Project Butterfly
of Damocles
A twelve-year retrospective on the supply chain attack surface — DEF CON 22 to the Glasswing Doctrine
In 2014 I stood at DEF CON and showed the internet’s foundational software was held together by wishful thinking, volunteer labor, and collective myth. Last week Anthropic’s unreleased frontier model validated that at industrial scale. This month two nation-state actors proved the security tooling itself is now the attack surface. And Project Glasswing — the response — may be the most consequential single policy decision to date in the history of open source software security. Whether that consequence is good or catastrophic depends entirely on decisions no one has made yet.
Episode guide
Three narrative threads across Season 3
Thread 1 — The vulnerability story (Ep. 01, 02, 06)
The quantitative case that OSS infrastructure was never as secure as the collective myth suggested — from Exim’s 13,000 criticals in 2014 to TensorFlow’s 700+ in 2026. The data through two generations of infrastructure.
- Ep. 01 — Internet infrastructure: DEF CON 22 dataset
- Ep. 02 — Supply chain: transitive dependency risk
- Ep. 06 — ML stack: the new load-bearing walls
Thread 2 — The attack story (Ep. 03, 04, 05)
How nation-state actors operationalized the threat models security researchers had been publishing for years. XZ as the template, Trivy as the pivot, Axios as the broadside. The playbook was published. They read it.
- Ep. 03 — XZ Utils: the two-year playbook
- Ep. 04 — Trivy/TeamPCP: the cascade
- Ep. 05 — Axios/UNC1069: the locksmith operation
Thread 3 — The response story (Ep. 07, 08, 09)
Glasswing as policy precedent. The compliance cliff. Whether the voluntary restraint that produced Glasswing is more durable than the restraint that left the bugs unfixed for 27 years. The substrate change at the governance layer.
- Ep. 07 — Glasswing: the doctrine
- Ep. 08 — The compliance cliff
- Ep. 09 — The pattern: season finale
Recurring themes across Season 3
Who is actually responsible for securing open source infrastructure?
Nobody — because everybody assumes somebody else will handle it. Introduced at DEF CON 22 to explain Heartbleed, the Everybody / Somebody / Nobody / Anybody parable returns in every episode as the structural explanation for why each generation of infrastructure inherits the same failure mode.
Why does security diligence become an attack surface?
Because the most security-conscious organizations ran Trivy most frequently and had the greatest exposure when Trivy itself was compromised. The XZ maintainer was trustworthy — which is exactly why the attack worked. Security posture becomes a vulnerability vector when the tools enforcing it are the target.
Why do the same infrastructure security failures repeat across technology generations?
The fairy dust doesn’t disappear — it moves one abstraction layer higher. 2014: “everyone’s looking at the code.” 2024: “our tooling is trustworthy.” 2026: “our AI deployment is safe.” The pattern is structurally identical. Only the substrate changes.
Why are open source maintainers prime targets for nation-state attacks?
Because the ROI is extraordinary: two weeks of social engineering to own 100 million weekly downloads; two years to own a transitive dependency of sshd. The economics of targeting volunteer maintainers only improve as package footprints grow and audit resources stay flat.
Why do good intentions fail to produce durable security governance?
Because both the OSS social contract and the Glasswing Doctrine are built on voluntary behavior at a scale they were never designed for. Glasswing is voluntary restraint by one company. The OSS model was voluntary contribution by scattered individuals. Season 3 asks what “durable” actually requires structurally.
What changes when AI discovers vulnerabilities faster than humans can patch them?
Discovery becomes essentially free, and the entire bottleneck shifts to everything that comes after: triage, coordination, patch development, deployment, compliance reporting. Mythos finds bugs at machine speed; patches still ship at human speed. Season 3 maps the 18-month gap between those two curves.
Supplementary Resources & Artifacts
For different ways to digest this information and explore the underlying thoughts, I compiled the original briefing documents, infographics, and multimedia discussions accompanying the series.
Multimedia & Discussions
Listen to or watch deep-dive discussions breaking down the key concepts from the series.
Executive Briefings & Decks
The original documents framing the strategic implications of Project Glasswing.
“Project Butterfly of Damocles: named for the moment you realize the sword has been hanging above the internet since 1998 — and that the thread was always a volunteer with a day job.”
— Series concept note, April 2026Episodes 01–06 are available now. Episodes marked [Forthcoming] (07–09) are scheduled for Q2–Q3 2026 and describe events as they develop — they are clearly distinguished from the documented incidents in earlier episodes. Subscribe to the Morphogenetic SOC newsletter at securesql.info for release notifications.
The author is affiliated with Project Glasswing. This post reflects the author’s independent analysis and opinions only, and does not represent the views or positions of Anthropic, Project Glasswing, or any Glasswing launch partner. Readers should weigh the author’s affiliation accordingly when evaluating assessments of the initiative’s merits and limitations.