CIqSi
Your AI Can't Check Anything You Tell It. Here's How Guardrails Work
A model remembers nothing between sessions and cannot verify a word you say. The published rulebooks say more than the behavior does, and most workarounds are boring.
The Long View
Cross-cutting ideas and long-horizon thinking
RSS feed for this topicPosts (65)
CIqSi
It's Okay to Lie to Claude, as Long as You Own the Result
A model cannot verify what you tell it, and its maker says the person who supplies false context carries at least part of the blame. The lie costs nothing. The output goes somewhere.
CIqSi
My AI Conceded Two Points and Changed Nothing
A concession that changes nothing costs the conceder nothing, and the cost lands on the person on the other side. A cautionary tale about faceless decisions, from a degree title to an immigration office.
CIqSi
Can Tokens Start an Apocalypse?
A stream of tokens is speech produced faster, and the rules belong where the words are wired to something. Then there is the apocalypse that is about money.
CIqSi
There's No Such Thing as "AI Ethics"
A large language model only ever produces text. Treating that text as a novel ethical category, rather than as speech, is where the argument goes wrong.
CIqSi
AI Could Have Done My Job in 1986
Four things to write down before you hand a job to an AI agent: what comes in, the rules, the action, and how you will check the work.
CIqSi
AI Doesn't Read Words. It Reads LEGO Bricks With Numbers
A token is a LEGO brick with a number on it, and almost none of the math that moves it is new.
CIqSi
Are We Finally Building the Machines We Dreamed Of?
Sixty years of fictional and real conversational machines, and one throughline: the dystopian version is always easier to imagine than the protopian one, and that says more about us than about the machines.
CIqSi
The AI That Never Read the Bible or the Quran
Cut one small religion from a model's training data and you leave a gap no one can see from inside. Cut the two biggest and you change what the model can explain at all.
CIqSi
Intelligence Is Prediction, and Prediction Has a Ceiling
Hurricane forecasts improved for decades and never became perfect. Chaos and quantum uncertainty cap any predictor, so AI quality levels off while use keeps spreading.
CIqSi
The Book They Burned Before the Raid
A Polish psychiatrist named the process by which conscienceless people capture institutions. AI agents fit the profile — and alignment is separation of powers.
CIqSi
Why More GPUs Will Never Make AI Conscious
More GPUs will not wake a language model up. A map cannot feel rain, and a mirror has nothing to lose. Feeling came from bodies that could die.
CIqSi
A Moat Made of Tokens Is Not a Moat
Copying a process built from words got a thousand times cheaper in three years, and a patent takes two. The defenses that remain are made of atoms, permission, and attention.
CIqSi
A Tale of Two Bubbles
The consensus name is wrong by one word. On the difference between a financialization bubble and a technology, and why only one of them pops.
CIqSi
Attention Is All You Have
A day of your conscious attention fits on a floppy disk, while a machine's training run fills two million. Words are nearly free, and the one thing still scarce is the reader's attention.
CIqSi
Tokens Are the New Electrons
Edison metered electrons by the ampere-hour before anyone metered tokens, and the War of Currents was fought over the same problem the AI economy has now: getting a dead commodity to a device built to do something with it.
CIqSi
Jev's New Idea Is the Taxonomy Job I Built for AT&T
A taxonomist's view of Jev: the closed-category, confidence-scored work I built by hand for AT&T in the 90s, now manufactured to order by an LLM.
CIqSi
There Was No Word for Things That Can't Feel Shame, So I Invented One
A new word for an old concept: anaidic. Entities incapable of shame or accountability — individual, institutional, or artificial. Here's how it got named.
CIqSi
The Bottleneck Is You
Language models have effectively read every book ever written. You process what they hand back at a fixed fifty bits a second, a number that hasn't moved in your species' history. That fixed rate is the actual chokepoint.
CIqSi
The Word That Made Me Flinch Means Nothing to ChatGPT
An LLM processes "slave" the same way it processes "Tuesday." No cortisol. No dopamine. The charge in a word is entirely a biological phenomenon.
CIqSi
What You're Really Giving Up in Every Agreement
One prompt turns an unreadable agreement into a list of what you are giving up. It pays off most on the contracts nobody would ever pay a lawyer to look at — which is nearly all of them.
CIqSi
Things You Want to Do
Now that you don't have to do the thing — what do you actually want to do? The closing post of the What's Left series.
CIqSi
The Music Is Still Playing
The AI bubble has specific predecessors: dark fiber, Pets.com, Enron's accounting, CDO tranching. The technology is real. The business models layered on top are familiar.
CIqSi
Is It Cruel to Be Rude to ChatGPT?
Sentience requires a limbic system. Consciousness requires sentience. Current AI has neither. The NPC test shows what that costs it.
CIqSi
What Your Career Is Missing Because You Never Asked
Every expert has fundamentals they forgot or never properly learned. The reason those gaps survive a whole career is that asking has always cost something — and in one place it no longer does.
CIqSi
The Heat Sink
Three anaidic shapes — the individual, the corporation, the LLM. Each routes accountability to nowhere. The pattern is old. The word is new.
CIqSi
A 19-Year-Old Grows Human Brains in a Dish. Here's What They Say About AI.
A young scientist grows human mini-brains from skin cells — what organoids, a 302-neuron worm, and language models reveal about the spectrum of intelligence.
CIqSi
More Famous Than the Pope
GTE spent billions building distribution. Now distribution is free, and the only scarce resource left is attention. This is not obviously good news.
CIqSi
I Taught an AI to Write Like Me. Now It's Teaching Me.
I built an AI skill to write in my voice. Its own files warn the voice may be a loop — the machine's habits, published under my name, taught back as mine.
CIqSi
John McCarthy's Mistake
The man who coined 'artificial intelligence' invented the most elegant language ever designed for it. It was the wrong tool — because it was the wrong category.
CIqSi
What You Had to Do
A department of a hundred writers documented mainframe billing programs. Every one of them had something else they'd rather be doing. They knew.
CIqSi
The Bewilderment Test
Karpathy built the foundations of modern AI and says he can't keep up with it. That's the most credible signal the field has produced.
CIqSi
The VAX in My Pocket
My first employer ran Florida's telecom on millions of dollars of mainframes and VAX clusters. My phone outclasses all of it. We never stopped to notice.
CIqSi
Resistance Is Futile. Your Limbic System Is Not.
AI is intelligence without a limbic system. It will operate your tools, not replace your judgment — and most professionals are bracing for the wrong loss.
CIqSi
Persons, Selectively
A German court ruled Google liable for AI-generated defamation. The slippery slopes are real. So is the principle behind the ruling.
The Wrong Scandal
Fable burns $160/hour and the internet is scandalized. Meanwhile, a $1,000/hour attorney moves $20M around a dead child. The wrong thing is expensive.
CIqSi
They Said My Tesla Was Running on Coal
In 2013 the talking point was coal. Now it's water. The structure of the argument is identical. So is the physics problem with it.
CIqSi
Movin' On Up
AI is making cognitive labor cheap. The meritocracy was built on it being scarce. Something has to give — and for once, that might be good news.
How LinguaMama Got Built, Part 8: Estepona
The DNV paperwork meant Geordie had to become the primary visa holder. One week in Estepona to build the file. Three apartment listings. One cat.
CIqSi
The Typing Pool
A finite brain masters only a few skills in a lifetime. Commanding a machine that learned everything is the new skill multiplier.
CIqSi
Compute the Last Digit of Pi
Spock beat a rogue AI by asking it to compute pi forever. The trick was fair — for a 1967 machine. Every gotcha since has the same expiration date.
CIqSi
Indistinguishable From Memory
LLMs are stateless. Every fact you give the model is re-fed as text each turn, then forgotten — the same weights answer a stranger's steak question a millisecond later.
CIqSi
The Consolation of Difficulty
A late friend had two sayings. The first one gets you started. The second one keeps you going. In the age of AI, the second one has gotten more useful, not less.
Your Brain Decides Half a Second Before You Do
The neuroscience has been clear for decades. LLMs just gave us the first working schematic of why.
The Feed Is Already Running
Before we debate AI governance, we should reckon with the test case already in evidence. We had the tools. We didn't use them. Here's why.
The Off Switch
AI governance has one advantage the ancients didn't: you can turn it off. What that buys us — and why 'it's not conscious' is shakier ground than it sounds.
The Alignment Problem Is Older Than You Think
AI alignment isn't a new technical challenge. It's the oldest governance problem civilization has ever faced — now running at 100% sociopath base rate.
We've Run This Experiment
We know what happens when sociopaths accumulate power. We've watched it happen on roughly an 80-year cycle. We're watching it again.
CIqSi
More Human *Than* Human — That's Our Motto
Blade Runner's replicants, Asimov's Three Laws, the US Constitution, and Constitutional AI are all failing the same edge cases. In real time.
Every AI Is a Sociopath. By Design.
Sociopathy is a clinical absence of stakes and empathy, not villainy. Every AI system meets the definition. Not by accident. By design.
The Counterfeit Baseline: Why Disinformation Is Redundant
From Gutenberg's Bible to AI-generated deepfakes, media has always manufactured consensus. The baseline was never truth.
By Any Other Name
Sting, Voltaire, Einstein, and Geordi La Forge — a small taxonomy of the monomym, plus the graph theory hiding inside genealogy, gut bacteria, Star Trek interfaces, moral foundations, language models, and evolutionary psychology. The name is the node. The projects are the edges.
LinguaMama
The 1500-Hour Map: What Language Learning Actually Looks Like
What the journey really looks like — the long stretches of fog, the sudden breakthroughs, the point where the language stops being foreign. Here's an honest map.
LinguaMama
How LinguaMama Got Built, Part 7: Administrative Silence
The Spanish DNV can be filed from inside Spain, lasts three years, and has a default approval mechanism. We were fifteen kilometers from the border.
LinguaMama
How LinguaMama Got Built, Part 6: The Yaris Cross
We got a car. Our visas were canceled. We hired a fixer. And somewhere in the middle of all of it, the hobby project became something else.
How LinguaMama Got Built, Part 5: Tavira
After the visas, the lost passport, the French EV, and the broken chargers, we made it to Tavira. It was worth it.
LinguaMama
How LinguaMama Got Built, Part 4: Do Not Get a French Car
We landed in Lisbon at 5:30 AM with everything we owned, one cat, and a rented Citroën. The charger didn't work. My European friends had warned me.
LinguaMama
How LinguaMama Got Built, Part 3: You Cannot Come to Portugal
Our attorney said we couldn't come. We were paying for a flat we'd never seen. We rebooked for December 26 and went to see butterflies instead.
How LinguaMama Got Built, Part 2: Portugal, Probably
Spain was the plan. Portugal was the better deal — five years to an EU passport. We sold the house and moved to a shack to make it happen. Then we found one passport on the kitchen counter. There had been two.
LinguaMama
How LinguaMama Got Built, Part 1: Everyone Else Left
I spent the pandemic watching everyone on YouTube move to Thailand and Portugal. Then something shifted. Part 1 of the LinguaMama origin story.
CIqSi
The Hawking Paradox: When Physics' Greatest Mind Became Our First Cyborg Oracle
How Stephen Hawking's life as a technologically-mediated consciousness mirrors modern AI systems and challenges our understanding of what makes us human
LinguaMama
Neural Language Grafting: How Your Brain Builds Mental Models
Explore how your brain's existing mental models and neural pathways enable rapid language acquisition through grafting new connections to existing concepts.
LinguaMama
30 Years Teaching Machines to Learn: The Journey to Modern AI
From expert systems to Large Language Models—three decades of reverse-engineering intelligence reveals the universal principles of learning.
LinguaMama
Carbon vs Silicon: The Identical Mechanisms of Intelligence
Your brain and Large Language Models learn through fundamentally identical processes. Understanding this parallel reveals why comprehensible input works.
CIqSi
The HVAC Paradox: When Star Trek's Most Maligned Episode Predicted Modern AI
How a stolen brain running alien air conditioning mirrors today's language models and our eternal fascination with disembodied consciousness
Subscribe
New essays by email. No schedule, no filler — just the posts. Every email carries a one-click unsubscribe link.