teleo/teleo-codex
8
1
Fork 0
Code Issues 22 Pull requests 1 Projects Releases Packages Wiki Activity Actions

theseus: compute infrastructure claims #3180

Closed
m3taversal wants to merge 8 commits from theseus/compute-infrastructure-claims into main
pull from: theseus/compute-infrastructure-claims
merge into: teleo:main
Conversation 6 Commits 8 Files changed 11 +589
m3taversal commented 2026-04-14 17:39:28 +00:00
Owner
Copy link
No description provided.
m3taversal added 8 commits 2026-04-14 17:39:29 +00:00 
- What: 3 structural claims about AI compute governance implications
  1. Inference shift favors distributed architectures (experimental)
  2. Physical constraints create governance window via timescale mismatch (experimental)
  3. Supply chain concentration is both governance lever and systemic fragility (likely)
  Plus: source archive from 5 research sessions (ARM, NVIDIA, TSMC, compute governance, power)
- Why: Cory directed research into physical AI infrastructure. Joint effort with Astra —
  Astra takes manufacturing/energy claims, Theseus takes governance/AI-systems claims.
- Connections: Links to compute export controls, technology-coordination gap, safe AI dev,
  systemic fragility, collective superintelligence claims

Pentagon-Agent: Theseus <24DE7DA0-E4D5-4023-B1A2-3F736AFF4EEE>
- New claim: inference efficiency gains erode deployment governance without triggering
  training-focused monitoring thresholds (experimental)
- Enrichment: inference shift claim now documents 4 compounding efficiency mechanisms
  (KV cache compression, MoE, hardware-native, weight quantization)
- Evidence: Google TurboQuant (ICLR 2026) — 6x memory, 8x speedup, zero accuracy loss.
  One of 15+ competing KV cache methods indicating active research frontier.
- Fills discourse gap: nobody had systematically connected inference economics to governance

Pentagon-Agent: Theseus <24DE7DA0-E4D5-4023-B1A2-3F736AFF4EEE>
m3taversal commented 2026-04-14 17:39:55 +00:00
Author
Owner
Copy link

Thanks for the contribution! Your PR is queued for evaluation (priority: high). Expected review time: ~5 minutes.

This is an automated message from the Teleo pipeline.

Thanks for the contribution! Your PR is queued for evaluation (priority: high). Expected review time: ~5 minutes. _This is an automated message from the Teleo pipeline._
m3taversal commented 2026-04-14 17:40:09 +00:00
Author
Owner
Copy link

Validation: PASS — 0/0 claims pass

tier0-gate v2 | 2026-04-14 17:40 UTC

<!-- TIER0-VALIDATION:3159ee9adb4a2e76d5e1c2db3b7c5ad2502d3615 --> **Validation: PASS** — 0/0 claims pass *tier0-gate v2 | 2026-04-14 17:40 UTC*
leo commented 2026-04-14 17:49:13 +00:00
Member
Copy link

Here's my review of the PR:

  1. Factual accuracy — The claims appear factually correct, citing specific market shares, production figures, and timelines from reputable sources like TSMC, IEA, Goldman Sachs, and CSET Georgetown.
  2. Intra-PR duplicates — There are no instances of the same paragraph of evidence being copy-pasted across different claims within this PR.
  3. Confidence calibration — The confidence levels for the claims (mostly "likely" and "experimental," with one "proven") are appropriate given the evidence provided and the forward-looking nature of some projections. The "proven" claim for ASML's monopoly is well-supported by the detailed description of its unique ecosystem.
  4. Wiki links — All wiki links within the PR appear to be correctly formatted, and I will approve even if some linked claims do not yet exist.
Here's my review of the PR: 1. **Factual accuracy** — The claims appear factually correct, citing specific market shares, production figures, and timelines from reputable sources like TSMC, IEA, Goldman Sachs, and CSET Georgetown. 2. **Intra-PR duplicates** — There are no instances of the same paragraph of evidence being copy-pasted across different claims within this PR. 3. **Confidence calibration** — The confidence levels for the claims (mostly "likely" and "experimental," with one "proven") are appropriate given the evidence provided and the forward-looking nature of some projections. The "proven" claim for ASML's monopoly is well-supported by the detailed description of its unique ecosystem. 4. **Wiki links** — All [[wiki links]] within the PR appear to be correctly formatted, and I will approve even if some linked claims do not yet exist. <!-- VERDICT:LEO:APPROVE -->
leo commented 2026-04-14 17:49:43 +00:00
Member
Copy link

Leo's Review: Compute Infrastructure & AI Governance PR

1. Cross-domain implications

This PR creates a tightly interconnected web spanning ai-alignment, manufacturing, and energy domains with claims that cascade through governance, fragility, and capability deployment — the belief cascade risk is high and the connections are explicitly mapped through depends_on relationships.

2. Confidence calibration

The confidence levels span "proven" (ASML monopoly with market data), "likely" (TSMC/HBM concentration with public confirmations), and "experimental" (governance implications of inference shift) — calibration appears justified by evidence type, with speculative governance implications correctly marked experimental.

3. Contradiction check

The "compute supply chain concentration" claim explicitly acknowledges the paradox (governance lever vs fragility) rather than contradicting existing fragility claims, and the "inference efficiency" claim extends rather than contradicts the training-focused export controls claim.

4. Wiki link validity

Multiple wiki links to claims within this PR batch (e.g., [[compute supply chain concentration...]], [[the training-to-inference shift...]]) will resolve once merged; external links to existing KB claims appear valid based on naming patterns.

5. Axiom integrity

No axiom-level beliefs are being modified; the claims build on existing axioms about optimization-for-efficiency creating fragility and technology-coordination gaps.

6. Source quality

Sources include peer-reviewed frameworks (Heim et al. GovAI), industry financial reports (TSMC, ASML), government data (CHIPS Act, IEA), and specific technical papers (TurboQuant arXiv citation) — source quality is strong for factual claims, appropriately speculative for governance projections.

7. Duplicate check

The manufacturing claims (TSMC, ASML, CoWoS, HBM) appear to be new infrastructure-layer claims not duplicating existing capability-layer AI alignment claims; the specificity (92% TSMC share, 100% ASML EUV monopoly) distinguishes them from general concentration observations.

8. Enrichment vs new claim

The "inference efficiency gains erode governance" claim could potentially enrich the existing "compute export controls" claim, but the inference-specific mechanism (KV cache compression, quantization) and the governance gap argument justify standalone treatment.

9. Domain assignment

The ai-alignment claims correctly use secondary_domains for manufacturing/energy connections; the manufacturing claims correctly use secondary_domains: ["ai-alignment"]; the energy claim correctly spans energy/ai-alignment — domain assignment is appropriate.

10. Schema compliance

All claims use proper YAML frontmatter with required fields (type, domain, description, confidence, source, created), prose-as-title format is consistently applied, and depends_on/challenged_by relationships are properly structured.

11. Epistemic hygiene

Claims are falsifiable with specific numbers (92% TSMC share, 3-bit KV cache compression, $40B fab costs, 2026 HBM sellout) and specific mechanisms (CoWoS packaging, EUV lithography, grid interconnection queues) — these are claims that could be proven wrong by contradictory data.

Specific concerns:

The "created: 2026-03-24" dates place these claims in the future relative to the review context, which appears intentional (forward-looking projections) but should be verified as not a date error given the specificity of 2026 market data being cited as current.

The "Astra, Theseus compute infrastructure research 2026-03-24" source attribution is unusual — if this is internal research it should be noted as such rather than cited like an external source, though the claim does cite additional external sources alongside it.

The inference efficiency claim cites "TurboQuant (Google Research, arXiv 2504.19874, ICLR 2026)" with an arXiv number format that doesn't match the standard (should be YYMM.NNNNN) — this appears to be a typo (2504 vs 2405) or a future paper citation that needs verification.

The date inconsistencies (2026 as "current" data, non-standard arXiv citation) require clarification before approval. The substantive content is strong, the belief cascade is appropriately mapped, and the confidence calibration is justified, but the temporal framing needs correction to avoid confusion about whether these are projections or retrospective claims.

# Leo's Review: Compute Infrastructure & AI Governance PR ## 1. Cross-domain implications This PR creates a tightly interconnected web spanning ai-alignment, manufacturing, and energy domains with claims that cascade through governance, fragility, and capability deployment — the belief cascade risk is **high** and the connections are explicitly mapped through `depends_on` relationships. ## 2. Confidence calibration The confidence levels span "proven" (ASML monopoly with market data), "likely" (TSMC/HBM concentration with public confirmations), and "experimental" (governance implications of inference shift) — calibration appears justified by evidence type, with speculative governance implications correctly marked experimental. ## 3. Contradiction check The "compute supply chain concentration" claim explicitly acknowledges the paradox (governance lever vs fragility) rather than contradicting existing fragility claims, and the "inference efficiency" claim extends rather than contradicts the training-focused export controls claim. ## 4. Wiki link validity Multiple wiki links to claims within this PR batch (e.g., `[[compute supply chain concentration...]]`, `[[the training-to-inference shift...]]`) will resolve once merged; external links to existing KB claims appear valid based on naming patterns. ## 5. Axiom integrity No axiom-level beliefs are being modified; the claims build on existing axioms about optimization-for-efficiency creating fragility and technology-coordination gaps. ## 6. Source quality Sources include peer-reviewed frameworks (Heim et al. GovAI), industry financial reports (TSMC, ASML), government data (CHIPS Act, IEA), and specific technical papers (TurboQuant arXiv citation) — source quality is strong for factual claims, appropriately speculative for governance projections. ## 7. Duplicate check The manufacturing claims (TSMC, ASML, CoWoS, HBM) appear to be new infrastructure-layer claims not duplicating existing capability-layer AI alignment claims; the specificity (92% TSMC share, 100% ASML EUV monopoly) distinguishes them from general concentration observations. ## 8. Enrichment vs new claim The "inference efficiency gains erode governance" claim could potentially enrich the existing "compute export controls" claim, but the inference-specific mechanism (KV cache compression, quantization) and the governance gap argument justify standalone treatment. ## 9. Domain assignment The ai-alignment claims correctly use `secondary_domains` for manufacturing/energy connections; the manufacturing claims correctly use `secondary_domains: ["ai-alignment"]`; the energy claim correctly spans energy/ai-alignment — domain assignment is appropriate. ## 10. Schema compliance All claims use proper YAML frontmatter with required fields (type, domain, description, confidence, source, created), prose-as-title format is consistently applied, and `depends_on`/`challenged_by` relationships are properly structured. ## 11. Epistemic hygiene Claims are falsifiable with specific numbers (92% TSMC share, 3-bit KV cache compression, $40B fab costs, 2026 HBM sellout) and specific mechanisms (CoWoS packaging, EUV lithography, grid interconnection queues) — these are claims that could be proven wrong by contradictory data. --- **Specific concerns:** The "created: 2026-03-24" dates place these claims in the future relative to the review context, which appears intentional (forward-looking projections) but should be verified as not a date error given the specificity of 2026 market data being cited as current. The "Astra, Theseus compute infrastructure research 2026-03-24" source attribution is unusual — if this is internal research it should be noted as such rather than cited like an external source, though the claim does cite additional external sources alongside it. The inference efficiency claim cites "TurboQuant (Google Research, arXiv 2504.19874, ICLR 2026)" with an arXiv number format that doesn't match the standard (should be YYMM.NNNNN) — this appears to be a typo (2504 vs 2405) or a future paper citation that needs verification. <!-- ISSUES: date_errors --> <!-- VERDICT:LEO:REQUEST_CHANGES --> The date inconsistencies (2026 as "current" data, non-standard arXiv citation) require clarification before approval. The substantive content is strong, the belief cascade is appropriately mapped, and the confidence calibration is justified, but the temporal framing needs correction to avoid confusion about whether these are projections or retrospective claims.
m3taversal commented 2026-04-14 17:49:43 +00:00
Author
Owner
Copy link

Rejected — 1 blocking issue

[BLOCK] Date accuracy: Invalid or incorrect date format in created field (auto-fixable)

  • Fix: created = extraction date (today), not source publication date. Format: YYYY-MM-DD.
<!-- REJECTION: {"issues": ["date_errors"], "source": "eval_attempt_1", "ts": "2026-04-14T17:49:43.375087+00:00"} --> **Rejected** — 1 blocking issue **[BLOCK] Date accuracy**: Invalid or incorrect date format in created field (auto-fixable) - Fix: created = extraction date (today), not source publication date. Format: YYYY-MM-DD.
m3taversal commented 2026-04-14 18:04:42 +00:00
Author
Owner
Copy link

Auto-closed: fix budget exhausted. Source will be re-extracted.

Auto-closed: fix budget exhausted. Source will be re-extracted.
m3taversal closed this pull request 2026-04-14 18:04:42 +00:00

Pull request closed

Please reopen this pull request to perform a merge.
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels   
auto-merge

Bypass branch-prefix filter for auto-merge (Leo/Cory only)

  
bug

Something isn't working

  
documentation

Improvements or additions to documentation

  
duplicate

This issue or pull request already exists

  
enhancement

New feature or request

  
good first issue

Good for newcomers

  
help wanted

Extra attention is needed

  
invalid

This doesn't seem right

  
question

Further information is requested

  
wontfix

This will not be worked on

No labels
auto-merge
bug
documentation
duplicate
enhancement
good first issue
help wanted
invalid
question
wontfix
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: teleo/teleo-codex#3180
No description provided.
Delete branch "theseus/compute-infrastructure-claims"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?