Every Question Is a Physical Act

Summary of Anderson (2026), Observation IS Protection: A First-Principles Derivation from

Information Thermodynamics. Full paper with 34 references available upon request.

The Simplest Truth in Environmental Protection A bit is an answer to a yes/no question. That is all a bit is. Not a computer term. The smallest irreducible piece of information in the physical universe. Claude Shannon defined it in 1948. John Archibald Wheeler built his entire theory of physics on it:

“What we call reality arises in the last analysis from the posing of yes-no questions.”

String together answers to lots of yes/no questions and you have an environmental protection system. String together lots and lots and lots of yes/no questions and you have a universe.

That sounds too simple. It is not. There is a depth beneath this simplicity that is genuinely difficult to grasp. But the finding itself, at its deepest depth, is simple:

Every question asked about the environment—whose answer reaches the gate—is a physical act of protection.

Not “enables” protection. Not “correlates with” protection. Is protection. This is derived from experimentally verified physics. Zero unverified assumptions. And it has been sitting in the equations, unseen, for sixteen years.

Why a Question Is Physical A question is physical in three experimentally verified ways. It costs energy—Landauer (1961) proved that any computational act dissipates at minimum kBT ln(2) ≈

2.87 × 10⁻²¹ joules per bit, confirmed by Bérut (Nature, 2012). Its answer has thermodynamic value—Sagawa and Ueda (2008) proved that mutual information gained through measurement enables work extraction of kBT per bit, confirmed by

Toyabe (Nature Physics, 2010) and Koski (PNAS, 2014).

And it completes a chain that is already built and waiting. A valve mechanism already installed. A signal path already wired. A schedule already operational. Natural attenuation pathways already running. All idle—until the question is asked and its answer reaches the gate. The question is the only missing piece. When it arrives, the valve closes, the schedule adjusts, the pathway activates. Everything else was already there.

One honest limit: where no gate exists—no valve, no pathway, no infrastructure—questions identify what needs to be built, but the building still costs real energy. The claim holds where the infrastructure already exists. For most industrial facilities and for all of nature’s own processes, it does.

A question costs joules, produces joules, and moves matter.

It is as physical as a wrench. It is ten billion times cheaper.

Everything Is Already in Motion The universe has more ways to be disordered than to be ordered. Overwhelmingly more. That is the Second Law of Thermodynamics—not a law about energy, but a law about counting. Any system left alone drifts toward disorder, not because disorder is powerful, but because disorder is numerous.

Imagine a building with ten million rooms. One contains what you need. The rest are empty. Without knowing which room, you wander and never find it. Someone tells you: room 7,432,891. Those words did not push you. Your legs moved you. But those words determined which room you ended up in. That was everything.

Information is an address in possibility space.

The rooms are possible configurations of matter. Natural processes—wind, water, chemistry, microbial degradation—are always moving. Industrial infrastructure—valves, control systems, treatment pathways—is always operational. People are always working. Everything is in motion. All of it will carry molecules somewhere no matter what. Human activity generates entropy that is not going to stop. The question is not whether it is produced. It is where it goes. Without questions reaching gates

→ uncontrolled. With questions reaching gates → directed. The questions do not push anything. They provide addresses. The walking was already happening.

The Gap Thousands of environmental questions are asked every second. Monitoring stations, satellites, CEMS, stream gauges, weather networks, IoT sensors, TEMPO, TROPOMI,

GHGSat—and human professionals reviewing data with their own eyes and judgment.

The raw questioning rate has never been higher.

Most of those answers go to databases. They do not close valves. They do not adjust schedules. They do not connect to any gate. A satellite measures NO over a city—the answer goes to a research archive. A CEMS records a spike—the answer goes to a quarterly compliance report reviewed weeks later. A stream gauge shows a critical threshold—nobody sees it in time.

The questions are being asked. The answers exist. The wire between the answer and the gate does not.

That wire is everything. In the Szilard engine—the experimentally verified system that proves a question is a physical act—observation and gate configuration are one event. No gap. The circuit is complete. In environmental management, the circuit is broken. Questions are asked by one system, answers stored in another, gates operated by a third. Without the wire, an observation is data. With the wire, an observation is protection.

The cost of this disconnection is staggering. The Bond-Bit Asymmetry—the ratio between the energy cost of moving a molecule and the energy cost of knowing where it is—is approximately 10 billion to 1 at current technology (derived from

Landauer’s principle and C–H bond energies; check the arithmetic in the companion paper). Every answer that reaches a gate saves ten billion times more than the answer cost. Every answer that goes to a database instead saves nothing. And that ratio doubles every 2.6 years (Koomey, IEEE, 2011) while chemistry costs remain fixed forever.

AI Is the Wire AI completes the circuit.

Not primarily because AI asks more questions—though it does, through physicsbased inference that reconstructs environmental states where no sensor exists. But because AI connects answers to gates at the speed the physics requires.

Sensor detects valve degradation → AI routes signal → valve closes. Satellite detects inversion forming → AI adjusts emissions schedule → plume disperses safely. Stream gauge shows drought threshold → AI triggers priority protocol → water reaches critical users.

No committee between. No weeks-long review. The observation configures the gate.

The measurement-actuation collapse completes. This is the Szilard engine at planetary scale.

AI contributes three things that no prior system could:

Integration. Satellite data, ground sensors, permit files, meteorological forecasts, regulatory history, 11 million environmental documents—combined into a single intelligence. A raw sensor reading alone is one answered question. Combined with all sources, it answers hundreds.

Inference. Physics-informed neural networks reconstruct environmental states where no sensor exists. Between monitoring stations—which is almost everywhere—the state was previously unknown. AI fills the gaps using the laws of physics. This is mathematically proven: boundary measurements determine interior states (BardosLebeau-Rauch, 1992).

Coupling. This is the critical one. AI routes answers to gates automatically, at machine speed. Sensor → AI → gate signal. The “decision” is embedded in the algorithm, exactly as in the Szilard engine. No latency between question and action.

The wire is built.

Why Now: Three Convergences This finding was derivable from existing physics since 2010. Nobody derived it—because the environmental profession and information thermodynamics have never met. Now three things have converged for the first time:

The physics was verified. Toyabe (2010) and Koski (2014) proved experimentally that information extracts real physical work. In 2024, Pruchyathamkorn et al. (Nature

Chemistry) demonstrated the first macroscale Maxwell’s Demon—driving material transport over centimeters using only information. Before 2010, this was theory. Now it is measured fact, at increasing scale.

The sensors arrived. TEMPO provides hourly atmospheric coverage of North America. GHGSat monitors 4 million facilities. IoT networks span entire watersheds.

The questions are now being asked at planetary scale. What is missing is the wire.

AI arrived. Machine intelligence can now integrate all sources, infer states where no sensor exists, and route answers to gates at machine speed. AI is the wire. And it can ask questions in thousand-dimensional spaces no human could conceptualize—genuinely new questions that expand what is possible to ask.

None works alone. Together they complete the circuit.

What We Are Building EnviroAI is building the wire. Not another sensor network. Not another database. The intelligence that connects questions already being asked—by satellites, sensors, monitoring equipment, and human professionals—to gates already built and waiting.

And that asks new questions where no sensor exists, using the laws of physics to reconstruct what is happening between the instruments.

The sensors exist. The gates exist. The answers exist. The connection between them does not. That connection is what turns a thousand environmental databases into a single protective system.

The Bottom Line For 50 years, the environmental profession assumed protection means physical intervention. Build the scrubber. Install the liner. Move the molecules. The physics says this is the most expensive possible approach—by a factor of ten billion.

The alternative is not doing less. It is completing the circuit. Every question whose answer reaches a gate is a physical act of protection. Every address provided is a gate configured. Every gate configured is a destination changed—from disorder toward order—powered by processes and infrastructure already in motion.

This is a theorem derived from experimentally verified thermodynamics, hidden for sixteen years in a gap between two fields that never met. They just met.

What This Does Not Claim It does not claim questions replace all infrastructure. Where no valve, no pathway, no control system exists, questions identify the need but do not satisfy it.

The claim holds where the control chain already exists—which, for most industrial facilities and all natural processes, it does.

It does not claim questions reverse past damage. Once contamination has dispersed, the Second Law requires real work to unmix. Questions dramatically reduce remediation cost by directing intervention precisely, but the zero-cost optimum is available only through prevention. The question must be asked before the entropy is produced.

It does not claim every question is answered correctly. A miscalibrated sensor provides wrong information, which misconfigures the gate and can make things worse.

The physics requires mutual information—correct correlation between question and reality. Quality assurance is not eliminated by this framework. It is made more important.

It does not claim the answer alone is sufficient. An answer that goes to a database and is never connected to a gate is not a completed protective act. The measurement-actuation collapse requires the full circuit: question, answer, gate. The wire between the answer and the gate is what makes observation into protection.

Without it, observation is data.

The extension from laboratory-scale to planetary-scale is analogical. The thermodynamic principle is verified at microscale (Toyabe, 2010) and demonstrated at centimeter scale (Pruchyathamkorn et al., Nature Chemistry, 2024). Planetary application is a framework claim supported by the physics but not yet directly verified at that scale.

Every question whose answer reaches the gate is a physical act of protection.

The sensors exist. The gates exist. The wire between them does not.

We are building the wire.

Key References Anderson, J. (2026). Observation IS Protection. EnviroAI Working Paper. [Full derivation]

Bérut, A. et al. (2012). Experimental verification of Landauer’s principle. Nature, 483.

Koomey, J.G. et al. (2011). Electrical efficiency of computing. IEEE Annals, 33(3).

Koski, J.V. et al. (2014). Experimental Szilard engine. PNAS, 111(38).

Pruchyathamkorn, J. et al. (2024). Macroscale Maxwell’s Demon. Nature Chemistry, 16(9).

Sagawa, T. & Ueda, M. (2008). Phys. Rev. Lett., 100, 080403.

Shannon, C.E. (1948). Bell System Technical Journal, 27(3).

Toyabe, S. et al. (2010). Information-to-energy conversion. Nature Physics, 6.

Wheeler, J.A. (1990). Information, physics, quantum.