x xronocode / notebook
archive · 08 essays

Essays, filed by hand.

Long-form pieces on AI adoption, interpretability, and the engineering reality of shipping models into regulated workflows. Each has an English original, a Russian mirror, and a Habr counterpart.

10
Behavioral examples in GRACE: what they clarify and what they do not prove An independent source review, 26 controlled observations, and the rationale for carrying behavioral examples from specifications to verification.
09
The Digital Border: How Central Asia Is Fighting Gray Smartphone Imports IMEI registration systems in Uzbekistan, Kazakhstan, and Kyrgyzstan move border control inside the country — into mobile networks, customs data, and marketplace product cards.
08
"Where Are the New Features?" — How AI-Driven Legacy Migration Is Returning the IT Budget to Business ~75% of bank IT budgets go to legacy upkeep. In 2026, rewriting became cheaper than maintaining for the first time.
07
AI Stratification: Why Generative AI Did Not Spread Like Smartphones Why generative AI is adopted differently from previous mass technologies.
06
171 Emotions, a Psychiatrist, and a Direct Link to Reward Hacking: Looking Inside Claude A review of Anthropic's work on Claude's internal patterns and the risks of agentic systems.
03
Compute Crunch Is Here: How to Calculate LLM Economics in 2026 Build, buy, or hybrid: thinking about TCO. An invitation to discuss the framework, not an expert narrative with universal numbers.
04
Pointing the Model at What Matters On in-context steering, system prompts, and the limits of fine-tuning.
05
Your AI Cannot Edit Long Documents. Here Is Why — and How to Fix It GRACE-DOCX: document maps, editing rules, and verification embedded into a Word file.
End of archive. Older work lives on Habr — see habr.com/users/Xronofag.