Bucket C: Therapeutic Delivery

Bucket C contains the most advances whose maturity depends on breakthroughs that are not currently visible, and a bucket's overall timeline is paced by its slowest critical advance.

Two advances do the pacing. C4 (neural reconnection) is the longest-horizon item in the entire study at 30-50+ years, classified very-high-risk, because CNS regeneration requires several breakthroughs simultaneously — neutralising the inhibitory environment, reactivating dormant growth programs, providing supportive cellular substrate, and solving the specificity problem of getting axons to their correct targets — with no clear path to any of them at clinical scale. C3 (architectural tissue reconstruction) is similarly gated, at 25-40 years, by the vascularisation problem: thick tissue cannot survive without an internal blood supply, and building that hierarchy in vivo remains unsolved.

The other Bucket C advances are on faster timelines — C1, C5, and C6 are all in the 10-20 year range — but the bucket as a whole reaches full maturity only when its hardest members do. This is why the bucket assessment is honest that full Bucket C capability is a multi-decade proposition, while still identifying substantial near-term value in the faster advances.

Acknowledging this pacing honestly is essential to good roadmap design: a programme that promised full therapeutic capability on a 15-year timeline would be over-promising on C3 and C4, and the four-flag honesty system exists precisely to keep that distinction visible to funders and the public.

The shift is conceptual, not just incremental, and it reshapes everything downstream from regulation to clinical practice.

Current advanced therapies — CAR-T, the approved CRISPR therapies, regenerative-medicine products — are narrow-scope: one therapy targeting one disease, designed and validated for that specific indication, delivered under direct clinical supervision. Each is a product.

AIHS requires multi-scope platforms: therapies designed to be selected and adjusted by the system in response to real-time diagnostic input, delivered in coordination with other therapies (the C1 problem), reversible if they turn out to interact badly (the C5 problem), and safe enough for at least partially autonomous use (the C6 problem). The conceptual move is from "therapy as product" to "therapy as platform capability."

This shift has consequences far beyond the lab. Regulatory pathways currently expect product-level review of a fixed therapy for a fixed indication — they have no established framework for a platform that selects and combines interventions adaptively. Manufacturing must shift from custom batches per disease to modular production of composable components. And clinical practice changes: clinicians become AIHS operators supervising an adaptive system rather than prescribers choosing a fixed therapy. Much of the Phase I and Phase II Bucket C work is building the infrastructure — regulatory, manufacturing, clinical-workflow — to support this shift, not just the molecular science.

Significantly — C6's principles are not a separate layer bolted on at the end but a set of design constraints that propagate into every other Bucket C advance.

C1 (multi-target intervention) must produce explainable, auditable intervention selections, which constrains the AI architectures it can use — a maximally capable but opaque selector is unacceptable. C2 (universal tissue access) must respect the protective function of the barriers it crosses; methods that simply destroy the blood-brain barrier to get a payload in are off the table regardless of efficacy. C3 (architectural reconstruction) must build in mechanisms to halt or reverse a reconstruction that goes wrong inside a living patient. C4 (neural reconnection) must verify that connections are correct before they consolidate, because incorrect wiring is actively harmful. C5 (bounded reversible agents) is itself a component of the safety architecture. C7 (accelerated healing) must respect the patient's overall physiological state — the thermal ceiling — rather than maximising healing rate in isolation.

The cumulative effect is that C6-aware design takes longer and yields less raw capability per unit of investment than C6-blind design would. A team optimising purely for therapeutic potency would build more impressive demonstrations faster. But it would build capability that cannot actually be deployed, because an autonomous medical agent that cannot be explained, bounded, halted, and reversed is exactly the protomolecule failure mode the study is organised to avoid. C6 is the discipline that turns impressive-but-undeployable into deployable, and the study treats it as a first-class investment from the start for precisely that reason.

The bucket assessment identifies the most likely — and most desirable — development trajectory as modular: increasingly capable suites of specialised therapeutic tools coordinated by an external decision system, rather than a single unified molecular agent that does everything. This is both a prediction and a recommendation, and the reasoning matters.

It sidesteps the hardest problems. A single omnicompetent therapeutic substance would have to internally integrate diagnosis, decision-making, and every intervention modality — concentrating all of AIHS's hardest unsolved problems into one agent. A modular suite lets each tool be developed, validated, and deployed independently, with the coordination handled by an external system that can be inspected and controlled. The hard problems are separated rather than compounded.

It is far safer. This is the decisive argument. A single self-directing agent with the full range of AIHS therapeutic capability, operating inside a patient, is precisely the protomolecule nightmare — an autonomous, omnicompetent biological agent whose behaviour cannot be fully predicted or externally bounded. A modular architecture keeps the decision-making in an external, inspectable, overridable system (subject to the C6 safety architecture) while the in-body tools remain specialised, bounded (C5), and individually controllable. The coordination intelligence stays outside the body where it can be supervised; only specific, reversible effectors go inside.

It still delivers most of the clinical value. The modular trajectory captures the great majority of what AIHS is meant to achieve — coordinated, diagnostic-driven, adaptive intervention — without requiring the omnicompetent agent at all. The study's recurring theme is that the valuable parts of the AIHS vision and the dangerous parts are separable, and that good architecture deliberately keeps them separate. The modular suite is the concrete expression of that principle in the therapeutic bucket.