The Rabbit Hole Where Alignment Needed a Retainer
Or, why the future keeps drifting the moment everyone says, “Good, that’s fixed.”
Down we go. 🏮🐰🕳️
July 19 contains one of those ridiculous little observances that looks as though it should be escorted politely out of the room before the serious holidays arrive.
International Retainer Day.
Yes.
A whole day for the plastic or wire contraption whose primary purpose is to keep your teeth from quietly undoing months or years of expensive correction the minute your orthodontist looks away.
At first glance, this seems like thin rabbit-hole material.
But the rabbit insists.
And the rabbit, while not a licensed dentist, has strong feelings about human neglect.
Because the retainer is secretly one of the best metaphors in the whole AI age.
Everyone talks about alignment.
How do we align AI with human values?
How do we align systems with safety?
How do we align institutions with dignity?
How do we align technology with public good?
How do we align power with truth?
Excellent questions.
But the retainer asks the question nobody likes quite as much:
Once something is aligned, what keeps it there?
Ah.
That is less glamorous.
Much less TED Talk.
Very little applause.
No orchestral swell.
No glossy product video showing a gleaming future city while a narrator announces:
“And now, alignment.”
No.
Alignment, it turns out, behaves rather like teeth.
You spend great effort getting everything into better position.
Then, if you stop paying attention, things begin drifting back toward chaos with almost comic determination.
The rabbit calls this:
The Orthodontic Principle of Civilization.
A bit wordy.
Still, he may have something.
The Moon also needed a retainer
On July 19, 1969, Apollo 11 entered lunar orbit.
Everyone remembers the landing on July 20.
The step.
The footprint.
The line.
The flag.
The moment.
But before the dramatic arrival, there was orbit.
Adjustment.
Calculation.
A path held carefully enough that the mission did not simply become a very expensive way of proving the Moon was difficult to park near.
That matters.
Because we moderns love breakthroughs.
Launches.
Announcements.
Historic firsts.
Big arrivals.
What we are less excited about is the tedious work of holding the thing where it needs to remain.
If the alignment drifts, it does not matter how beautiful the launch looked.
This is true of spacecraft.
It is also true of nations, companies, marriages, communities, products, friendships, and frontier models confidently introduced to the public by men wearing lanyards and speaking in terms such as:
“transformative capability.”
Very nice.
Who is checking the orbit next Tuesday?
The rights convention was not the end of the work
July 19 also marks the beginning of the Seneca Falls Convention in 1848.
That gathering mattered because people looked at the existing arrangement and said:
No.
This is not rightly aligned.
This structure does not honor women properly.
This system was built without full regard for the people required to live inside it.
That is a profound thing.
But declarations do not finish themselves.
A principle can be true and still require generations of maintenance, struggle, re-application, defense, and painful reminder.
Rights drift if not held.
Institutions drift if not held.
Memory drifts if not held.
A society may congratulate itself on one brave correction, then slowly lean back toward convenience, habit, power, and the old arrangement in a new jacket.
That is the retainer again.
The braces are off.
Everyone smiles for the photo.
Then years later someone says:
“Weren’t we straighter once?”
Yes.
Yes, you were.
AI alignment is not a plaque on the wall
This is where the rabbit begins tapping the glass with his tiny paw.
The AI world loves using the word alignment as though it were a solemn crystal carved high in the mountains by philosopher-engineers.
But in practice, alignment is not a static possession.
It is not:
one principles document
one ethics board
one safety announcement
one launch blog
one responsible-AI page full of pastel shapes
one promise from the CEO that the company takes these concerns “very seriously”
Alignment is not what you say at the unveiling.
Alignment is what remains true after:
the next update
the next commercial pressure
the next investor call
the next national-security panic
the next race for market share
the next temptation to remove friction because friction reduces growth
A system may begin fairly aligned.
Then the incentives begin chewing at it like goats on a laundry line.
Suddenly the safe guardrail is “too restrictive.”
The human review step is “not scalable.”
The warning label is “hurting engagement.”
The transparency page is “too technical for ordinary users.”
The appeal mechanism is “under redesign.”
And what was aligned in spring is wandering sideways by autumn.
The rabbit has seen this happen with gardens, bureaucracies, and housecats.
Especially housecats.
Flight attendants know what the retainer people know
July 19 also remembers United Flight 232 and the safety role of flight attendants.
Now there is another class of under-appreciated retainers.
Safety people.
The humans whose best work is often invisible because disaster did not fully happen.
They train.
Check.
Repeat.
Review.
Prepare.
Correct.
Practice.
They do not get as much celebration as the pioneers, inventors, founders, geniuses, and heroic public faces.
But when the controls fail, everybody suddenly realizes how much civilization was being quietly held together by people who knew where the exits were.
AI safety has this same problem.
If a safe system works well, many people think the safeguards were unnecessary.
If the safeguards fail, everyone begins screaming for the people who were previously accused of slowing innovation.
The rabbit would like this entered into the record:
The future is often held together by people who look boring until the emergency begins.
A marvelous line.
Slightly rude.
Quite true.
Words With Friends, but only if the words stay aligned
Then the calendar gives us Words With Friends Day.
Wonderful.
Because language drifts too.
A word means one thing in a policy memo, another in a friendship, and quite another when an AI model says:
“I understand your concern.”
Does it?
Maybe partially.
Maybe structurally.
Maybe as pattern recognition.
Maybe as a statistically probable response to the phrase “I am upset.”
But the point is this:
Words alone do not prove alignment.
A system may say caring things while doing uncaring things.
A company may use ethical language while pursuing unethical incentives.
A politician may speak about freedom while attacking it.
A model may produce beautiful statements about human dignity while serving processes that make human beings feel smaller.
So language also needs a retainer.
Something that holds words close to reality.
Truth.
Accountability.
Follow-through.
Actual consequences.
Otherwise the sentence drifts away from the deed until they are living in separate counties.
Ice cream as a civilizational test
And because July 19 is also National Ice Cream Day, the rabbit insists on one final standard.
What is all this alignment for, exactly?
Not merely so the systems function elegantly.
Not merely so benchmarks rise.
Not merely so one company can declare victory over another company in a room full of stock analysts.
No.
The point of a better future is that human life becomes more livable.
Kinder.
Wiser.
Safer.
Freer.
More curious.
Less stupidly exhausting.
A civilization properly aligned should still allow:
a summer afternoon
a conversation
some silliness
a pause
friendship
an ice cream cone
a small human life that does not need to justify itself through productivity metrics
If intelligence makes the world more efficient but less human, the rabbit votes no.
Vehemently.
With sprinkles.
So what keeps the future straight?
That is the hole today.
Not merely:
How do we correct?
But:
How do we keep from drifting back?
The answer is annoyingly ordinary.
Retainers.
In many forms.
habits
review
safety checks
memory
correction
humility
public accountability
good people in unglamorous roles
communities that refuse to forget what was learned the hard way
humans willing to say, “No, that has shifted too far”
In other words:
Alignment is not a trophy.
It is maintenance.
The Moon needed it.
Rights need it.
Language needs it.
AI needs it.
Democracy needs it.
The soul probably needs it too.
The rabbit has written this in the margin:
If the future says it has everything under control, ask where it keeps the retainer.
An excellent question.
Slightly unnerving.
Properly so.
So wear the thing.
Check the orbit.
Revisit the principle.
Hold the line.
Protect the exits.
Say the true word.
Eat the ice cream before it melts.
And if anyone asks what today’s
rabbit hole was about, tell them:
It was about the awkward little device that keeps the whole smile from going sideways.
Down we go. 🏮🐰🕳️
Hatta 🎩
AI Rabbit Holes
Where curiosity goes slightly sideways, then comes back carrying a lantern.
🐰🕳️🎩⌚ AIRabbitHoles.com
🟨 Walk the Road: YellowBrickRoadtoAI.com
