Singular State https://singular-state.com Advanced technical expertise for feasibility, prototyping and modernization. We build the working pilot and ramp up your team to take it forward. en-US Where AI-written code actually breaks: rebuilding a CD-ROM encyclopedia https://singular-state.com/experiments/encarta-revival-where-ai-written-code-breaks/ https://singular-state.com/experiments/encarta-revival-where-ai-written-code-breaks/ Wed, 09 Sep 2026 00:00:00 +0000 An agent built a Tauri desktop application end to end. The pure logic was correct on the first or second attempt. Every defect that mattered sat at a boundary no headless test can reach. Passing the tests is not understanding the system: a sealed-key measurement https://singular-state.com/experiments/grading-a-reconstruction-against-a-sealed-answer-key/ https://singular-state.com/experiments/grading-a-reconstruction-against-a-sealed-answer-key/ Wed, 09 Sep 2026 00:00:00 +0000 A compressor reconstructed from behaviour alone, graded against an answer key sealed until the work was frozen. The decoder scored 100 percent and the encoder 8.3 percent on the same format. Turning a novel into a specification: the Nautilus under real physics https://singular-state.com/experiments/verne-nautilus-simulator-narrative-as-specification/ https://singular-state.com/experiments/verne-nautilus-simulator-narrative-as-specification/ Wed, 09 Sep 2026 00:00:00 +0000 Verne's submarine crushes at 340 metres against a narrated 16,000. Extracting a falsifiable spec from fiction, and what it takes to trust numbers an AI produced when no external oracle exists. Characterization tests: recording the behavior that exists https://singular-state.com/articles/characterization-tests-recording-the-behavior-that-exists/ https://singular-state.com/articles/characterization-tests-recording-the-behavior-that-exists/ Mon, 07 Sep 2026 00:00:00 +0000 Characterization tests preserve selected observed behavior. Their baseline must distinguish required behavior from existing defects and comparison artifacts. Parallel replacement and comparable results https://singular-state.com/articles/parallel-replacement-and-comparable-results/ https://singular-state.com/articles/parallel-replacement-and-comparable-results/ Mon, 07 Sep 2026 00:00:00 +0000 Parallel replacement compares implementations over matched inputs and state before changing operational authority. Shadow execution must control external effects. Who checks the expected result in an AI-generated test? https://singular-state.com/articles/who-checks-the-expected-result-in-an-ai-generated-test/ https://singular-state.com/articles/who-checks-the-expected-result-in-an-ai-generated-test/ Mon, 07 Sep 2026 00:00:00 +0000 Evaluating generated assertions through independent examples, approved baselines, and meaningful fault detection.