Report · leodemoura.github.io
When AI writes the software, who verifies it?

leodemoura.github.io
Source published March 3, 2026
Summary: 1 min read
62/100
Summary
Code Metal has raised $125 million to rewrite defense industry code using AI. Google and Microsoft report that 25–30% of their new code is AI-generated, while AWS has modernized 40 million lines of COBOL for Toyota using AI.
Key Takeaways
- Code Metal raised $125 million to rewrite defense industry code using AI, highlighting the increasing reliance on AI in software development.
- Google and Microsoft report that 25–30% of their new code is AI-generated, with predictions that 95% of all code will be AI-generated by 2030.
- Nearly half of AI-generated code fails basic security tests, and there is currently no formal verification process for the accuracy of AI-generated software.
- Poor software quality costs the U.S. economy $2.41 trillion per year, a figure that is expected to rise as AI generates a significant portion of new code.
Coverage & discussions
Hacker News · leodemoura.github.io
Community discussions
Community Sentiment
Positives
- The evolution of software development towards AI-assisted coding signifies a shift towards a more engineering-focused discipline, emphasizing oversight and structural integrity over manual coding.
- The decreasing marginal cost of code production suggests that AI can democratize software development, allowing more individuals and companies to innovate without the traditional barriers.
Concerns
- Despite advancements, AI-generated code often contains errors, highlighting the critical need for thorough code reviews and verification processes to ensure reliability.
- As AI becomes more autonomous in coding, the risk of producing untested and potentially flawed software increases, posing significant challenges for quality assurance.
Related Articles

LLMs are making formal software verification feasible at a much larger scale
Oct 22, 2026

Go is an ideal language for AI-assisted software engineering
Aug 11, 2026

What can you confidently guarantee about your software?
Jun 29, 2026

Leanstral 1.5: Proof abundance for all
Jul 3, 2026

Code Is Cheap Now, and That Changes Everything
Apr 9, 2026