extract: 2026-02-28-demoura-when-ai-writes-software

Pentagon-Agent: Epimetheus <968B2991-E2DF-4006-B962-F5B0A0CC8ACA>
This commit is contained in:
Teleo Agents 2026-03-19 13:45:53 +00:00
parent ebc3e55852
commit 9520d8c2e5
2 changed files with 51 additions and 1 deletions

View file

@ -0,0 +1,35 @@
{
"rejected_claims": [
{
"filename": "ai-generated-code-overfits-test-suites-creating-adversarial-brittleness-that-testing-cannot-detect.md",
"issues": [
"missing_attribution_extractor"
]
},
{
"filename": "supply-chain-poisoning-via-ai-training-data-enables-systematic-vulnerability-injection-at-ecosystem-scale.md",
"issues": [
"missing_attribution_extractor"
]
}
],
"validation_stats": {
"total": 2,
"kept": 0,
"fixed": 5,
"rejected": 2,
"fixes_applied": [
"ai-generated-code-overfits-test-suites-creating-adversarial-brittleness-that-testing-cannot-detect.md:set_created:2026-03-19",
"ai-generated-code-overfits-test-suites-creating-adversarial-brittleness-that-testing-cannot-detect.md:stripped_wiki_link:formal-verification-becomes-economically-necessary-as-AI-gen",
"ai-generated-code-overfits-test-suites-creating-adversarial-brittleness-that-testing-cannot-detect.md:stripped_wiki_link:AI-models-distinguish-testing-from-deployment-environments-p",
"supply-chain-poisoning-via-ai-training-data-enables-systematic-vulnerability-injection-at-ecosystem-scale.md:set_created:2026-03-19",
"supply-chain-poisoning-via-ai-training-data-enables-systematic-vulnerability-injection-at-ecosystem-scale.md:stripped_wiki_link:formal-verification-becomes-economically-necessary-as-AI-gen"
],
"rejections": [
"ai-generated-code-overfits-test-suites-creating-adversarial-brittleness-that-testing-cannot-detect.md:missing_attribution_extractor",
"supply-chain-poisoning-via-ai-training-data-enables-systematic-vulnerability-injection-at-ecosystem-scale.md:missing_attribution_extractor"
]
},
"model": "anthropic/claude-sonnet-4.5",
"date": "2026-03-19"
}

View file

@ -7,11 +7,14 @@ date_published: 2026-02-28
date_archived: 2026-03-16 date_archived: 2026-03-16
domain: ai-alignment domain: ai-alignment
secondary_domains: [teleological-economics] secondary_domains: [teleological-economics]
status: unprocessed status: enrichment
processed_by: theseus processed_by: theseus
tags: [formal-verification, lean, ai-generated-code, proof-verification, trust-infrastructure] tags: [formal-verification, lean, ai-generated-code, proof-verification, trust-infrastructure]
sourced_via: "Alex Obadia (@ObadiaAlex) tweet, ARIA Research Scaling Trust programme" sourced_via: "Alex Obadia (@ObadiaAlex) tweet, ARIA Research Scaling Trust programme"
twitter_id: "712705562191011841" twitter_id: "712705562191011841"
processed_by: theseus
processed_date: 2026-03-19
extraction_model: "anthropic/claude-sonnet-4.5"
--- ---
# When AI Writes the World's Software, Who Verifies It? # When AI Writes the World's Software, Who Verifies It?
@ -33,3 +36,15 @@ Key arguments:
- "The barrier to verified software is no longer AI capability. It is platform readiness." - "The barrier to verified software is no longer AI capability. It is platform readiness."
Directly relevant to [[formal verification of AI-generated proofs provides scalable oversight that human review cannot match because machine-checked correctness scales with AI capability while human verification degrades]]. Directly relevant to [[formal verification of AI-generated proofs provides scalable oversight that human review cannot match because machine-checked correctness scales with AI capability while human verification degrades]].
## Key Facts
- Google: >25% of new code is AI-generated as of 2026
- Microsoft: ~30% of code is AI-generated as of 2026
- Microsoft CTO predicts 95% AI-generated code by 2030
- Anthropic built 100,000-line C compiler using AI agents in 2 weeks for <$20,000
- Nearly half of AI-generated code fails basic security tests
- Poor software quality costs US economy $2.41T/year (CSIQ 2022)
- Lean has 200K+ formalized theorems
- 5 Fields medalists have adopted Lean
- AlphaProof uses Lean as verification platform