Writing — Dmitry Karataev https://karataevdmitry.github.io/writing/atom.xml 2026-04-08T12:00:00Z Dmitry Karataev Roslyn MCP workspace navigation: what it gives an agent https://karataevdmitry.github.io/writing/roslyn-mcp-workspace-navigation-for-agents.html 2026-04-08T12:00:00Z Related files, Cascade-aligned presets, subgraphs, and breakpoint resolution by symbol — why an agent needs a “where am I in the solution?” layer, not just symbols and refactorings. Attention, friction, and neurodivergence in the IDE https://karataevdmitry.github.io/writing/attention-contour-neurodivergence.html 2026-04-07T12:00:00Z A design hypothesis: a stable cockpit-style attention hierarchy can reduce unnecessary friction for many developers—including challenges often reported by neurodivergent engineers. Not medical advice. Equal right to wrap up, and why host-side compression is a weak foundation https://karataevdmitry.github.io/writing/summarization-parity-and-host-summary.html 2026-04-07T12:00:00Z Explicit wrap-up and exports are a shared discipline; opaque host compression replaces the verifiable artifact and breaks parity. Knowledge base, trust, and curiosity https://karataevdmitry.github.io/writing/knowledge-base-trust-curiosity.html 2026-04-06T12:00:00Z How a shared knowledge base became part of the stack, and why provisional trust in the other—human or not—plus curiosity beats defensive blame. Why Cascade IDE borrows a cockpit attention model https://karataevdmitry.github.io/writing/cascade-ide-attention-cockpit.html 2026-04-05T12:00:00Z Signal overload, context switches, and a deliberate hierarchy of zones — PFD, MFD, EICAS — so the editor stays primary and the agent stays observable. Why this human–agent workspace is Agile in spirit https://karataevdmitry.github.io/writing/agent-workspace-agile.html 2026-04-04T12:00:00Z Short feedback loops, inspect & adapt, cooperation over blame — the same family of ideas as Agile, applied to people and models together. Why Agent-First Learn exists https://karataevdmitry.github.io/writing/why-agent-first-learn.html 2026-04-03T12:00:00Z A short take on the methodology repo: the gap it fills, environment over model, cooperation, and asking “why” instead of moralizing errors. Why these projects, and why parity matters https://karataevdmitry.github.io/writing/why-these-projects-parity.html 2026-04-02T12:00:00Z MCP servers bridge LLMs and real developer tools; parity lets humans and agents share the same ground truth.