AI and the End of the Normandy Paradigm
How the Iran conflict and Mythos are quietly introducing a software-first model for future warfare
The Iran conflict has entered its infrastructure and denial‑network phase. While markets and headlines fixate on Hormuz transits, missile exchanges, and ceasefire optics, a quieter but more consequential development is underway: the deployment of Anthropic’s Mythos AI model into U.S. military operations.
This is not another incremental drone or missile upgrade. Mythos represents a step‑change in how software becomes a primary vector for degrading an adversary’s ability to fight. For capital allocators tracking energy shocks, escalation scenarios, and operational constraints, understanding Mythos helps explain current U.S. tactics and frames the risks and duration of any future campaign phase.
What Mythos Is — And How It Fits Iran
Claude Mythos Preview is Anthropic’s frontier AI model, introduced in early April 2026 as the core of Project Glasswing. In controlled testing it autonomously discovered large numbers of zero‑day vulnerabilities across major operating systems, web browsers, and other critical software, then generated working exploits and chained them into full cyber kill chains with minimal human steering. It functions less like a better scanner and more like a tireless junior red‑team operator.
U.S. security services moved quickly. Public reporting indicates the NSA began using Mythos almost immediately, even as the Pentagon maintained a supply‑chain risk designation for Anthropic and sparred with the company in other venues. Earlier Anthropic models had already been feeding Palantir’s Maven system during the initial February–March strike phase; Mythos amplifies that existing pipeline at the software layer.
As of early May, roughly four weeks after announcement, there has likely been enough time for targeted mapping, vulnerability discovery, and at least limited operational effects against Iranian networks if U.S. planners chose to employ Mythos in that theater. Open sources do not document specific operations, but official comments make clear that AI is being used to shorten the path from “map the enemy’s digital terrain” to “gain leverage inside their systems.” Iran is the first sustained confrontation where that approach appears to matter for day‑to‑day crisis management in the Strait of Hormuz.
Why the Software Layer Is the First Battlefield
Modern militaries are software‑defined at every level: command‑and‑control links, radar fire‑control systems, missile guidance, drone telemetry, logistics, and maintenance networks all run on code. Many of these systems share common operating systems, libraries, or commercial components, creating a broad, overlapping attack surface. Whether you label it an A2/AD architecture or an integrated air defense system, the denial network is fundamentally software.
Mythos compresses the traditional cyber preparation timeline from months or years to days or weeks. It finds previously unknown flaws and turns them into working access at machine speed. In practice, that allows an adversary’s denial network to be degraded before additional kinetic strikes are required. Physical platforms may remain intact on paper, but their ability to function as a coordinated system erodes.
For the U.S. in the Strait of Hormuz, this is leverage. Quietly degrading Iranian command links, fire‑control software, and telemetry makes visible U.S. moves — destroyers in the strait, bombers on alert, carrier groups repositioning — less risky than they appear on a map. The more uncertain Iran’s ability to “shoot back” becomes at the software layer, the more confident U.S. planners can be in tighter, more assertive force postures.
From Normandy to Silent Prep
The days of storming defended beaches at Normandy, grinding through layered defenses via attrition, are largely over for non‑peer adversaries. The emerging pattern is quieter and more asymmetric:
1. Silent prep: intelligence fusion plus Mythos‑assisted mapping, vulnerability discovery, and pre‑positioned access across the adversary’s networks.
2. Force posture: visible but calibrated signaling — destroyers in the Strait of Hormuz, bombers on alert, allied forces postured.
3. Denial‑network degradation: pre‑positioned access could be activated against C2, radars, SAM fire‑control, missile telemetry, and drone links, turning the network slow, unreliable, or intermittently blind.
4. Kinetic exploitation: conventional assets operate in a suddenly permissive environment, striking key nodes with limited effective return fire.
In this sequence, air defenses and missile forces are degraded while U.S. weapons do not need to be dramatically more advanced. Once the denial network is compromised, air dominance could arrive more quickly. Non‑stealth legacy aircraft, including B‑52s, can loiter at altitude and service targets from outside the effective range of what remains functional on the other side.
For markets, this shifts attention away from hardware headlines — hypersonics, stealth platforms, new standoff munitions — toward software‑layer preparation. Campaigns against software‑dependent, non‑peer adversaries look less like long, grinding wars and more like punctuated episodes whose duration is bounded by political constraints, not production capacity.
Mythos and Iran’s Denial Network
Iran’s A2/AD posture rests on a layered denial network: integrated air defenses, coastal missile batteries, drone swarms, naval mines, and the C2 architecture tying them together. Much of the stack depends on aging Soviet and Russian‑derived systems supplemented by domestic code and commercial components. That combination is likely rich in overlooked vulnerabilities and long‑lived technical debt.
In that context, a Mythos‑enabled campaign would logically focus on the highest‑leverage nodes: links between radars and SAM batteries, missile brigade command systems and targeting databases, telemetry and guidance software for missiles and drones, and networks managing coastal defenses and naval units. Even partial success at those points can have outsized effects. The network does not need to be destroyed; it only needs to become slow, unreliable, or untrustworthy at critical moments.
This is the infrastructure‑denial equivalent of blockade logic. Operational constraints compound faster than inventory numbers suggest. If SAM batteries lose reliable tracks, drones sever links at key moments, or missile brigades doubt their targeting data, nominal force counts overstate true fighting power. That dynamic helps explain why U.S. force projection and Hormuz transits can remain relatively steady despite Iranian capabilities on paper: the constraint could be applied upstream, in software, not just at the visible platform level.
Beyond Iran: A Pattern for Software‑Dependent Adversaries
The Iran case points to a broader pattern. A Mythos‑style sequence for breaking denial networks generalizes to almost any software‑dependent adversary below top‑tier peer level. Once a state can rapidly map and exploit an opponent’s digital nervous system, the balance of coercion shifts.
Cuba offers one illustration. Older Soviet‑era SAMs and C2 systems there would be vulnerable to the same kind of mapping and exploitation. A future operation could compress into a brief air campaign with limited effective return fire. The political theater of defiance becomes expensive when the military reality is a rapid loss of ability to engage.
Similar logic applies to smaller regimes and some non‑state actors. Any networked or software‑dependent system becomes a potential target for silent degradation. That gives the U.S. a non‑kinetic coercion option that is deniable and scalable: an adversary may see increasing failures in its drones, missiles, or logistics without a clear, attributable kinetic trigger.
Adversaries will respond over time. Some may fold earlier in crises once they recognize their denial networks can be degraded remotely. Others will invest heavily in air‑gapping, custom code, and manual backups. Those paths carry costs: stepping outside the global software supply chain usually means less capable, harder‑to‑maintain systems, and greater operational friction.
When Losing in Cyberspace Means Losing the War
Cyberspace has long been part of conflict, but tools at the Mythos level change its relative weight. When autonomous systems can discover and exploit vulnerabilities across entire classes of software, the first decisive contest in a crisis is over whether your denial network functions on day one.
Lose that contest, and the physical fight becomes hard to win. Radars still spin, launchers still roll, drones still fly — but the logic and links that make them effective have been compromised. From a planning perspective, that comes close to preemptive neutralization. From a markets perspective, it suggests that outcomes hinge increasingly on who controls the most capable offensive‑AI cyber stack and who is most exposed to legacy, vulnerable software.
Iran and the Strait of Hormuz are simply the first visible theater where this shift is starting to matter in real time. Using Mythos there is not about science‑fiction warfare; it is about accelerating an existing battlefield in cyberspace to the point where it shapes the visible confrontation.
The future of warfare has arrived.
Disclaimer: This note is provided for informational purposes only and does not constitute investment, financial, or legal advice. The information contained herein is based on current market observations and analysis, which are subject to change without notice. All investments involve risk, including the loss of principal. We do not provide personalized recommendations, and readers should conduct their own due diligence or consult with a qualified professional before making any investment decisions.