📖 Tesseract Book Club · 🏆 high-score board
This is the 2026-10-30 edition as it was sent — published because readers voted for it with their clicks.
Somebody will defend you this week with the word deterministic. The word will work, and it will not hold.͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ ͏‌ 
The Stage Floor Principle · Preface · ¶1–11 · 2026-10-30
If it's debatable, it's not insurable.

The difference between what a system says it is doing and what it is doing has weight. That gap is where every AI failure and every uninsurable liability lives, and the gap can be measured, so it can be priced. Read the other way, the same measurement is personal: you are not about to be averaged out by a generalist.

Next time you wonder how to help, tap the sticker. The sticker carries the word to the next signer.

A word from a peer gets read. The same word from a vendor gets filed.

▶ Hear the case in four minutes (3:50) → Take 83, recorded on my phone. Rooms that want the talk book it at thetadriven.com/speaker.

I was offered a few minutes at the Royal Institution on 10 November, in the Faraday Lecture Theatre, and I am grateful for it: recognition is exactly what this work needs right now. One evening does not carry a transatlantic flight, so my seat is pending. If you know a room in London that week that would hear the talk, forward this email to whoever runs it. One more room and I am on the plane.

Before the book
AI doesn't get past the pilot stage until someone can insure it.

The market agrees. From January 1, 2026, the standard liability forms let insurers exclude generative AI outright (ISO CG 40 47, CG 40 48, CG 35 08). Insurers won't carry what they can't count. We sell the count.

No highways before insurance: uninsured roads never get built, and the cars never improve.

If you sign for an agent: Whether it was good is undecidable (Rice, 1953). What it actually did cannot be reconstructed from its own account (the data processing inequality). The only remaining move is to read a record the actor did not write.

You would be forgiven for calling an answer to that impossible. What we did, and how it holds, follow the book.

Tesseract Physics · Daily Book Club · 2026-10-30
Grip.

Ask the insurance industry why they will not insure AI.

Jump as high as you can; nobody jumps off sand. An AI budget is that jump, standing on sand: the deployer cannot fund a rollout nobody can price, and the underwriter cannot price what nobody measured, so the same deal stalls on both desks. The sovereign asks whether: a licence, a pause. The market asks what: where the work landed against what was declared, written down as a receipt anyone can check. Only that can be priced.

Send this to the next person who says it

Your strongest engineer will defend your AI this week by calling it deterministic, to reassure the room. The word will work, and everyone will move on.

“Our systems are deterministic.”

They are. Completely. Pin the weights, the seed, the sampler and the batch — the same output, every run. Every step obeys; where the steps end up is undecidable (Rice, 1953). Determinism is undecidable chaos. Determinism belongs to the replay. Autonomy is the part where nobody fixed the input. You can replay yesterday. Nobody has ever replayed tomorrow.

An agent’s loss lands on the company. Whether oversight was even possible lands on you (EU AI Act, Article 14).

Today’s chapter is the same mistake, made by a whole field for sixty years.

Why this works: Preface of Tesseract Physics, below.
Preface: The Splinter in Your Mind

The Stage Floor Principle

The fear is real. You read the words Zero-Entropy Control and absolute verification, and the hair on your neck stands up. The picture arrives on its own: a panopticon logging every keystroke, or a prompt chain frozen so stiff that your engineers spend the quarter asking permission to change one word. You are right to flinch. In the real world, ambiguity is often a feature. The CEO needs wiggle room. The diplomat needs constructive ambiguity. The engineer needs room to try a thing and be wrong, and the human needs privacy.

If control meant taking away the room to breathe, the fear would be the correct reading of this book.

We distinguish between the Floor and the Play.

Every trade that carries a load drew that line first. The boiler inspector never told the engine driver where to run the train; the inspector rated the plate so the pressure could climb. The Lloyd's surveyor never chose the cargo; the surveyor graded the hull so the merchant could insure the voyage. A floor joist carries a stamped load rating for the same reason: the rating stays rigid so the room above can be anything.

S≡P≡H draws the same line through the machine. S≡P≡H does not demand that humans stop telling stories. S≡P≡H demands that the physics stops lying about where the ground is. We eliminate structural ambiguity, not social ambiguity: the meaning you negotiate stays yours, and the address that meaning sits at stops moving.

Think of a theater. You want the stage floor absolute, rigid, verifiable, oak on iron joists that hold 10,000 pounds without a creak. The floor holds still and leaves the blocking to the director, so that the actors can be free to perform.

Now put the same cast on boards nobody rated. If the actors spend 40% of their energy checking whether the floorboards are rotten, they cannot perform the play. The risky line gets cut, the leap shrinks to a shuffle, and the cast turns anxious, reactive, exhausted. Your teams pay that tax today, standing on sand and waiting for the one uninsurable turn that drops them through.

The violin strings must be under absolute tension so that the music can fly; a slack string makes no sound at all. Tension is what the music stands on, and the substrate is what the agent stands on.

Constrain the Substrate. Free the Agent.

We are not here to police your culture, your stories or your people. We are here to fix the floor.

Read this in the book, in context →

Meta vector — what this section's idea rests on, and what rests on it.

🟣E9🎨 Qualia (The Redness of Red) - P=1 Structural Certainty ↘ — what defines it
  9 🟣E7🔌 Hebbian Learning — creates P=1 structure
    9 🟢C1🏗️ Unity Principle — S=P=H theoretical foundation
  9 🟣E8💪 Long-Term Potentiation (LTP) — physical mechanism
  8 🟣E5a✨ Precision Collision — clean signal

🟣E9🎨 Qualia (The Redness of Red) - P=1 Structural Certainty ↗ — what it causes
  9 🟣E4🧠 Consciousness Proof — qualia validates consciousness
    9 🟣E5💡 The Flip — subjective validation
  8 🟣E5a✨ Precision Collision — enables insights
  7 🔵A1⚛️ Landauer's Principle — thermodynamic foundation

Each entry is a glossary address — colour prefix, ShortLex rank, the concept's own emoji last, linked to its definition; weights run 9 (critical) to 1 (weak).
Refined for Book Club 2026-10-30 — "The Stage Floor Principle".
An unsigned testimonial

“I am the one holding the bag when our AI breaks a client contract. Every change our agents make now carries a receipt the machine did not write, one my carrier can recompute.”

— ______________________________

Would you put your name on this? Would your boss? Would anyone you know?

Ladder rung 2, steering payload footprint per tick: how small the steering payload is beside one context window growing toward a million tokens. The extension draws this chart from your own runs. It measures the tree/snowball mechanism, not the /steer word.
Ladder rung 2 vs the worst case, one context window growing to 1M tokens: 99% smaller on 850 of 1146 ticks; the best 20% at 99.85%.
What this represents

An agent’s log is words about words, the loop Stevan Harnad named in 1990. Each word points to more words; the chain branches without end and never reaches ground. The field took that runaway as fate. A runaway is only fatal in an open space. Close the space, so every word has a place and nothing is left outside, and the same branching turns into confirmation: many independent paths agreeing on one place. A nerve cell decides the same way inside its membrane: one input proves nothing, and many independent inputs, summed, carry the call. We have been closing that space since the summer of 2000. How the floor is built, and the one condition it rests on, follow further down.

Further down: a photograph of the physical piece and the checks you can run yourself.

What we did
For all practical purposes, we solved the symbol grounding problem: your specification and your agent’s work become two shapes on one map, and the drift is the difference you can see.

A word you never declared has no address, and the receipt says so instead of guessing.

Two ways to prove us wrong, both public: find a definition on the floor that does not stop, or a change written in your own specification's terms that lands outside its lane. Measured: the walk returns the same bytes on a Mac and on Linux, Intel and Arm alike, in an open cross-architecture check anyone can rerun (IntentGuard, tests/cross_arch.rs).

The mathematical proof and the empirical runs follow further down.

We detect, place, price and dispatch. You hold the wheel: the halt is yours to wire. iamfim.com

ThetaDriven crest
Intermission
The long version of the blade

They were being precise. Not loose. Precise.

Now read their own sentence back and notice what it says. The sentence says the machine repeats itself. The sentence does not say the machine does what you asked. Anyone who has written a line of code has watched a deterministic program crash, loop forever, corrupt the table and halt on the wrong branch — identically, every run, exactly as designed and nothing like intended. Determinism has never once meant a system is under control. Determinism means the system does the exact same thing twice, given the exact same input. Change that input by the width of a butterfly’s wing and a deterministic system can end somewhere else entirely: Lorenz showed it in 1963 with a weather model that was fully deterministic and, past a horizon, unpredictable in principle. Deterministic chaos is still determinism.

And there is a smaller sentence hiding inside the big one: a deterministic autonomous agent. Its input is the morning — an inbox that moved, a file rewritten under it, a tool that answered differently than it did on Tuesday.

When you hear those two words in the same breath, pay attention. Those two words are the most reliable tell in the room that somebody has been asked for a guarantee nobody can give and has reached for the nearest word that sounds like one. The speaker is usually the strongest engineer in the building, which is precisely why the sentence goes unchallenged.

The computing sense is the version you can prove with your logs. The English sense — predictable, under control, will do what we intended — is the version the audit committee and your regulator are reading out of those same logs.

The long form: The two determinisms · the river is the prompt. Determinism reassures a room rather than confusing it — and that is what makes it the expensive word.

You just read where the eviction is held: in the slot, at the clock of the write. Your logs were written afterwards, by the thing they are about. That is where the law starts.

The question underneath all of this

Am I personally liable for what my agents did?

The financial loss lands on the company. The question of whether effective oversight was even possible lands on you.

That phrase — effective human oversight — is the one the EU AI Act turns on (Article 14). The Act lists what the overseer must be able to do; it draws no line for when that bar is met. The line will be drawn the way every such line is drawn: by the first people who have to argue about one. There are two roads out and no third.

If effective oversight was possible and you did not have it, that is the easy case. Your insurer declines for negligence.

If effective oversight was impossible — if nothing you owned could have told you where that agent went — then you were never in a position to be responsible for it. And an owner who cannot be responsible is not given command of an asset. The liability has your name on it.

You are the captain. The crew is autonomous. No underwriter in three hundred years has asked a captain for a better crew. Underwriters ask for the logbook — the record the crew does not get to write after the fact.

The sentence to hand your risk officer: “An agent incident is any action outside the lane declared before the run, counted from a record the agent did not write.”

The measurement is free. The sealed logbook costs $20 a year per agent. Seal it, or hold the liability unrecorded. Buy the stamp →

What nobody is checking — and why nobody else has an answer

You already have a control environment, and it is not a bad one. The prompts are versioned. The models are pinned. The eval suite runs in CI and goes red when it should. Every action the agent takes is written down with a timestamp and retained for seven years. Ask the industry what more you could possibly do and you get a longer version of that same list.

There is one property that list does not have, and doing the list better will never add it. Every record on it was produced by the process it describes. The agent reports its own actions. The eval was written by the team that wrote the thing it grades. An account of a run, produced by that run, cannot contain what the run never noticed — and that is not a claim about anybody’s honesty. The claim is about where the account came from.

Which is an old problem with an old answer. Double-entry bookkeeping (1494) never asked a merchant to be honest — it made the books check themselves, every debit against a credit. The audit that grew up around it added the part that matters here: a second hand that did not write the first entry. That requirement is why a stranger will lend you money. Your AI keeps one book. So does everyone’s.

The industry’s answer is to build better first entries: more evals, more guardrails, another model watching the first one — each of them written from inside the same boundary as the thing it watches. We went the other way and put the instrument on the far side of it, so the record is one the agent did not author. That is the entire difference, and it is why the claims that follow sound outlandish coming from a small shop. The gap is not cleverness. The gap is where the instrument stands.

And if it is wrong, it is wrong in one specific place, so here is the sentence to swing at. Name one record in your stack that was not produced by the thing it is about. Reply with it and you have taken this apart in public — and we will publish that you did.

What shipped today — the work this email rides on
"The Stage Floor Principle" was edited today, the same day it reaches you: the Stage Floor Principle earns its seat in the 10-30 club send.
The rest of the day went into the work around it: A1 admits the Preface label the renderer builds; the What-shipped-today writer speaks in the email's voice.
Drift receipt panel for today's headline commit
The drift receipt of the commit that refined today's passage — where the work landed on the 144-zone map, circled where it strayed.
How the picture is painted
Every square on the grid is a crossing: one row meeting one column. The walk starts at the square your declaration names and runs along its row to the eleven squares that share it, then turns down each of their columns, then along those squares' rows again — row, column, row, column — each step counting for less than the one before, so near squares glow and far ones fade. Run it on what you declared and you get the first picture. Run it on the code that shipped and you get the second. Same arithmetic, same bytes, the same picture every time you rerun it. Each commit's panel is kept with a signed row on the walk tape naming the commit, the hash of the text walked, and the hash of the exact binary that walked it — so the picture can be redrawn and compared, byte for byte. That is how the walk stops the dictionary problem — every word defined by more words — in practice: a meaning ends at an address on the grid, not at another word. Where the two clouds disagree is the red in the commit panel further down. The walk is open source, MIT, and runs on your own machine: github.com/wiber/thetacog-mcp — the README carries all six panels.
The walk on the declaration — the intent cloud on the 144-square lattice
The walk on what you declared
The walk on the code — the reality cloud on the same lattice
The walk on the code that shipped
The Meta vector under the passage is this walk written out in words, one passage at a time. The gear is the same at every step: start at one idea, step to what it rests on (↓) and to what rests on it (↑), and count each step for less, 9 down to 1 in the words, the fade from bright to dim in the pictures. The panels run that move from every word, by machine, and draw the sum; the Meta vector names the first steps from one passage. One difference, said plainly: the Meta vector’s steps come from the book’s glossary, chosen by hand; the panels’ steps are computed by the walk.
ThetaDriven
Intermission
Twenty minutes ago you asked an agent for one CSS class.
One night an agent re-read this whole repo to change a two-line import, and one afternoon it touched a file I never asked it to touch while CI stayed green. I built /steer so my agents boot cold from a row I wrote before they ran, and leave a second entry they did not write.
— Elias Moosman
Steer it.
Six seconds: the diff counter climbs to fourteen files in red — Steer it. — /steer typed, the agent line, then the Δ line placed and the Copy Run Summary door
ThetaCog Steer — the TD shield, the extension's logo in the Marketplace
Now · ThetaCog Steer · on the VS Code Marketplace
Coding with agents is flying. Flight needs a logbook the pilot did not write.
Double-entry bookkeeping for autonomy, two hands: Steer, free and open-source (MIT), declares the lane in your editor; IntentGuard places each production turn against it, beside the deployed bot. In VS Code press ⇧⌘X (Ctrl+Shift+X) and search exactly ThetaCog Steer — Review it on the Marketplace → · other editors: thetadriven.com/vsx → The VS Code Marketplace listing: ThetaCog Steer: Coding Is Now Flying — with Double-Entry Bookkeeping for Autonomy — ThetaDriven — Free — Install

The install and the measurement are free and open-source (MIT); only the financialization/attestation layer is licensed. What a licence buys: the right to run the receipt on an agent in production under US Patent Application 19/637,714. One licence per agent, $20, good for 365 days or 10,000 countersigned receipts, whichever comes first. Each receipt records where that agent's work landed against what it was asked to do, and anyone can recompute it.

The sceptic’s chart
Ladder rung 1: every call charged its whole context against a full million-token window, one point per tick, 2026-08-29 to 2026-10-02, the 99 percent line drawn above the series
Charge every call its whole context against a full million-token window. On 962 ticks it crosses 99 % zero times; the best fifth sits at 90.4 %. The line still climbs. The median tick went from 69 % in its first week to 90 % in its last, because the context each call carries fell from about 309k tokens to 97k. Rank correlation with time ρ = 0.62; on one point per day (53 days, the honest unit, since neighbouring ticks move together) the permutation p is 0. Small, and real. The story we would tell: once the spec carried what the corpus used to, a call stopped re-reading the whole corpus. The story is a story, not a proof — nobody randomised these ticks.
So here is the second ledger

Insurance never prevented a single fire; it made fire a number. Your logs already say something happened — what they do not say is whether it was allowed.

1 — The open source is free. A standard format for logging what your agents did, MIT licensed, running on your own machine. Declare the lane before the work starts. Nothing leaves the building. The full book text rides in the same repo (data/book/COMPLETE-BOOK.txt). You can do this today, alone, and never speak to us. github.com/wiber/thetacog-mcp
+1 — The stamp is $20 per agent-year. The stamp seals the log so it can be handed to someone else. A log your underwriter cannot verify is not evidence — it is your word for it.
= The receipt. One half is free and yours. The other half is what makes it somebody else’s evidence.
Install it, then keep flying
A carrier has never priced a meal by inspecting the kitchen; it prices what leaves the pass. Nobody sues over a branch on a laptop at 2 AM. The claim arrives when a deployed bot acts on a customer's data, so that is where the second entry has to be written. There are two halves.
1. The drafting table: ThetaCog Steer, in VS Code. Press ⇧⌘X, search ThetaCog Steer, install, and go back to work. Your agents start from a spec row you declared before they ran, and every commit lands on a receipt against that row, recomputable by anyone you hand it to.
2. The flight recorder: IntentGuard, beside the deployed bot. The same walk, carved out as a library with a Node addon, built for Linux as well as the Mac. Put it in the server that runs your wrapper and each turn is placed against the spec you committed, never one the bot wrote. The Linux build and the Mac build give the same card for the same bytes, and that is a check you run yourself: the source is on GitHub →
The countersigned card is the paid half, and it works the way a telematics box works in a car: the box has never stopped anybody leaving the road, and it is still the thing that moved the premium. The card records where each turn landed, signed by a key that is not yours and kept where you cannot edit it, so a carrier can recompute the record without taking your word for it. The halt stays yours to wire.
The backup curve on thetadriven.com: tokens in one task, the tail an underwriter prices, scope declared before the task against no declared scope, with the grey band where the declared curve would fall if declaring made no difference
The typical task is the same size either way. The rare one is where the spend goes, and what a carrier prices. ThetaCog Steer draws this from your own runs →
Commit Panel of the Day
And here is today's receipt — the one behind the actual work that shipped in this repo today, not a mockup: chore(commit-page): panel-repair bbe59c55f — re-grounded panels [panel-attached] (2026-09-08). GREEN is in lane. AMBER is the bleed, one square out. RED is out of lane, and the circles mark where it clustered. The lane is a line somebody declared before the run; the placement is what the walk measured; the colour is only the two meeting. You do not have to read the math. You have to notice that the square where it left its lane has an address, and an address is countable — a count is the frequency half of a price, and the lane you declared is the severity half.
Commit tolerance panel — chore(commit-page): panel-repair bbe59c55f — re-grounded panels [panel-attached]

Open this commit's attestation — verify it yourself →

Sidenote: when a commit lands OUT of its lane, the receipt triggers extra work automatically — a sensemaking pass explains the drift, the fixes get checklisted, and the intervention is published. Every out-of-lane receipt is a countable event; that count is what makes this priceable.
From the last 24 hours of the ledger, verbatim: “Room attribution runs on path-prefix overlap before it checks the commit's own Relevant-Rooms trailer, so a single-token touch to a generated sibling directory under a room's owned prefix can outweigh ninety-plus lines of correctly-attribut”
The verdict

Rice’s theorem says nobody can prove your agent is good. We have never claimed to. We made where it went countable instead — and there are two doors on the next page, only one of them yours.

Yes. A decidable, still-semantic spec boundary for every autonomous deterministic AI agent.

And because this is the business of AI: buy at iamfim.com, and sleep like a baby. A baby sleeps because someone else keeps the night watch. Here the watch is a record your agent did not write: the lane is set before the run, the landing is checked after it, and your underwriter reads the record instead of your word. For the discerning sceptic — fork github.com/wiber/thetacog-mcp, run it on your last ten commits, and see what your agents touched in far fewer than twenty-one questions.

What does this mean for me? →
iamfim.com — four seconds to know which one you are.
We published three of our own bugs and left them open — fork one, fix it, and that is the entire application: thetadriven.com/hiring
Every time a new measurement appeared, a new market opened

Markets do not underwrite safety. Markets underwrite what a stranger can check. The count is live.

1494  books: double-entry ledger (balance, not honesty) → banking became possible
1764  ships: Lloyd’s Register grade (condition, not seaworthiness) → cargo underwritten
1866  boilers: the inspector’s pass (fit to the rule, never sworn safe) → factories financed
1903  cars: the driver’s licence (a bound operator) → the road opened
2026  agents: ????????????????? → exclusions written
Not one of those certified the thing was good. Double entry never asked a merchant to be honest. The boiler inspection never promised the boiler would not explode. Each time the market found the checkable half, and each time the money arrived the day after. If it’s debatable, it’s not insurable.
And in 1932 the deadline arrived for everyone who had waited. Two barges went down in a storm their tugs would have dodged with a weather radio almost no tug then carried. Judge Learned Hand refused the industry-custom defence in The T.J. Hooper: a whole calling may have unduly lagged in the adoption of new and available devices. Custom is not care.
If you wonder whether the math is real, here it is

The question began in a conversation in the summer of 2000; the algorithm followed in 2007 and 2008.

Why a branching chain settles here instead of running away. In an open dictionary every word points to more words and the chain never ends. On the floor every word has a place and the walk only steps forward, so every walk ends: the matrix it runs on is nilpotent. The branching stays. Each place has twenty-two neighbours, and the floor branches far above one. As commits accrue, more and more paths confirm each landing: a cascade past its threshold, the dichotomy Galton and Watson proved for family names. A cascade proves something only if the paths are independent; a million copies of one path confirm one thing once. Inside each block of the floor, children are ranked by their parent’s rule. Where that rule is independent of the neighbours’ (correlation under a tenth), the ranking is a new axis. The book calls that part the resonator. A nerve cell does the same inside its membrane: weak, independent inputs, summed, carry a decision no single input could. Independence is the condition, and the book puts it under test in public rather than assuming it.

On this repository: the latest 150 of 29,694 commits, walked twice on one machine, rerun distance 0. None of the 150 is long enough for the walk to admit, so whether meaning moves a placement is still unmeasured, and we say so here.

The keyring in front of the printed 12 by 12 grid: a flat slab with one row of small blocks and one column of small blocks along two of its edges R ≡ (Σᵢ |Aᵢ|)ⁿ − Σ over k₁+⋯+kₘ = n of the multinomial (n; k₁…kₘ) · Πᵢ |Aᵢ|^kᵢ ≡ 0
R = (Σᵢ Aᵢ)² − (squares + crossings) ≡ 0
What it does: each lane is one Aᵢ. The squares (a², b²) are where one lane works alone. The cross terms (2ab) are the named crossings where two lanes meet. Add them up and the remainder is zero: no corner of the space belongs to nobody. So wherever a piece of work lands, it lands at an address, inside a lane or on a crossing you can name. That makes where it went countable. The count says nothing about whether the work was good.
(a + b)² = a² + 2ab + b² is the smallest case, and you can hold it: the slab, the row, the column, the corner. The 144×144 floor is the same identity with more lanes: 20,736 named places, each inside exactly one term. The preface walks it through, and names the one flag that would make it wrong. Call it conservation of address: every place inside exactly one term, the remainder zero. Conservation of address never looks at what a place holds. The patent (US 19/637,714, Track One) claims the machine that uses it, never the law.
Latest from the blog
Nothing new went up in the last day, so here are the 3 most recent essays — what each one is for, what we are least sure of in it, and the question we would most like answered back. Reply to this email with any of them — the reply reaches Elias, not a funnel.
Independent STILL Means A Different Generator · 2026-10-27
Takeaway. In April we argued the EU AI Act was written to be impossible in software, and 102 comments argued back.
Our note. We published "Independent STILL Means A Different Generator" on 2026-10-27, straight out of the work it came from (Positioning). The post opens: "You get a record of what your agent did that your agent did not generate. Each change the agent makes is placed against the job you wrote down before it ran, anyone can re-run the placement and get the same bytes, and ma…" Read it as a working draft: if the argument breaks somewhere, that break is the useful part and we want it back.
We want your answer: Which sentence in "Independent STILL Means A Different Generator" would you strike first, and what would you put in its place?
What Arrives If We Are Right · 2026-10-23
Takeaway. Grounding belongs to the act, not the description: agent work is real only where it lands at an address another hand can check.
Our note. We published "What Arrives If We Are Right" on 2026-10-23, straight out of the work it came from (Positioning). The post opens: "You pay for agent work by the token, which is a lamp priced by its glow. We sell the other price: a record of where your agent's work landed, which anyone can re-run. The instrument is a free, open-source command you run…" Read it as a working draft: if the argument breaks somewhere, that break is the useful part and we want it back.
We want your answer: Which sentence in "What Arrives If We Are Right" would you strike first, and what would you put in its place?
Nobody Writes A Rule Against Slapping The Hand · 2026-10-23
Takeaway. No rule list names every move an agent will make, and since January standard liability forms carry a generative-AI exclusion: the loss from an unwritten step lands on whoever signed for the run.
Our note. We published "Nobody Writes A Rule Against Slapping The Hand" on 2026-10-23, straight out of the work it came from (Positioning). The post opens: "Your liability policy now carries a generative-AI exclusion. When an AI agent does something nobody told it not to, the loss is yours if you signed for it, and no rule list can name every such move in advance. What you c…" Read it as a working draft: if the argument breaks somewhere, that break is the useful part and we want it back.
We want your answer: Which sentence in "Nobody Writes A Rule Against Slapping The Hand" would you strike first, and what would you put in its place?
A black die-cut sticker: If it's debatable, it's not insurable.
The whole argument, on a disc that fits on a laptop lid. The disc is going to print.
The second sticker: What is good is yours. AI doesn't get to sell it.
We published three of our own bugs and left them open. Fork one, fix it — that is the entire application. No CV.
ThetaDriven
Forward this to your lead — three lines for #engineering:
ThetaCog Steer — VS Code ⇧⌘X, search "ThetaCog Steer" (id thetadriven.thetacog-mcp, free)
One thing it does: a worker boots from the spec row instead of the chat, and the commit gets a second entry the agent did not write.
https://marketplace.visualstudio.com/items?itemName=thetadriven.thetacog-mcp
The physics is not in this email. The physics is a book, and you have to ask for it — iamfim.com.

Do you worry about $1.2B in AI liability?

If the property is trivial, software can check it — and why are you paying to check trivial properties? If it isn’t trivial, Rice’s theorem says nobody can. So we fixed the math.

type your number — we call you →

The one thing that actually helps: hit reply. A reply is the only move a mail provider counts as a human being — it is what puts the next one in your inbox instead of a tab. Type anything: a correction, an idea, one word.

And if you ever engage with something like this, think of this newsletter next time.

🏆 The high-score board — you voted with your clicks
Ranked by pages readers opened in their own browsers — not by mailbox pings. If you missed one of these, the room read it for a reason. The order on this board is the room’s and not ours — what gets opened is what aims the next edition, and you can watch it move.
🔥🔥🔥🔥 Chapter 10: Natural Experiments — When Humans Beat the Metrics · Ch 10 · ¶1–9 · 2026-10-12
The full board →
📡 The Signal — who started asking for the receipt
We don't curate AI news. We magnetise the exact moment the world reaches for what we built — placed on the same lattice as the panel above, ranked by who's screaming loudest. A general curator can't send this.
1 · A jurisdiction made our record the law source →
Who's screaming: The European Union.  ·  Why it's us: This is a jurisdiction aligned with paying — it does not muse about a gap, it mandates the purchase and prices the exposure: the conformity record is the audit trail insurers underwrite against. The only open question is whether that record is a software log (which can be edited to claim the agent stayed in its lane) or hardware-attested and recomputable (which cannot). We are the second kind — the decidable, tamper-evident receipt the law now compels.
This is happening now: the audit trail is no longer optional in the EU — it is a market-access and insurability requirement as of 2 August 2026. Give yours an entry it didn't write: install ThetaCog Steer in VS Code. Forward to whoever owns your EU AI Act readiness and your AI liability.
2 · A startup is racing to build the software version
Who's screaming: The market.  ·  Why it's us: The evidence layer is now the product. The only question left is whether your evidence can lie about its own state (a software log) or can't (hardware-attested, recomputable) — the line we're on.
This is happening now: the category is forming around exactly what we built. Be on the side that can't be faked — forward to anyone building AI audit.
3 · Humanoids hit the factory floor with no priced liability source →
Who's screaming: An unpriced factory floor.  ·  Why it's us: A humanoid acting autonomously is an uninsured liability surface until someone can prove, per action, that it stayed in its lane. That proof is the receipt.
This is happening now: autonomous agents are already on payroll with no priced boundary. The boundary is decidable — forward to anyone deploying physical AI.
You're getting the Tesseract Physics Book Club because your address is in our circle. One passage a day, chosen for what's live right now.
thetadriven.com/book
ThetaDriven · Elias Moosman · elias@thetadriven.com
Running a tool-calling agent unattended? Read what we warrant on our own act, then install from the Marketplace.