Developer overhauls publishing pipeline and adds silent-failure detection across systems

A developer spent a weekend tracing why their automated dev-log pipeline had gone silent for over ten days, finding that the final save step had been quietly stalling on a pending permissions approval. The fix prompted a broader restructuring: publishing was migrated from a defunct Tistory API to GitHub Pages and Dev.to, with Tistory retained only as an archive. On Sunday, an A/B test on financial data processing was evaluated; despite a tempting result in a high-error subset, an AI consultant confirmed it was the same intervention measured under different conditions, so only the full-sample result was adopted as the official metric. Separately, a safety filter exposed an existing bug where internal prompt instructions leaked into final reports, and two timeout failures in an automated trading verification unit prompted new per-round timing logs with early-warning alerts. Across all fixes, the week's unifying theme was converting silent failures — processes that stopped or misfired without raising any signal — into observable, audible events.