What I am thinking deeply about.

Key question

**When a probabilistic component is placed inside a request path that users depend on, what changes structurally in the architecture, and which of the resulting failure modes are already documented in the software engineering literature rather than novel?** The second half of the question matters more than the first. There is a widespread assumption in current practice that AI-native architecture is unmapped territory. The premise examined here is that a substantial portion of it was mapped between 2015 and 2020, in papers about machine learning systems, and that the practical failure of current products is a failure to read rather than a failure of available knowledge.

Key question

**If a multi-step agent workflow is modelled as a long-running distributed transaction rather than as a prompting problem, which established patterns apply, and what does that model imply about where human checkpoints belong?** A subsidiary question follows from it: does the transaction model give a design-time test for checkpoint placement, replacing the current practice of placing checkpoints by intuition about model trustworthiness?

Key question

**Which interface mechanisms are empirically supported as improving the calibration of a user's reliance on an AI system, and which widely deployed mechanisms have been tested and found not to?** The framing of the question is deliberate. It does not ask how to increase trust, because the literature's consistent position is that increasing trust is not the objective and that both excessive and insufficient trust are failures with distinct costs.

Key question

**Across the available empirical studies of AI-assisted software development, what outcome variable was measured, on what population and task type, and what do the results support once those differences are made explicit?** A second question follows: which decision-relevant quantity is currently unmeasured?

Key question

**Which assumptions encoded in a conventional design system does generated content violate, and what primitives would allow a system to absorb those violations once rather than once per feature?**