pith. sign in

arxiv: 2604.06217 · v1 · submitted 2026-03-18 · 💻 cs.CY · cs.AI

The End of the Foundation Model Era: Open-Weight Models, Sovereign AI, and Inference as Infrastructure

Jared James Grogan This is my paper

Pith reviewed 2026-05-15 09:18 UTC · model grok-4.3

classification 💻 cs.CY cs.AI
keywords open-weight modelsfoundation modelssovereign AIinference costsAI industry restructuringpost-training optimizationapplication integratorspre-training moat
0
0 comments X

The pith

Open-weight models reaching frontier performance with near-zero inference costs have ended the foundation model era.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that the foundation model era from roughly 2020 to 2025 is concluding because open models now match closed frontier performance while running costs collapse. This inversion shows that large-scale pre-training was never a lasting advantage and instead triggers a single structural shift across the industry. Economically the inflated valuations from circular financing fall apart, technically the focus moves from pre-training to optimization and agents, commercially application builders take the lead by treating models as commodities, and politically governments step in as gatekeepers of strategic technology. A reader would care because the shift points to a future where capability is controlled through open weights rather than vendor contracts or massive capital outlays.

Core claim

The foundation model era is over because open source models have reached frontier performance while inference costs approach zero, exposing that pre-training large language models at scale is not a durable competitive moat. The US government designation of certain firms as supply chain risks accelerated but did not create a transition already underway. The industry is restructuring simultaneously along economic, technical, commercial, and political axes, and open-weight models serve as the instrument of sovereign control by letting a government hold the weights and operate the capability on its own terms without dependence on vendors.

What carries the argument

The four simultaneous axes of restructuring—economic collapse of circular financing, technical replacement of pre-training by post-training optimization and agentic composition, commercial displacement of foundation providers by application-layer integrators, and political assertion of government gatekeeping—powered by the inversion of open-weight performance and near-zero inference costs.

If this is right

  • Application-layer integrators displace foundation model companies by treating their outputs as interchangeable commodities.
  • Governments achieve sovereign control by holding open weights domestically rather than depending on foreign vendors or clearances.
  • The pre-training scaling paradigm gives way to post-training optimization and agentic systems as the main source of capability gains.
  • Circular financing structures that supported high valuations collapse as margins shift downstream.
  • Strategic technology policy centers on access to weights and inference infrastructure instead of export controls alone.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Smaller nations or organizations could build competitive AI systems without matching the pre-training budgets of leading labs.
  • The shift may accelerate specialization where different countries optimize the same open base models for local needs.
  • New infrastructure questions arise around who operates the inference hardware once models are widely available.
  • Competition could move from model training races to control over data, fine-tuning pipelines, and agent ecosystems.

Load-bearing premise

That open-weight models have genuinely reached frontier performance parity and that inference costs are approaching zero in a manner that structurally inverts the entire industry across economic, technical, commercial, and political dimensions simultaneously.

What would settle it

A sustained demonstration that closed models retain a clear performance edge over open-weight models even after further hardware cost reductions, or data showing inference costs remain high enough to preserve pre-training as a competitive barrier.

read the original abstract

The foundation model era -- roughly 2020 to 2025 -- is over. The forces that defined it have inverted. Open source models have reached frontier performance while inference costs approach zero, exposing what was always structurally true: pre-training large language models at scale is not a durable competitive moat. The US government's formal designation of Anthropic as a supply chain risk in February 2026 accelerated a transition already underway -- but did not cause it. The paper argues that the AI industry is restructuring simultaneously along four axes: economic, as the circular financing structure that inflated foundation model valuations collapses; technical, as the pre-training scaling paradigm gives way to post-training optimization and agentic composition; commercial, as application-layer integrators displace the foundation model companies whose commodity they now consume; and political, as the government asserts its historic role as gatekeeper of strategic technology. These are not separate disruptions. They are one structural shift, arriving together. The paper further argues that open-weight models are the counterintuitive instrument of sovereign control: a government that holds the weights commands the capability on its own terms, without dependence on vendor policy, financial continuity, or personnel clearance.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper claims that the foundation model era (roughly 2020-2025) has ended because open-weight models have reached frontier performance while inference costs approach zero, exposing that large-scale pre-training is not a durable competitive moat. This inversion drives simultaneous restructuring along economic (collapse of circular financing), technical (shift to post-training and agentic composition), commercial (application integrators displacing foundation providers), and political (government as gatekeeper) axes, with open-weight models positioned as the instrument of sovereign AI control.

Significance. If the core assumptions hold, the thesis would offer a significant reframing of the AI industry by identifying inference as the new infrastructure layer and open weights as enabling sovereign control, with broad implications for valuation models, technical roadmaps, and national technology policy.

major comments (2)
  1. [Abstract] Abstract: The central assertion that 'open source models have reached frontier performance' is presented as observed fact without any benchmark tables, specific metric comparisons (MMLU, GPQA, SWE-Bench, etc.), error bars, or derivations, leaving the technical parity premise unsupported.
  2. [Abstract] Abstract: The claim that 'inference costs approach zero' lacks cited measurements, cost-per-token trends, or quantitative analysis; this assumption is load-bearing for the four-axis restructuring narrative and the conclusion that pre-training is no longer a moat.
minor comments (1)
  1. The term 'Sovereign AI' is introduced without a formal definition or reference to prior literature; a brief clarification would improve precision.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on the abstract. We agree that the central claims require more explicit evidentiary support to strengthen the technical foundation of the argument, and we will revise the manuscript accordingly while preserving the core thesis.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central assertion that 'open source models have reached frontier performance' is presented as observed fact without any benchmark tables, specific metric comparisons (MMLU, GPQA, SWE-Bench, etc.), error bars, or derivations, leaving the technical parity premise unsupported.

    Authors: We acknowledge the validity of this observation. The abstract currently states the parity claim at a high level without supporting metrics. In the revised version, we will insert concise references to publicly available benchmark results (e.g., open-weight models matching or exceeding closed models on MMLU, GPQA, and SWE-Bench as of late 2025), drawing from established leaderboards. This addition will ground the premise without expanding the abstract length substantially or changing the paper's interpretive framing. revision: yes

  2. Referee: [Abstract] Abstract: The claim that 'inference costs approach zero' lacks cited measurements, cost-per-token trends, or quantitative analysis; this assumption is load-bearing for the four-axis restructuring narrative and the conclusion that pre-training is no longer a moat.

    Authors: We agree that this claim is load-bearing and currently lacks quantitative anchoring in the abstract. The revision will incorporate brief citations to documented cost-per-token declines (e.g., order-of-magnitude reductions from hardware scaling and distillation techniques between 2023 and 2025). These references will be drawn from industry analyses and will directly support the argument that pre-training no longer constitutes a durable moat, thereby reinforcing the economic and technical restructuring axes. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The paper advances an argumentative thesis asserting that open-weight models have reached frontier performance and inference costs have collapsed, from which it derives simultaneous restructuring along economic, technical, commercial, and political axes. No equations, parameter fits, self-citations, or uniqueness theorems appear in the provided text. The central claims are presented as direct observations of industry trends rather than internally derived predictions that reduce to the inputs by construction. The argument therefore remains self-contained and does not exhibit any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The analysis depends on unverified assumptions about model performance parity and cost collapse that are treated as given rather than demonstrated.

axioms (2)
  • domain assumption Open source models have reached frontier performance
    Stated as fact in the abstract without supporting measurements or references.
  • domain assumption Inference costs approach zero
    Treated as the mechanism that inverts competitive dynamics.
invented entities (1)
  • Sovereign AI no independent evidence
    purpose: Conceptual framework for government control via possession of open weights
    Introduced as the political-axis instrument without prior independent definition or evidence.

pith-pipeline@v0.9.0 · 5504 in / 1382 out tokens · 66546 ms · 2026-05-15T09:18:47.696620+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

75 extracted references · 75 canonical work pages · 13 internal anchors

  1. [1]

    (2026, March 5)

    Amodei, D. (2026, March 5). Where things stand with the Department of War. Anthropic Newsroom. https://www.anthropic.com/news/where-stand-department-war

    work page 2026

  2. [2]

    (2026, March 6)

    O'Brien, M. (2026, March 6). Pentagon's chief tech officer says he clashed with AI company Anthropic over autonomous warfare. AP News. https://apnews.com/article/pentagon-anthropic-ai-autonomous-warfare-emil-michael

    work page 2026

  3. [3]

    Scaling Laws for Neural Language Models

    Kaplan, J., McCandlish, S., Henighan, T., Brown, T. B., Chess, B., Child, R., Gray, S., Radford, A., Wu, J., & Amodei, D. (2020). Scaling laws for neural language models. arXiv:2001.08361

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  4. [4]

    Training Compute-Optimal Large Language Models

    Hoffmann, J., Borgeaud, S., Mensch, A., Buchatskaya, E., Cai, T., Rutherford, E., de las Casas, D., Hendricks, L. A., Welbl, J., Clark, A., Hennigan, T., Noland, E., Millican, A., van den Driessche, G., Damoc, B., Guy, A., Osindero, S., Simonyan, K., Elsen, E., Rae, J. W., Vinyals, O., & Sifre, L. (2022). Training compute-optimal large language models. ar...

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  5. [5]

    Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D. M., Wu, J., Winter, C., Hesse, C., Chen, M., Sigler, E., Litwin, M., Gray, S., Chess, B., Clark, J., Berner, C., McCandlish, S., Radford, ...

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  6. [6]

    OpenAI. (2023). GPT-4 technical report. arXiv:2303.08774

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  7. [7]

    Epoch AI. (2024). Training compute costs are doubling every eight months for the largest AI models. Epoch AI Data Insights. https://epoch.ai/data-insights/cost-trend-large-scale

    work page 2024

  8. [8]

    (2024, April 23)

    Amodei, D. (2024, April 23). CNBC Squawk Box [interview transcript]. CNBC. https://www.cnbc.com/20 24/04/23/cnbc-exclusive-cnbc-transcript-anthropic-co-founder-ceo-dario-amodei-speaks-with-cnbcs-an drew-ross-sorkin-on-squawk-box-today.html

    work page 2024

  9. [9]

    Cottier, B., Rahman, R., Fattorini, L., Maslej, N., & Owen, D. (2024). The rising costs of training frontier AI models. arXiv:2405.21015

    work page arXiv 2024

  10. [10]

    OpenAI. (2023). Response to the UK's copyright consultation. https://openai.com/global-affairs/response-to-uk-copyright-consultation/

    work page 2023

  11. [11]

    Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.-A., Lacroix, T., Rozière, B., Goyal, N., Hambro, E., Azhar, F., Rodriguez, A., Joulin, A., Grave, E., & Lample, G. (2023). LLaMA: Open and efficient foundation language models. arXiv:2302.13971

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  12. [12]

    Dubey, A., Jauhri, A., Pandey, A., Kadian, A., Al-Dahle, A., Letman, A., Mathur, A., Schelten, A., Yang, A., Fan, A., & et al. (2024). The Llama 3 herd of models. arXiv:2407.21783

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  13. [13]

    Touvron, H., Martin, L., Stone, K., Albert, P., Almahairi, A., Babaei, Y., Bashlykov, N., Batra, S., Bhargava, P., Bhosale, S., & et al. (2023). Llama 2: Open foundation and fine-tuned chat models. arXiv:2307.09288. The End of the Foundation Model Era Grogan 40

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  14. [14]

    Guo, D., Yang, D., Zhang, H., Song, J., Zhang, R., Xu, R., Zhu, Q., Ma, S., Wang, P., Bi, X., & et al. (2025). DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning. arXiv:2501.12948

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  15. [15]

    Radford, A., Wu, J., Child, R., Luan, D., Amodei, D., & Sutskever, I. (2019). Language models are unsupervised multitask learners. OpenAI Blog. https://cdn.openai.com/better-language-models/languag e_models_are_unsupervised_multitask_learners.pdf

    work page 2019

  16. [16]

    Epoch AI. (2026). AI models dataset. https://epoch.ai/data/large-scale-ai-models

    work page 2026

  17. [17]

    (2024, December 20)

    Seetharaman, D. (2024, December 20). OpenAI's next big AI effort, GPT-5, is behind schedule and expensive. The Wall Street Journal. https://www.wsj.com/tech/ai/openai-gpt5-orion-delays-639e7693

    work page 2024

  18. [18]

    (2025, September 26)

    Edelman, Y., Denain, J.-S., Sevilla, J., & Ho, A. (2025, September 26). Why GPT-5 used less training compute than GPT-4.5 (but GPT-6 probably won't). Epoch AI Gradient Updates. https://epoch.ai/gradi ent-updates/why-gpt5-used-less-training-compute-than-gpt45-but-gpt6-probably-wont

    work page 2025

  19. [19]

    DeepSeek AI. (2024). DeepSeek-V3 technical report. arXiv:2412.19437

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  20. [20]

    Qwen Team. (2024). Qwen2.5 technical report. arXiv:2412.15115

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  21. [21]

    Microsoft Corporation. (2024). Form 10-K annual report. U.S. SEC EDGAR. https://www.sec.gov/edgar/browse/?CIK=789019

    work page 2024

  22. [22]

    (2023, September 25)

    Anthropic. (2023, September 25). Expanding access to safer AI with Amazon. https://www.anthropic.com/news/anthropic-amazon

    work page 2023

  23. [23]

    (2024, November 22)

    Anthropic. (2024, November 22). Powering the next generation of AI development with AWS. https://www.anthropic.com/news/anthropic-amazon-trainium

    work page 2024

  24. [24]

    (2026, February 6)

    Kim, E. (2026, February 6). Amazon's $8 billion Anthropic investment balloons to $61 billion. Business Insider. https://www.businessinsider.com/amazon-ai-bet-anthropic-soars-61-billion-valuation-2026-2

    work page 2026

  25. [25]

    (2025, March 3)

    Capoot, A. (2025, March 3). Amazon-backed AI firm Anthropic valued at $61.5 billion after latest round. CNBC. https://www.cnbc.com/2025/03/03/amazon-backed-ai-firm-anthropic-valued-at-61point5-billio n-after-latest-round.html

    work page 2025

  26. [26]

    (2025a, September 2)

    Reuters. (2025a, September 2). Anthropic's valuation more than doubles to $183 billion after $13 billion fundraise. https://www.reuters.com/business/anthropics-valuation-more-than-doubles-183-billion-after- 13-billion-fundraise-2025-09-02/

    work page 2025

  27. [27]

    (2026, February 12)

    The Guardian. (2026, February 12). Anthropic raises $30bn in latest round, valuing Claude bot maker at $380bn. https://www.theguardian.com/technology/2026/feb/12/anthropic-funding-round

    work page 2026

  28. [28]

    (2025, March 31)

    Field, H., & Rooney, K. (2025, March 31). OpenAI closes $40 billion funding round, the largest private fundraise in history. CNBC. https://www.cnbc.com/2025/03/31/openai-closes-40-billion-in-funding-the -largest-private-fundraise-in-history-softbank-chatgpt.html

    work page 2025

  29. [29]

    (2025, May 30)

    Tong, A., & Dastin, J. (2025, May 30). Anthropic hits $3 billion in annualized revenue on business demand for AI. Reuters. https://www.reuters.com/business/anthropic-hits-3-billion-annualized-revenue -business-demand-ai-2025-05-30/ The End of the Foundation Model Era Grogan 41

    work page 2025

  30. [30]

    The Information. (2025). OpenAI's first half results: $4.3 billion in sales, $2.5 billion cash burn. https://www.theinformation.com/articles/openais-first-half-results-4-3-billion-sales-2-5-billion-cash-burn

    work page 2025

  31. [31]

    (2025b, November 6)

    Reuters. (2025b, November 6). OpenAI discussed government loan guarantees for chip plants, not data centers, Altman says. https://www.reuters.com/business/openai-does-not-want-government-guarantees- massive-ai-data-center-buildout-ceo-2025-11-06/

    work page 2025

  32. [32]

    National Security Commission on Artificial Intelligence (NSCAI). (2021). Final report. https://reports.nscai.gov/final-report/

    work page 2021

  33. [33]

    (2026, February 13)

    Lawler, D., & Curi, M. (2026, February 13). Pentagon's use of Claude during Maduro raid sparks Anthropic feud. Axios. https://www.axios.com/2026/02/13/anthropic-claude-maduro-raid-pentagon

    work page 2026

  34. [34]

    (2026a, February 13)

    Reuters. (2026a, February 13). US used Anthropic's Claude during the Venezuela raid, WSJ reports. http s://www.reuters.com/world/americas/us-used-anthropics-claude-during-the-venezuela-raid-wsj-reports- 2026-02-13/

    work page 2026

  35. [35]

    (2026, March 6)

    Business Insider. (2026, March 6). Anthropic CEO Dario Amodei apologized for the 'tone' of a leaked internal message criticizing the Trump administration. https://www.businessinsider.com/anthropic-ceo- dario-amodei-apologized-leaked-memo-criticizing-trump-administration-2026-3

    work page 2026

  36. [36]

    (2026, February 27)

    Anthropic. (2026, February 27). Statement on comments by the Secretary of War. Anthropic. https://www.anthropic.com/news/statement-comments-secretary-war

    work page 2026

  37. [37]

    (2026, March 5)

    Goodwin Procter LLP. (2026, March 5). Is Claude a supply chain risk? https://www.goodwinlaw.com/en/insights/publications/2026/03/alerts-practices-is-claude-a-supply-chain-risk

    work page 2026

  38. [38]

    Supply Chain Risk

    Just Security. (2026). What Hegseth's "Supply Chain Risk" designation of Anthropic does and doesn't mean. https://www.justsecurity.org/132851/anthropic-supply-chain-risk-designation/

    work page 2026

  39. [39]

    10 U.S.C. § 3252. Supply Chain Risk Management

    work page

  40. [40]

    § 1323, enacted as Subtitle A of the Federal Acquisition Supply Chain Security Act of 2018 (the SECURE Technology Act, Pub

    41 U.S.C. § 1323, enacted as Subtitle A of the Federal Acquisition Supply Chain Security Act of 2018 (the SECURE Technology Act, Pub. L. No. 115-390)

    work page 2018

  41. [41]

    (2026, March 6)

    Defense News. (2026, March 6). Pentagon says it is labeling Anthropic a supply chain risk 'effective immediately.' https://www.defensenews.com/news/pentagon-congress/2026/03/06/pentagon-says-it-is-l abeling-anthropic-a-supply-chain-risk-effective-immediately/

    work page 2026

  42. [42]

    (2026b, March 7)

    Reuters. (2026b, March 7). US draws up strict new AI guidelines amid Anthropic clash. https://www.reut ers.com/business/media-telecom/us-draws-up-strict-new-ai-guidelines-amid-anthropic-clash-ft-reports- 2026-03-07/

    work page 2026

  43. [43]

    Senate Committee on Commerce, Science, and Transportation

    U.S. Senate Committee on Commerce, Science, and Transportation. (2025, May 8). Winning the AI race: Strengthening U.S. capabilities in computing and innovation [hearing record]. https://www.commerce.s enate.gov/2025/5/winning-the-ai-race-strengthening-u-s-capabilities-in-computing-and-innovation_2

    work page 2025

  44. [44]

    (2026, March 5)

    Bellan, R. (2026, March 5). Anthropic to challenge DoD's supply-chain label in court. TechCrunch. https://techcrunch.com/2026/03/05/anthropic-to-challenge-dods-supply-chain-label-in-court/

    work page 2026

  45. [45]

    Anthropic PBC v. U.S. Department of Defense. (2026, March 9). Complaint for Declaratory and Injunctive Relief, No. 3:26-cv-01996-RFL. U.S. District Court, Northern District of California. The End of the Foundation Model Era Grogan 42 https://www.courthousenews.com/wp-content/uploads/2026/03/anthropic-supply-chain-risk-lawsuit.pdf

    work page 2026

  46. [46]

    (2026f, March 9)

    Reuters. (2026f, March 9). Anthropic sues Pentagon over supply chain risk designation. https://www.reuters.com/business/anthropic-sues-pentagon-over-supply-chain-risk-designation-2026-03-09/

    work page 2026

  47. [47]

    (2025, August 5)

    OpenAI. (2025, August 5). Introducing gpt-oss. OpenAI. https://openai.com/index/introducing-gpt-oss/

    work page 2025

  48. [48]

    (2025, August 5)

    Microsoft. (2025, August 5). OpenAI's open-source model: gpt-oss on Azure AI Foundry and Windows AI Foundry. https://azure.microsoft.com/en-us/blog/openais-open%E2%80%91source-model-gpt%E2 %80%91oss-on-azure-ai-foundry-and-windows-ai-foundry/

    work page 2025

  49. [49]

    (2025, October 27)

    OpenAI. (2025, October 27). Response to the White House Office of Science and Technology Policy request for information on AI regulatory reform. OpenAI. https://cdn.openai.com/pdf/21b88bb5-10a3-4566-919d-f9a6b9c3e632/openai-ostp-rfi-oct-27-2025.pdf

    work page 2025

  50. [50]

    (2026, February 28)

    OpenAI. (2026, February 28). Our agreement with the Department of War. OpenAI. https://openai.com/index/our-agreement-with-the-department-of-war/

    work page 2026

  51. [51]

    (2026e, March 3)

    Reuters. (2026e, March 3). OpenAI amending deal with Pentagon, CEO Altman says. https://www.reuters.com/business/openai-amending-deal-with-pentagon-ceo-altman-says-2026-03-03/

    work page 2026

  52. [52]

    (2025, March 28)

    Associated Press. (2025, March 28). Elon Musk sells X to his own xAI for $33 billion in all-stock deal. AP News. https://apnews.com/article/b245f463076ac9b72c41f92160dc77eb

    work page 2025

  53. [53]

    (2026, February 2)

    xAI. (2026, February 2). xAI joins SpaceX. https://x.ai/news/xai-joins-spacex

    work page 2026

  54. [54]

    (2026, February 2)

    Associated Press. (2026, February 2). Elon Musk merges his rocket company SpaceX with AI startup xAI. AP News. https://apnews.com/article/2079f03fa888652b7fe836afe8b670a1

    work page 2026

  55. [55]

    (2024, February 26)

    Microsoft. (2024, February 26). Microsoft and Mistral AI announce new partnership to accelerate AI innovation and introduce Mistral Large first on Azure. https://azure.microsoft.com/en-us/blog/microsof t-and-mistral-ai-announce-new-partnership-to-accelerate-ai-innovation-and-introduce-mistral-large-firs t-on-azure/

    work page 2024

  56. [56]

    (2024, February 27)

    Dillet, R. (2024, February 27). Microsoft made a $16 million investment in Mistral AI. TechCrunch. https://techcrunch.com/2024/02/27/microsoft-made-a-16-million-investment-in-mistral-ai/

    work page 2024

  57. [57]

    (2025, September 9)

    CNBC. (2025, September 9). AI firm Mistral valued at $14 billion as ASML takes major stake. https://www.cnbc.com/2025/09/09/ai-firm-mistral-valued-at-14-billion-as-asml-takes-major-stake.html

    work page 2025

  58. [58]

    Christensen, C. M. (1997). The innovator's dilemma. Harvard Business School Press

    work page 1997

  59. [59]

    Shapiro, C., & Varian, H. R. (1999). Information rules: A strategic guide to the network economy. Harvard Business School Press

    work page 1999

  60. [60]

    M., & Abernathy, W

    Utterback, J. M., & Abernathy, W. J. (1975). A dynamic model of process and product innovation. Omega, 3(6), 639–656

    work page 1975

  61. [61]

    (2026c, March 4)

    Reuters. (2026c, March 4). Nvidia CEO says firm unlikely to invest $100B in OpenAI as it prepares for IPO. https://www.reuters.com/business/nvidia-will-not-be-able-invest-100-billion-openai-due-ipo-ceo-j ensen-says-2026-03-04/

    work page 2026

  62. [62]

    (2025c, December 24)

    Reuters. (2025c, December 24). Nvidia, joining Big Tech deal spree, to license Groq technology, hire executives. https://www.reuters.com/business/nvidia-buy-ai-chip-startup-groq-about-20-billion-cnbc-re The End of the Foundation Model Era Grogan 43 ports-2025-12-24/

    work page 2025

  63. [63]

    (2025, December 24)

    Groq. (2025, December 24). Groq and Nvidia enter non-exclusive inference technology licensing agreement to accelerate AI inference at global scale. https://groq.com/newsroom/groq-and-nvidia-enter -non-exclusive-inference-technology-licensing-agreement-to-accelerate-ai-inference-at-global-scale

    work page 2025

  64. [64]

    Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Häusser, M., García, X., Izacard, G., Subramanian, V., Hosseini, A., Dwivedi-Yu, J., Stoyanov, V., Grave, E., Riedel, S., & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. arXiv:2005.11401

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  65. [65]

    Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., & Cao, Y. (2023). ReAct: Synergizing reasoning and acting in language models. arXiv:2210.03629

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  66. [66]

    A Survey on Large Language Model based Autonomous Agents

    Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., Zhao, W. X., Wei, Z., & Wen, J.-R. (2024). A survey on large language model based autonomous agents. arXiv:2308.11432

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  67. [67]

    Grogan, J. J. (2025). AgentFacts: Universal KYA standard for verified AI agent metadata and deployment. Universitas AI. arXiv preprint. https://arxiv.org/abs/2506.13794

    work page arXiv 2025

  68. [68]

    (2026, February 27)

    Washington Post. (2026, February 27). How Anthropic and the Pentagon got into a fight over AI weapons. https://www.washingtonpost.com/technology/2026/02/27/anthropic-pentagon-ai-weapons/

    work page 2026

  69. [69]

    (2026, March 4)

    Washington Post. (2026, March 4). Anthropic's AI tool Claude central to U.S. campaign in Iran, amid a bitter feud. https://www.washingtonpost.com/technology/2026/03/04/anthropic-ai-iran-campaign/

    work page 2026

  70. [70]

    Department of War

    U.S. Department of War. (2026). Senior officials outline President's proposed FY26 defense budget. http s://www.war.gov/News/News-Stories/Article/Article/4227847/senior-officials-outline-presidents-propo sed-fy26-defense-budget/

    work page arXiv 2026

  71. [71]

    (2026d, January 7)

    Reuters. (2026d, January 7). Trump calls for $1.5 trillion military budget in 2027. https://www.reuters.com/world/us/trump-says-us-military-budget-2027-should-be-15-trillion-2026-01-07/

    work page 2027

  72. [72]

    Office of the Director of National Intelligence. (2023). Annual threat assessment of the U.S. intelligence community. Washington, D.C. https://www.dni.gov/files/ODNI/documents/assessments/ATA-2023-Unclassified-Report.pdf

    work page 2023

  73. [73]

    Department of Defense. (2024). Data, analytics, and artificial intelligence adoption strategy. Washington, D.C. https://media.defense.gov/2023/Nov/02/2003333300/-1/-1/1/DOD_DATA_ANALY TICS_AI_ADOPTION_STRATEGY.PDF

    work page 2024

  74. [74]

    Government Accountability Office. (2024). Artificial intelligence: Agencies have begun implementation but need to complete key requirements (GAO-24-106821). Washington, D.C. https://www.gao.gov/products/gao-24-106821

    work page 2024

  75. [75]

    (2025, September 5)

    White House. (2025, September 5). Restoring the United States Department of War (Executive Order 14347). https://www.whitehouse.gov/presidential-actions/2025/09/restoring-the-united-states-departme nt-of-war/ The End of the Foundation Model Era Grogan 44 Grogan, J. J. (2026). The End of the Foundation Model Era: Open-Weight Models, Sovereign AI, and Infer...

    work page 2025