have meant.
That is a strange kind of loss because it doesn’t feel like loss at first. It feels like help.
And that may be the most dangerous thing about it.
The New Creative Avoidance
We keep talking about AI as a productivity tool, but I think one of its most powerful functions is avoidance disguised as productivity.
It lets you remain near the work without entering the hardest part of the work.
You can keep prompting instead of choosing. You can keep generating instead of deciding. You can keep refining instead of risking. You can keep asking for another angle instead of admitting the angle you already have is the one that scares you because it actually says something.
This is where AI has made an old human problem worse.
Avoiding action has always been easy. People have always found ways to delay, research, revise, rehearse, plan, and prepare themselves into paralysis. The blank page has always had escape routes. AI just industrialized them.
Now avoidance comes with outputs.
That’s new.
You can spend an hour generating headlines and feel like you worked. You can create ten versions of a sales page and feel like you made progress. You can ask for a “stronger,” “more emotional,” “more human,” “more persuasive” version until you no longer know what you were trying to say in the first place.
This is the prompt-edit-repeat loop dressed up as creative diligence.
And it can feel very responsible, like you’re improving the work when you may actually be avoiding the moment where your own conviction has to become visible.
At some point, more output stops being possibility and becomes cover.
That’s when Almost Intelligence becomes a hiding place.
Why Almost Is So Hard to Resist
The reason this is hard to talk about is because AI really does help.
That’s why the usual anti-AI scolding falls flat. People know it helps. They have felt the relief. They have watched it rescue a bad afternoon, clarify a muddled idea, or give them language when their own language was jammed somewhere behind fatigue, doubt, and too many browser tabs.
So the issue isn’t whether AI should be used.
That question is boring now.
The real issue is what kind of human you become while using it.
Do you become sharper, more honest, more discerning, more willing to wrestle with your own raw material?
Or do you become dependent on the machine’s ability to make your half-formed thoughts look finished before you’ve actually finished thinking them?
That’s the line in the sand.
And it’s not always easy to see because the outputs can be genuinely good. Sometimes better than what you would have produced alone.
But “better” is not a simple word anymore. Better for what? Better by whose standard? Better because it carries more truth, or better because it removed the awkwardness that made the piece alive? Better because it sharpened your conviction, or better because it made your uncertainty sound confident enough to publish?
This is where the old conversation about AI quality breaks down.
The question is not just “Is this good?”
The better question is “What did this cost me to accept?”
The First Bid Problem
One of the most dangerous things AI does is establish a baseline before you realize a baseline has been established.
The first output is not just a draft. It’s a bid. I talk about this more in my upcoming book, The Humanizers Strike Back.
It’s a bid for the tone.
A bid for the level of risk.
A bid for the emotional temperature.
A bid for how much specificity the work is going to carry.
A bid for how much of you is actually required.
And because the first bid is usually coherent, clean, and plausible, it can quietly reset your expectations. You start editing from its reality instead of your own. You start improving its premise instead of interrogating whether the premise belongs there at all.
This is how authorship gets softened.
Not because the machine has become self-aware like in the Terminator movies. But because it gives you something acceptable before you’ve had to fight for something true.
And “acceptable” is oh so tempting, because it lets you move on. It gives you the little dopamine hit of completion. It helps you avoid the ugly private stretch where the better idea is still unreachable and the easy idea is sitting right there, smiling politely, ready to be approved.
That’s why “never accept the first bid” is not just a prompting tactic. It’s a sovereignty practice.
And the first bid is where Almost Intelligence tries to set the terms.
Your job is to refuse those terms long enough for the real work to show up.
Almost Content, Almost Conviction
The internet is not filling up with terrible AI content.
That would almost be easier.
It is filling up with competent, readable, almost-content. Content that has the shape of usefulness, with the right pacing, the right formatting, the right implied authority, the right little gestures toward vulnerability. Content that knows how to behave like content.
And that may be worse than bad content because bad content at least announces its failure and it’s easy to spot and reject.
Almost-content slides through.
It clogs the feed with things that are not wrong enough to reject and not alive enough to remember. It creates the sensation that everyone is saying something while very few people are risking anything. It turns insight into a posture. It turns conviction into a tone setting. It turns human experience into a design element.
This is why people are developing that strange sixth sense around AI-generated writing. They may not be able to prove it. They may not be able to point to one sentence and say, “There, that’s the crime scene.” But they can feel the absence of pressure.
Nothing seems to have been endured.
Nothing seems to have been chosen at a cost.
Nothing seems to be at stake beyond engagement.
That’s the difference. AI can imitate the architecture of meaning, but it cannot supply the human burden underneath it. And when humans use it carelessly, we start producing work with architecture and no burden.
A house no one has lived in.
The Real Fight
AI = Almost Intelligence.
Almost is useful. Almost is powerful. Almost can help you move faster, think wider, and see possibilities you might have missed.
But almost cannot become the ceiling.
Not for your voice or your work or your judgment or your life.
Because once almost becomes good enough, the fight shifts from whether machines can imitate humans (they can) to whether humans will keep insisting on the parts of themselves that cannot be conveniently imitated.
The parts with memory, risk, contradiction, obsession, taste, grief, humor, defiance, and lived consequence.
The parts that make the work harder to produce and harder to fake.
That is where the real human signal still lives.
If this hit a little too close to home, that’s the point.
This is exactly the turf I’m treading on in my new book, The Humanizers Strike Back, which launches May 28.
It’s about the quiet coup happening inside our own creative standards — where the machine hands us a smoother version of ourselves, and we mistake that for progress.
We’re not going to put up with that any more.
I’m putting together the First Wave list now for people who want updates, behind-the-scenes notes, launch bonuses, and first access to what’s coming around the book.
Because I don’t want this to be just another book floating around the internet.
I want it to be a reminder that Almost Intelligence only wins if we let almost become enough.



You nailed it, Andy! "The danger is that it comes close enough to make the miss feel irrelevant."