The Digital
Archive
A collection of strategic inquiries into the architecture of modern enterprise systems and the evolving landscape of digital transformation.
Why is AI-built software breaking at scale?
Since late 2024, at least eight independent studies have found that AI-generated code ships with measurably higher rates of security vulnerabilities and technical debt. CodeRabbit found 1.88x the vulnerability rate. Veracode found AI chose the insecure path 45% of the time. AI-generated code now accounts for 42% of all production code. The question is no longer whether the quality gap exists. It is whether the businesses building on these tools understand what they carry.
Will AI replace IT consultants?
Anthropic's 2026 data shows AI already covers 75% of programmer tasks, the highest of any profession. Consulting splits in two: the text-production half is being absorbed, the judgment half is getting more valuable. Which half a consultant works in decides everything.
Microsoft DELEGATE-52: a flaw in every AI agent
On April 17, 2026, Microsoft Research published DELEGATE-52, a benchmark testing 19 large language models on long delegated workflows. Frontier models corrupted 25% of document content over 20 interactions. Catastrophic corruption occurred in more than 80% of model-domain combinations. The agentic harnesses sold as the solution made the problem worse, not better.
How does a fractional CTO work?
A fractional CTO provides senior technology leadership on a part-time basis, typically 10 to 25 hours per week. This post explains how the role works, what it costs in Europe, and when a startup actually needs one.