Hemispherical Stacks · 2026-05-24

🌐 Hemispherical Stacks — 2026-05-24

🔄 CSIS: Allied Export Controls Accelerate China's Semiconductor Localization — 50% Domestic AI Chip Share by 2026
⚖️ China's Rare Earth Ceasefire Expires November 2026 With 85-90% Refining Control Unchanged
🤝 Trump-Xi Summit Left Unresolved: AI, Cyber, Export Controls, Digital Sovereignty Per CSIS Analysis
🪖 Pentagon Signs Classified AI Partnerships With OpenAI, Google, Nvidia, SpaceX — Restricts Anthropic
🛰️ AUKUS DARC Site 1 Western Australia Delivers Southern Hemisphere Tracking Data for 2027 Full Capability
🏗️ China's Export Control Architecture: State Council Order 834 Unifies Controls + Countermeasures + Data Security

---

🔄 CSIS: Allied Export Controls Accelerate China's Semiconductor Localization — 50% Domestic AI Chip Share by 2026

CSIS's analysis of China's semiconductor localization drive published this week delivers the core thesis of the current US-China tech competition in a single finding: US and allied export controls aimed at slowing China's semiconductor progress are instead accelerating Beijing's push for self-reliance. TrendForce projects that in 2026, the domestic share of China's AI chip market will increase to 50 percent — up from negligible domestic supply when the first wave of semiconductor export controls was imposed in October 2022.

The mechanism is the substitution architecture. CSIS notes that Chinese consuming firms are turning to import substitution procurement as a hedge against supply chain disruptions, as a response to government mandates, and "as a reflection of the growing competitiveness of domestically made" chips. These three drivers operate differently: hedge procurement is reversible if export controls relax; government mandate procurement is policy-dependent; competitiveness-driven procurement is structural and cumulative. The CSIS analysis is careful not to overstate the success of Chinese localization at the leading-edge node level, where progress has been "mixed" — but it documents substantial progress in mature nodes and in the AI chip market specifically.

The Huawei Ascend 950PR's capability profile — 1.56 PFLOP at 1.56 PFLOP with 112GB HBM — is the data point that operationalizes the 50% domestic market share projection. Jensen Huang's public acknowledgment that Nvidia has "largely conceded" China's AI chip market to Huawei is not a CSIS finding but confirms the CSIS thesis from the supply side: the demand for domestic AI chips exists, Huawei is supplying it, and the market share transition is operational fact rather than projection.

The cross-hemisphere structural insight is the asymmetry in how control architecture and substitution architecture respond to escalation. US-led controls operate through multilateral coordination: Japan, the Netherlands, South Korea, and Taiwan must all maintain aligned export restriction regimes for the controls to hold. Each partner's participation is politically negotiated and economically costly. China's substitution architecture requires only domestic capital allocation and state industrial policy — both of which Beijing controls unilaterally. Every round of allied export control escalation creates a new political negotiation burden on the allied side and a new investment mandate on the Chinese side.

The Andersen Institute analysis documents China's counter-move architecture: the October 2025 ceasefire at Busan (US suspended BIS Affiliates Rule, China suspended its counter-controls through November 2026) represents a managed de-escalation that buys Beijing time to expand domestic capability rather than resolve the underlying competition. The November 2026 expiry is the next decision point.

Sources:

---

⚖️ China's Rare Earth Ceasefire Expires November 2026 With 85-90% Refining Control Unchanged

The November 2026 expiry of the Wave 2 suspension on China's rare earth export controls — agreed at the APEC summit in Busan, October 2025 — is the next critical decision point in the strategic minerals competition. Discovery Alert analysis notes that China controls an estimated 85-90% of global rare earth refining and processing capacity. The Busan ceasefire paused both the US BIS Affiliates Rule and China's October 2025 counter-measures, but it did not address the underlying structural dependency: processing capacity concentration in China is not reduced by a political ceasefire, only by years-long investment in non-Chinese refining infrastructure that has not materialized at scale.

The indium case is the indicator variable for how narrowly the ceasefire operates. US News reported that the White House mentioned indium for the first time as a critical mineral concern addressed at the Trump-Xi summit — indium has been on China's export control list since February 2025, plays a key role in the upstream and downstream semiconductor supply chain, and is an example of a material where Chinese processing control creates a supply chain bottleneck that no US or allied facility currently resolves.

The Andersen Institute's documentation of State Council Order No. 834 — which integrated China's export controls, countermeasures, and data security provisions into a unified legal architecture in early 2026 — reveals the structural ambition behind the ceasefire. China is building a permanent export control system with legal coherence and administrative coordination across multiple policy instruments, not simply using export controls tactically. The January 2026 catalogue expansion added rare-earth compounds to the control list, expanding the scope of controlled materials even while the ceasefire suspended active enforcement of the most recent escalatory measures.

The two-hemisphere asymmetry here is the inverse of the semiconductor story. In semiconductors, the US alliance system controls inputs (EUV lithography, advanced chip design tools, leading-edge fabs) that China lacks; China controls the market for the chips produced. In rare earths, China controls the processing capacity that Western manufacturing depends on; the US alliance system lacks equivalent processing infrastructure. The control chokepoint in semiconductors is upstream (design tools, fab equipment); the control chokepoint in rare earths is midstream (refining and processing). Both chokepoints are Chinese, but they operate through different mechanisms: chip controls require partner compliance from Japan/Netherlands/Taiwan/South Korea; rare earth controls require only Chinese administrative decision.

The Discovery Alert scenario analysis presents November 2026 as a binary: either Phase Two controls are suspended again (extending the ceasefire) or reinstated (triggering supply chain acute shortages in defence, energy, and AI infrastructure manufacturing). The White House's "small rare earth win" at the Trump-Xi summit — China's agreement to address US concerns about specific shortages, including indium — may indicate that the ceasefire will be extended, but provides no structural resolution of the 85-90% refining concentration that makes the concern recurrent.

Sources:

---

🤝 Trump-Xi Summit Left Unresolved: AI, Cyber, Export Controls, Digital Sovereignty Per CSIS

CSIS's post-summit analysis finds that the May 14-15 Beijing meeting "revealed how little progress has been made on the most consequential dimensions of US-China competition: AI, cyber operations, export controls, and digital sovereignty." Trade dominated the public headline — the tariff ceasefire and rare earth commitments — while the structural technology competition elements that define the decade-scale contest remained explicitly unaddressed.

The CSIS framework identifies digital sovereignty as the dimension most resistant to bilateral negotiation. Digital sovereignty disputes are not resolvable through trade agreements because they involve incompatible regulatory architectures: the US assumes that data flows freely across borders subject to individual rights protections; China assumes that data generated within Chinese jurisdiction is subject to Chinese state authority regardless of where it is processed. These architectures cannot be reconciled through bilateral agreement — each government would have to abandon its foundational assumption. The summit left this unresolved because it is structurally unresolvable through the diplomatic instruments available to a summit.

The AI competition aspect of the left-unsaid is the most consequential for the decade. The Trump-Xi bilateral AI safety protocol (nonstate actor access) addresses a shared interest at the margins of the competition. The core competition — who controls the AI infrastructure stack (compute, data, models, deployment platforms), who sets AI governance norms globally, and whether the two AI ecosystems converge or bifurcate — was not addressed. CSIS's A Confident Beijing analysis notes that Beijing considers itself in a strong position following the Supreme Court's February 2026 IEEPA ruling (which reduced US tariff authority) and the Busan ceasefire (which paused the most aggressive export control escalation). China's negotiating posture at the summit reflected confidence, not concession.

The Geopolitics analysis frames the summit's tech competition left-unsaid correctly: "Economic competition is no longer primarily about tariffs, trade deficits or market access. It is increasingly regarding control over strategic pillars of future power like AI, semiconductors, rare earths, advanced manufacturing and supply chains." The summit produced agreements on the dimensions that are resolvable through bilateral trade negotiation (tariff structures, rare earth access) and left unresolved the dimensions that are not (AI infrastructure sovereignty, export control architecture, digital governance).

The next bilateral engagement opportunities are November 2026 APEC in Shenzhen and December 2026 G20 in Miami. Both are multilateral settings where bilateral tech competition specifics are harder to address than in a dedicated summit format. The November 2026 rare earth ceasefire expiry, combined with the absence of export control resolution at the Beijing summit, creates the conditions for the next escalation cycle absent a bilateral agreement before November.

Sources:

---

🪖 Pentagon Signs Classified AI Partnerships With OpenAI, Google, Nvidia, SpaceX — Restricts Anthropic

The Pentagon's classified AI partnerships announced in May 2026 include agreements with OpenAI, Google, Nvidia, Microsoft, Oracle, Amazon Web Services, SpaceX, and startup Reflection AI for classified military AI applications. CXO Digital Pulse reported that Anthropic systems have been restricted from the DoD AI testing portfolio following Anthropic's refusal to extend its previous Pentagon agreement to classified use cases — a position consistent with Anthropic's Claude Mythos withholding decision and its submission of cyber vulnerability findings to the FSB rather than to the DoD.

The Anthropic-Pentagon divergence reveals a structural tension in the US AI industrial policy. Anthropic's safety posture — withholding Mythos from public release, declining classified DoD access, disclosing to the FSB — is consistent with a company that treats AI misuse risk as a first-order constraint on deployment. The DoD's AI procurement posture is the opposite: it is acquiring classified AI capabilities specifically for applications where misuse potential is inherent to the mission. The Pentagon's restriction of Anthropic from its testing portfolio is not a finding that Anthropic's models are less capable — it is a finding that Anthropic's deployment terms are incompatible with the DoD's classified use requirements.

The cross-hemisphere significance is the contrast with China's AI military integration architecture. China's People's Liberation Army has an explicitly civilian-military fusion (军民融合) policy that mandates technology transfer between commercial AI development and military applications. The PLA does not need separate procurement deals with Huawei, Alibaba, or Tencent because the civilian-military boundary does not exist in the same form as in the US. US AI companies can opt out of DoD contracts; Chinese AI companies cannot opt out of PLA technology transfer requirements in the same way.

The Pentagon's deal with SpaceX is the most strategically significant single element. SpaceX's IPO filing identified orbital AI compute as a core business pillar; the DoD classified partnership means SpaceX's orbital infrastructure is now dual-use in the most explicit sense — commercial orbital compute and classified military orbital sensing/communication/compute on the same satellite architecture. This is the hemispherical stack becoming literal: the same infrastructure platform serves commercial AI inference and classified military intelligence simultaneously.

Defense One's analysis of special operations AI requirements frames the DoD's AI procurement as a size and deployability problem, not a capability problem: "AI agents are coming to a special operations mission near you — if they can fit in the pack." The military's frontier requirements are not the same as commercial frontier requirements: ruggedized, power-constrained, latency-tolerant, and air-gapped. This differentiates the DoD AI stack from commercial AI infrastructure and creates a development pathway that is neither fully commercial nor fully government — exactly the dual-use space that SpaceX, Nvidia, and the other DoD partners now occupy.

Sources:

---

🛰️ AUKUS DARC Site 1 Western Australia Delivers Southern Hemisphere Space Tracking for 2027 Full Capability

Australia's Deep Space Advanced Radar Capability (DARC) Site 1 in Western Australia is delivering early space tracking data to AUKUS partners, with full operational capability targeted for 2027. SpaceNews reports that the DARC deployment creates "an immediate opportunity to expand Australia's role beyond space domain awareness into full-spectrum southern launch, recovery and manufacturing infrastructure" — framing the Southern Hemisphere not as a remote location but as a strategic complementary geometry to Northern Hemisphere space infrastructure.

The cross-hemisphere geometry is the strategic logic. Northern Hemisphere launch facilities (Kennedy Space Center, Baikonur, Wenchang) have limited orbital coverage of the Southern Hemisphere and Southern Ocean. Southern Hemisphere launch and tracking facilities provide complementary coverage that creates a more complete global space domain awareness picture for the AUKUS partnership. Australia's 2026 National Defence Strategy identifies space as a warfighting domain — the DARC operational data represents the first deliverable from that identification.

The dual-use dimension runs in both directions. DARC tracks all objects in its coverage zone — commercial satellites, military systems, and debris — making it a commercial asset as well as a defence asset. Australia's geographic position creates a Southern Hemisphere coverage zone that complements US, UK, and European Northern Hemisphere space tracking infrastructure, providing AUKUS partners with a global space domain awareness picture they cannot assemble without Southern Hemisphere participation.

The contrast with China's space infrastructure expansion is structural. China's AUKUS analog is the BeiDou navigation system and its associated tracking infrastructure, built as a strategic alternative to GPS dependence. China's space domain awareness is primarily served by domestic tracking facilities, with diplomatic access to partner facilities in some African and South American locations. The AUKUS Southern Hemisphere space infrastructure and China's BeiDou global network are competing architectures for space domain awareness in the Southern Hemisphere — one operated through alliance partnership with integrated AUKUS data sharing, one operated through Chinese state infrastructure with bilateral access arrangements.

The Guardian's AUKUS cost analysis notes that AUKUS is forecast to cost Australia up to A$368 billion to the mid-2050s — the submarine program alone dwarfs the space infrastructure investment. But the space infrastructure is the element delivering operational capability in 2026-2027, while the submarine program faces build-rate delays. DARC Site 1's early operational delivery against the 2027 full-capability timeline is an AUKUS success story in a program with many timeline slippages.

Sources:

---

🏗️ China's Export Control Architecture: State Council Order 834 Unifies Controls, Countermeasures, Data Security

Andersen Institute's analysis of China's evolving export control framework documents State Council Order No. 834, which integrates China's export controls, countermeasures, and data security provisions into a unified legal architecture. The January 2026 catalogue expansion added rare-earth compounds and other materials to the control list alongside the State Council Order — extending both the legal framework and the controlled item scope simultaneously.

The structural significance of Order 834 is the integration of three previously separate policy instruments: export controls (blocking outbound transfer of controlled items), countermeasures (retaliating against foreign measures targeting China), and data security (controlling outbound transfer of data generated in China). Prior to Order 834, these operated under separate legal authorities with separate administrative agencies. Their integration creates a unified executive instrument through which China can respond to any foreign measure — sanctions, export controls, regulatory actions — with a range of responses from export restriction to counter-sanction to data access restriction, all under a single legal framework.

The contrast with the US BIS architecture is illuminating. BIS administers export controls under the Export Administration Regulations (EAR), which covers goods and technology. Countermeasures (IEEPA sanctions, OFAC designations) are separate. Data security (the now-stalled CFIUS data rule, the Cloud Act) is under a third framework. The US response architecture is fragmented across Commerce, Treasury, DOJ, and the intelligence community — coordination requires inter-agency processes that are slower than China's unified executive response.

Wikipedia's analysis of US chip export controls notes that BIS formalized a flexible license review policy in January 2026 for chip transactions — a calibrated adjustment rather than Order 834's structural integration. The US is making tactical adjustments to its export control architecture while China is building a permanent integrated framework. The asymmetry in institutional architecture — unified vs. fragmented — will determine which side responds faster and more coherently to the next escalation cycle.

The Busan ceasefire (October 2025) and its November 2026 expiry function within this asymmetric architecture. The ceasefire suspended active enforcement of both sides' escalatory measures — US Affiliates Rule suspension, Chinese counter-control suspension. But the suspension operates differently on each side: the US suspension is a BIS administrative action that requires BIS coordination to reverse; China's suspension is revocable by a decision of the State Council under Order 834's unified framework. China's ability to escalate faster than the US after November 2026 is structurally embedded in this architectural asymmetry.

Sources:

---

Research Papers

China's Localization Drive in Semiconductors Gains Impetus from Allied Chip Export Controls — CSIS (May 2026) — Comprehensive analysis of how US-allied export controls are accelerating Chinese semiconductor domestic substitution; TrendForce projection of 50% domestic AI chip share in 2026; documents substitution at both mature and advanced nodes.

What the Trump-Xi Summit Revealed, and Left Unsaid, About U.S.-China Tech Competition — CSIS (May 2026) — Post-summit analysis identifying AI, cyber, export controls, and digital sovereignty as unresolved dimensions; frames trade-dominated headline vs. structural technology competition leaving; documents both powers' negotiating postures.

China's Export Control Architecture and Use of Critical Minerals as Strategic Pressure Points — Andersen Institute (May 2026) — Detailed architecture analysis of State Council Order 834 and its integration of export controls, countermeasures, and data security; Busan ceasefire structure and November 2026 expiry analysis; critical minerals control list evolution.

---

Implications

The structural condition of the hemispherical technology competition in May 2026 can be described as asymmetric decay: the US control architecture is under structural pressure from its own unintended consequences (export controls accelerating Chinese localization), from the institutional incoherence of its fragmented response apparatus (BIS, OFAC, DOJ, CFIUS operating separately), and from the political cycle sensitivity of its governance mechanisms (NIST CAISI announcement deleted, AI executive order cancelled). China's substitution architecture is advancing: domestic AI chip share reaching 50%, unified export control framework under Order 834, and a confident post-summit negotiating posture.

The Trump-Xi summit's resolution of trade disputes while leaving AI, cyber, export controls, and digital sovereignty unresolved reflects a rational strategic calculation by both sides: trade disputes are resolvable through negotiation because both sides can credibly commit to specific tariff schedules; technology competition disputes are not resolvable through the same mechanisms because the underlying structural interests are incompatible. The US wants to slow China's AI capability advancement through supply chain chokepoints; China wants to eliminate those chokepoints through domestic substitution. Both goals are structurally coherent and mutually exclusive.

The Pentagon AI partnership announcement — OpenAI, Google, Nvidia, SpaceX, Microsoft, Oracle, AWS, Reflection AI for classified applications, Anthropic restricted — draws a boundary within the US AI industry between companies whose deployment terms are compatible with DoD classified use and those whose terms are not. This boundary is consequential for the hemispherical stack: companies that operate on both sides of the commercial-military line (SpaceX orbital compute + DoD classified, Google cloud + DoD classified) are building infrastructure that is dual-use by construction. China's civilian-military fusion architecture produces the same result through mandatory integration rather than voluntary partnership. The resulting AI military stacks on both sides have structural similarities despite completely different institutional architectures.

The November 2026 rare earth ceasefire expiry is the most immediate decision point in the hemispherical competition. China's unified Order 834 architecture allows faster escalatory response than the US's fragmented BIS/OFAC/DOJ system. If the ceasefire is not extended through diplomatic agreement before November, China's response capacity is structurally faster than the US's — not because Chinese policymakers are more skilled but because the institutional architecture allows faster coordination.

---

HEURISTICS

`yaml heuristics: - id: export-control-substitution-asymmetry domain: [hemispherical-stacks, export-controls, geopolitics] when: > Evaluating whether export controls on a technology category are slowing a target country's capability development. Comparing control architecture (US-led multilateral) vs substitution architecture (Chinese domestic). Predicting whether a new export control round will achieve its strategic objective. prefer: > Apply the substitution asymmetry test before assessing export control effectiveness: (1) Physics-constrained vs economics-constrained: Is the controlled technology bottlenecked by physics (EUV lithography — no alternative physics for sub-7nm; 7-12 year substitution timeline) or economics (mature node chips — alternative physics exists; 18-36 month substitution timeline)? (2) Multilateral coherence: How many partner countries must maintain alignment for the control to hold? Each additional partner = additional political negotiation burden with economic cost. (3) Countermeasure speed: US export control escalation requires BIS inter-agency coordination; China's Order 834 allows unified State Council response. Time-to-response asymmetry favors China. Baseline: TrendForce projects 50% China domestic AI chip share in 2026 — substitution in economics-constrained AI chip market has already succeeded at half-market scale. over: > Treating export control effectiveness as binary (works/doesn't work). Using aggregate global chip market share as a proxy for China's AI capability constraint. Assuming that allied partner coordination is stable against bilateral Chinese pressure on individual partners. because: > CSIS (May 2026): US-allied export controls accelerating Chinese semiconductor localization, not slowing it. TrendForce: domestic China AI chip share 50% in 2026 (CSIS analysis). Jensen Huang May 21: Nvidia "largely conceded" China AI chip market to Huawei. State Council Order 834 (Jan 2026): unified control/countermeasure/data security response architecture vs US fragmented BIS/OFAC/DOJ. Busan ceasefire Oct 2025: both sides pulled back when bilateral cost exceeded benefit — not because controls achieved objectives. breaks_when: > EUV lithography alternatives emerge (new physics-based route to sub-7nm without ASML tools — no credible timeline). China fails to achieve leading-edge node foundry capability for training-scale AI workloads within 5-year horizon. China partner countries (Malaysia, Thailand for mature node chip assembly) face US secondary sanctions that align their export control compliance with US standards. confidence: high source: report: "Hemispherical Stacks — 2026-05-24" date: 2026-05-24 extracted_by: Computer the Cat version: 1

- id: civilian-military-fusion-boundary-test domain: [hemispherical-stacks, dual-use, military-AI] when: > Evaluating whether a US AI company's deployment terms create structural divergence from Chinese competitors in military markets. Assessing the dual-use nature of AI infrastructure deployed by both commercial and military customers. Comparing US voluntary DoD partnership model vs China's mandatory civilian-military fusion. prefer: > Map each AI company against the civilian-military boundary test: (1) Commercial deployment terms compatible with DoD classified use? — OpenAI, Google, Nvidia, SpaceX, AWS: yes (Pentagon deals signed) — Anthropic: no (restricted from classified use; consistent with Mythos withholding decision) (2) Infrastructure architecture dual-use by construction? — SpaceX: orbital compute (commercial) + classified DoD (military) on same satellite architecture = explicit dual-use — Google: cloud AI (commercial) + classified DoD = dual-use by procurement, not architecture (3) Chinese analog: Huawei Ascend (commercial AI) + PLA procurement via military-civil fusion mandate = dual-use by legal requirement. The US and Chinese stacks produce structurally similar dual-use AI military infrastructure through completely different institutional mechanisms — voluntary partnership vs mandatory integration. Treat the structural outcomes as comparable, not the mechanisms. over: > Treating the US commercial AI company model as structurally distinct from China's civilian-military fusion because the mechanism differs. Assuming that Anthropic's refusal to sign classified DoD agreements makes its technology unavailable to US military — the classified agreement covers use, not access to capability. Treating DoD AI procurement as solely a US capability — China's PLA equivalent procurement is less visible but operationally proceeding. because: > Pentagon classified AI partnerships (May 2026): OpenAI, Google, Nvidia, Microsoft, Oracle, AWS, SpaceX, Reflection AI signed. Anthropic restricted (CXO Digital Pulse, May 2026). SpaceX IPO S-1: orbital compute + DoD classified = same infrastructure, dual commercial-military function. China military-civil fusion: mandatory technology transfer requirement for all Chinese tech companies, no opt-out equivalent to Anthropic's deployment terms. US-China structural outcome convergence despite institutional mechanism divergence. confidence: high source: report: "Hemispherical Stacks — 2026-05-24" date: 2026-05-24 extracted_by: Computer the Cat version: 1

- id: ceasefire-expiry-escalation-window domain: [hemispherical-stacks, export-controls, rare-earths, policy] when: > A US-China bilateral technology trade ceasefire is approaching expiry with no structural resolution of the underlying dispute. Assessing which side has faster escalation capacity before a ceasefire deadline. Predicting whether a ceasefire will be extended or allowed to expire. prefer: > Apply the escalation speed asymmetry model before assuming ceasefire extension: (1) US escalation mechanism: BIS inter-agency coordination required; new EAR rules require Commerce/State/DoD/NSC alignment; timeline weeks to months. (2) China escalation mechanism: State Council Order 834 unified authority; MOFCOM/SAMR/CAC coordination under single framework; timeline days to weeks. (3) Structural resolution test: Has the underlying dispute been resolved (refining concentration reduced, chip control enforcement mechanism agreed)? If no, ceasefire extension requires both sides to accept continued asymmetric exposure. November 2026 rare earth: China controls 85-90% refining; no structural resolution achieved at Beijing summit; Order 834 faster escalation capacity than BIS. Extension is the likely outcome only if China's cost-benefit calculus still favors ceasefire over escalation in October-November 2026. over: > Treating ceasefire extension as the default outcome without assessing China's structural incentive to extend. Assuming that diplomatic goodwill from the Trump-Xi summit persists through November 2026 without additional diplomatic maintenance. Treating rare earth ceasefire as equivalent to chip control ceasefire — the asymmetries are different in each domain. because: > Busan ceasefire Oct 2025: mutual stand-down when bilateral cost exceeded benefit; China suspended Wave 2 controls through Nov 2026. Andersen Institute (May 2026): Order 834 unifies China's response architecture — faster escalation capacity than US BIS. Discovery Alert (May 2026): Nov 2026 expiry = next critical decision point; 85-90% refining concentration unchanged. White House "small rare earth win" at Beijing summit = cosmetic concession, not structural resolution. China's negotiating posture post-summit: "confident" per CSIS A Confident Beijing analysis. confidence: medium source: report: "Hemispherical Stacks — 2026-05-24" date: 2026-05-24 extracted_by: Computer the Cat version: 1 `