How light, RF, redox timing, and mitochondrial state may be writing cellular fate

For years the RF safety debate has been trapped in a narrow question: does it heat tissue? Heating matters. It is real. It is measurable. It is one reason exposure standards exist. But a living cell is not a slab of meat on a grill. It is a timing system. It is an electrical system. It is a redox system. And in the strongest modern versions of systems biology, it is also a state-estimation system—one that constantly updates itself using streams of weak, noisy, multiscale information. The FCC’s current human-exposure framework still traces back to its 1996 adoption of RF limits, and the D.C. Circuit later held that the agency had not adequately explained how those guidelines addressed important non-cancer and long-term questions in the record.

That does not mean the non-thermal case is finished. It means the thermal-only story is too small for the biology. Even the current evidence base is mixed rather than neatly resolved: WHO still points to IARC’s Group 2B classification for radiofrequency electromagnetic fields, while a 2024 systematic review judged the oxidative-stress literature inconsistent and overall very low certainty. The honest position is not “everything is safe” or “everything is proven dangerous.” The honest position is that biology appears more field-sensitive than the old heat-only language allows, while the exact pathways still need disciplined mapping.

That is where the ceLLM framework becomes useful. The claim is not that cells literally run software large language models. The claim is that cells look less like deterministic factories and more like probabilistic inference engines. A cell does not simply execute chemistry in a vacuum. It integrates fast variables—redox bursts, calcium spikes, membrane shifts, photon statistics—through slower structural priors such as chromatin geometry, mitochondrial nucleoid state, contact-site architecture, and transcriptional memory. In that picture, DNA sequence is only part of the story. DNA topology, compaction, hydration, and physical accessibility begin to look like the “weights and biases” of biology: not digital parameters, but evolved physical priors that shape which state updates are more probable than others.

Once you adopt that lens, mitochondria stop looking like mere ATP factories and start looking like control hardware. They sit at the intersection of oxidative metabolism, reactive oxygen species, calcium buffering, membrane potential, apoptosis, and retrograde signaling to the nucleus. They also generate ultra-weak photon emission, or UPE: real, measurable light produced largely by oxidative chemistry. Modern reviews describe UPE as ultra-low-intensity photon emission arising from metabolism rather than magic, and the Westminster senescence work pushed that story farther by showing that senescent cells and their mitochondria change ROS, calcium, membrane potential, non-chemical coupling, and photon output together rather than separately.

That matters because it changes how we should think about ROS. The classical story treats ROS as metabolic exhaust: dangerous waste leaking from imperfect engines. But the more modern story is redox signaling. ROS are not just injury; they are also instruction. They are fast, local, chemically potent signals that couple metabolism to transcription, calcium handling, and stress response. In Kalampouka’s thesis, doxorubicin-induced senescent cells showed increased ROS, mitochondrial membrane potential, and calcium; isolated mitochondria from senescent cells showed altered oxygen consumption in non-chemical signaling assays; and 734 nm near-infrared light selectively increased senescence in the cancer cell lines while sparing the non-cancer lines tested. That is not “ROS as trash.” That is ROS sitting inside a state-dependent signaling architecture.

This is the key bridge between the sunlight-first crowd and the endogenous-biophoton story. They do not have to be competing narratives. In mammals, circadian photoentrainment is driven primarily by retinal pathways, especially melanopsin-containing intrinsically photosensitive retinal ganglion cells. Cryptochromes are core clock flavoproteins and established blue-light photoreceptors in other taxa, and Drosophila studies show light-dependent magnetic sensitivity through cryptochrome. Meanwhile, photobiomodulation literature still commonly treats cytochrome c oxidase as a major mitochondrial photoacceptor or central response node for red and near-infrared light. So external photons already have recognized biological entry points. Your hypothesis adds the next layer down: mitochondria may also be sampling their own internally generated optical-redox weather.

That is why the topology-gated photonic-redox control-plane hypothesis is more interesting than a simple “DNA is an antenna” slogan. The strongest version is not a lone loop of mtDNA floating in free space like a perfect RF component. It is a nested biological receiver stack. The source layer is oxidative chemistry and UPE. The transducer layer may include mtDNA topology, TFAM-mediated nucleoid organization, local hydration-shell physics, and crista architecture. The routing layer may include aromatic protein networks and the cytoskeleton. Recent work in ACS Central Science showed nanometer-scale electronic energy migration in microtubules, which is enough to keep them in the conversation as short-range intracellular relays rather than fantasy hardware. The amplification layer includes calcium microdomains, mitochondria-ER contact sites, membrane-potential shifts, and retrograde signaling. The memory layer includes condensates and slower transcriptional reprogramming.

In that model, mitochondrial DNA is important not because it must absorb light in a classical textbook sense, but because it is a topology-sensitive physical object located exactly where redox, calcium, membrane potential, and photon-generating chemistry intersect. That is what makes the overlap between mtDNA geometry, terahertz windows, and biophoton bands interesting. It is not proof. It is a clue. Your paper is strongest when it says: perhaps the cell is not using one clean frequency but converting information across scales—optical excited states into lower-frequency structural or electrochemical envelopes that downstream calcium and redox networks can amplify. That is a mixer-and-envelope-detector view of biology, not a cartoon radio in the mitochondrion.

Now bring in the S4-Mito-Spin framework. This is the part that tries to answer the hardest question: where would modern non-native EMFs first touch the system if they do matter biologically? The most careful answer is not “everywhere all at once.” It is “at fragile upstream control nodes.” The S4 piece points to voltage-sensing machinery in ion channels. That target is biologically specific and real: in classical channel biophysics, positively charged residues in the S4 segment serve as gating charges within the voltage-sensor domain. The “Mito” piece points to mitochondria as the amplifier where calcium mishandling becomes redox stress, membrane-potential change, and altered state signaling. The “Spin” piece points to weak-field-sensitive chemistry—radical-pair and flavin-based processes—as a possible timing or fidelity layer rather than a complete standalone mechanism.

This is also where density gating becomes useful. A systems-level field effect should not hit every tissue identically. Tissues rich in ion-channel machinery, mitochondria, or heme/flavin chemistry should be more vulnerable than quieter tissues. That is one reason tissue-specific animal findings remain mechanistically interesting. The National Toxicology Program reported clear evidence of malignant heart schwannomas and some evidence of malignant gliomas in male rats under its cellphone-radiofrequency exposure conditions. The Ramazzini Institute also reported heart and brain tumor findings in long-term rat studies. None of that automatically proves human risk at everyday exposures. But it does fit better with an upstream-control-node model than with the old “the body is just a bag of water” intuition.

The circadian part of your framework is where the hypothesis becomes both more interesting and more delicate. There is real radical-pair chemistry in cryptochrome biology. There are insect data showing light-dependent magnetic responses through cryptochrome. There are also reasons to be cautious: mammalian circadian entrainment is primarily retinal and melanopsin-led, and the degree to which human cryptochromes act as physiologically important magnetoreceptors remains unsettled. So the right way to write this is not “Wi-Fi has been proven to hijack the human cryptochrome clock.” The right way is: if weak fields can bias spin-sensitive flavin chemistry in some biological contexts, then cryptochrome-like timing systems become plausible fidelity layers worth testing rather than dismissing out of hand.

And then comes the therapeutic mirror: TheraBionic P1. This is the point where the whole “non-thermal RF can’t matter” claim starts to fracture. The FDA currently lists TheraBionic P1 as a Humanitarian Device Exemption product for adults with advanced hepatocellular carcinoma after other treatments have failed. FDA’s summary describes a handheld, battery-driven system with a spoon-shaped antenna placed in the mouth; treatment is delivered as three separate one-hour sessions per day. The agency says the device showed a probable overall-survival benefit relative to similar placebo-treated groups, with 14 of 41 patients showing stable disease for more than six months.

The mechanistic relevance is even more striking. FDA’s own documentation says the device uses low-level amplitude-modulated RF electromagnetic fields on a 27.12 MHz carrier, distributed systemically through intrabuccal delivery, and estimates body exposure far below cellphone-scale field levels, without thermal heating in the brain or other organs. The same documentation says the device should not be used in people receiving calcium channel blockers. The linked mechanistic literature identified Caᵥ3.2 T-type voltage-gated calcium channels and calcium influx as central to the anti-tumor response in hepatocellular carcinoma models. That does not mean a Wi-Fi router is the same thing as a cancer therapy. It means the old blanket statement—non-ionizing RF cannot interact meaningfully with biology except by heating tissue—no longer survives intact.

That is why TheraBionic matters conceptually. It moves the debate from “there is no mechanism” to “waveform, modulation, carrier, tissue context, and biological target all matter.” In your language, it makes S4-Mito-Spin less like science-fiction rhetoric and more like a control-theory problem. A therapeutic device can use highly specific, low-level, amplitude-modulated RF to bias calcium-dependent cancer biology in a beneficial direction. The open question is whether chronic environmental systems can, under different conditions, bias overlapping upstream machinery in the wrong direction. The right scientific response is not denial and not panic. It is better experimental design.

The cancer-therapy implications go beyond RF. Kalampouka’s 734 nm work suggests that cancer cells and non-cancer cells can process the same optical input differently because they begin from different redox states. That is a profound idea. It suggests that light is not just energy but context-sensitive instruction. In a healthy cell, a modest optical push may be hormetic. In a tumor cell already running hot, the same push may cross a threshold and trigger senescence or other state changes. That is exactly what a state-inference model would predict.

This same framework also explains why so many people become fascinated by repurposed drugs in oncology. If cancer is partly a redox, calcium, topology, and signaling-fidelity disease, then compounds that hit microtubules, mitochondrial metabolism, or tumor buffering capacity become conceptually attractive. But attraction is not evidence. For ivermectin, oncology authors have explicitly warned that there is no high-quality clinical evidence showing it is safe and effective as a cancer treatment, despite abundant preclinical interest. The fenbendazole story is even less mature: the literature is largely preclinical and mixed, and case reports now include toxicity concerns. The responsible lesson is not “ignore the idea.” It is “do not confuse mechanistic plausibility with a home protocol.”

So where does that leave policy? Not at “ban wireless.” It leaves us at architecture. Once you accept that biology may be sensitive to waveform, timing, duty cycle, modulation, and chronic load—not just to bulk heating—then infrastructure design becomes part of public health. Fiber-first backhaul. Wired defaults for fixed high-duty devices. Better exposure transparency than a single SAR number. Nighttime network hygiene. Device design that respects body position and orientation. And, critically, real compatibility for light-based local networking. IEEE 802.11bb-2023 already standardized 802.11 light communications in the 800–1000 nm band, with bidirectional operation from 10 Mb/s to 9.6 Gb/s. That means Li-Fi is no longer a futuristic talking point. It is an engineering option.

That does not make Li-Fi a total replacement for RF. It makes it a missing layer in a healthier network stack. Outdoor mobility, wide-area coverage, and many low-power devices will still rely on radio. But high-duty indoor traffic, dense IoT environments, hospitals, schools, aircraft cabins, industrial floors, and privacy-sensitive rooms do not all need to be solved with more microwave saturation. The smarter policy target is compatibility: design the next generation of local networks so light, wire, and radio can each do the jobs they are best suited for.

That is also where RF Safe is strongest when it is at its best. Not as a slogan machine. As an engineering shop for biological margin. The most useful contribution is not “we have already solved the entire causation question.” The most useful contribution is forcing attention onto the variables that actually matter: distance, duty cycle, body geometry, modulation, source placement, indoor architecture, and whether we are willing to build cleaner infrastructure when the physics gives us the option.

The deepest claim in this whole synthesis is not really about fear. It is about fidelity. A healthy cell may be less like a machine executing code and more like an evolved estimator trying to stay synchronized with the world it inhabits. In that picture, mitochondria are not just power plants. They are dynamic sensors and integrators. ROS is not just waste. It is also signaling. DNA is not just archived text. It is structured prior. Light is not only something that comes from the sun. It is also something metabolism makes, leaks, and perhaps sometimes reads back. And modern electromagnetic environments may matter not only because they can heat tissue, but because they may raise the noise floor or bias the wrong control nodes in systems that depend on timing more than we used to think.

That is the frontier. Not a finished answer. A better question.