For All Mankind

What this becomes, and what would prove us wrong

Forecasts in deep tech are mostly wishful. So this page separates three things that usually get blurred together: what is already in motion, what we are betting on, and what physics will never allow regardless of funding.

01 — Horizons

Three timescales

Each horizon depends on the one before it. Nothing in the second is reachable without the hardware in the first, and the third is a clinical and regulatory problem as much as a technical one.

Fig 08 · Dependency, Not Prediction
HARDWARE 0 – 5 yrs sources · detectors arrays · integration CLINIC 5 – 15 yrs burns · margins · wounds endoluminal probes RESTORATION 15 – 30 yrs fascicular mapping limbs that feel gates ▶ gates ▶
Each stage is a precondition, not a phase of a marketing plan. If uncooled detector arrays do not arrive, the clinical horizon does not slip — it does not happen.
HORIZON I
Hardware — closing the gap
0 – 5 years · already in motion
  • Portable sources. Quantum cascade lasers now reach temperatures a compact thermoelectric cooler can hold. The engineering work is packaging, yield and lifetime rather than new physics — which is a very different kind of problem, and a much more tractable one.
  • Uncooled focal-plane arrays. The single highest-leverage development in the field. Terahertz imaging has historically meant a cooled single pixel raster-scanned over minutes. Video-rate uncooled arrays convert that into an instrument a person can hold and point.
  • Terahertz in silicon. CMOS and SiGe transceivers operating in the hundreds of gigahertz mean terahertz front ends that cost tens of dollars rather than tens of thousands. Cheap changes what applications are conceivable.
  • The 6G tailwind. Telecommunications is investing heavily in sub-terahertz components for the next wireless generation. We do not need that standard to succeed — we need the supply chain it creates. Somebody else is paying to industrialise our components.
  • Computational imaging. Compressive and single-pixel methods trade computation for detector count, and learned reconstruction extracts more from a starved signal than classical inversion does.
HORIZON II
Clinic — imaging without a dose
5 – 15 years · gated by Horizon I
  • Burn depth as a measurement. The first application where terahertz should become standard of care, because it answers a question clinicians currently answer by eye and the answer determines whether tissue is cut away.
  • Intraoperative margins. Tumour boundary determined while the patient is still open, rather than days later in pathology followed by a second operation. The clinical and economic case here is unusually clean.
  • Endoluminal probes. The honest answer to penetration depth is to stop trying to shine through a person and instead deliver the sensor to the tissue — catheter and endoscope-borne terahertz. This is where terahertz reaches organs.
  • Wound and ulcer monitoring at frequency. Zero dose means daily quantitative imaging becomes reasonable. Diabetic foot ulcers and graft viability are monitored on a schedule no radiation-based modality could support.
  • Cost collapse changes geography. A solid-state instrument has nothing like the economics of a cryogenic magnet in a shielded room. The target is diagnostic imaging a rural clinic owns outright — which is where "For All Mankind" stops being a slogan.
HORIZON III
Restoration — the wiring problem
15 – 30 years · gated by Horizons I and II
  • Fascicular mapping as routine imaging. Patient-specific nerve topography before the incision, turning electrode placement from exploratory craft into a deterministic procedure.
  • Bidirectional prosthetics as standard of care. Not a research demonstration for a handful of participants, but the default fitting — limbs with graded force control that report pressure and position back to their owner.
  • Sensation as the metric that matters. The field's own data is clear that feedback, not degrees of freedom, determines whether a prosthesis is used or abandoned. Embodiment is the outcome measure.
  • Bridged spinal pathways. Walking restored by decoding intent and stimulating below the lesion. Demonstrated in individuals already; the work is making it robust, implantable and affordable.
  • Sealed, ambient-powered implants. No wire through the skin — the dominant chronic failure mode — and no charging ritual. Routine living tops the device up.
02 — Our bets

Five assumptions we are exposed to

These are not predictions, they are positions. If we are wrong about one, it costs us years. Stating them plainly is how you find out early.

01 — Someone else industrialises our components

We are betting that 6G research and automotive sensing pull sub-terahertz devices into volume manufacturing, and that we inherit the supply chain. A vertically integrated terahertz company that must fabricate everything itself is a far worse business.

02 — Computation beats penetration

Water absorption is not going to improve. We are betting that reflection-mode coherent reconstruction, frequency agility and learned priors extract clinically useful depth from a signal that classical inversion treats as hopeless.

03 — Peripheral interfaces win over cortical ones

No craniotomy, no glial scarring against a cortical array, no asking the brain to learn an artificial code — and a signal already in the limb's native language. We are betting the lower-risk path is also the higher-fidelity one.

04 — Defence funding is the only patient capital for this

Closing a forty-year hardware gap is not a venture-timescale problem. We are betting that government programmes fund the components, and that the same components then serve medicine — the path GPS, the internet and modern imaging all took.

05 — Honesty is a competitive advantage

Terahertz has a history of overclaiming, and the field carries scar tissue from it. We are betting that publishing our limits attracts the physicists, program managers and clinicians who can actually tell the difference.

03 — Falsification

What would show this is wrong

A roadmap that cannot fail is not a roadmap. These are the observations that should make you — and us — stop.

Uncooled arrays do not arrive

If sensitive terahertz detection still requires cooling and single-pixel scanning in ten years, the clinical horizon does not slip — it closes. Every portable application assumes video-rate uncooled imaging.

Reconstruction plateaus at the surface

If computational methods cannot push useful imaging meaningfully past a millimetre in living tissue, terahertz remains a superb dermatology and intraoperative tool and never becomes a general diagnostic one. That is a real business — but it is not this one.

Chronic interfaces keep degrading

If foreign-body response and micromotion still erode neural signal quality over a handful of years, restoration is a temporary intervention rather than a life returned. Materials, not algorithms, decide this.

Nerve contrast is insufficient

Our fascicular mapping claim rests on terahertz sensitivity to myelin and bound water being strong enough to resolve individual fascicles in vivo. If that contrast does not hold outside excised tissue, the deterministic-placement argument collapses.

04 — Fixed points

What no amount of money changes

Four constraints are properties of the universe rather than of our engineering. Any roadmap that quietly assumes these will yield is not worth reading.

4.1 meV

Terahertz will always be non-ionising. Photon energy at 1 THz is roughly two thousand times below the threshold to strip an electron. That is permanent, and it is the safety argument that never needs revisiting.

~200 cm⁻¹

Water will always absorb it. The absorption of liquid water at terahertz frequencies is a material property. Better engineering routes around it; nothing removes it.

λ / 2

Diffraction sets the resolution. At 1 THz the wavelength is 300 µm, so far-field imaging resolves on the order of 150 µm. Sub-wavelength detail requires near-field probing, which only works at surfaces.

1 × 1

No aperture means no direction. A single antenna cannot infer bearing. Position requires many spatially separated radios or genuine array bandwidth — better processing of one link will never produce it.

The shape of the whole thing

Two hundred years of electromagnetism produced one band we never learned to use, sitting exactly where biology is most legible and radiation damage is impossible. The hardware to open it is arriving now, mostly funded by people who want faster wireless. What that hardware is pointed at is a choice — and we would rather it were pointed at finding people under rubble, at telling a surgeon where the burn actually ends, and at giving a hand back to someone who can feel it.