Writing
Short pieces on why the open-source MCP stack exists and what “parity” between people and agents is meant to achieve.
- agents
- cascade
- ide
- knowledge-base
- mcp
- methodology
- parity
- tooling
- accessibility
- agile
- cursor
- design
- roslyn
-
Why these projects, and why parity matters
MCP servers bridge LLMs and real developer tools; parity lets humans and agents share the same ground truth.
-
Why Agent-First Learn exists
A short take on the methodology repo: the gap it fills, environment over model, cooperation, and asking “why” instead of moralizing errors.
-
Why this human–agent workspace is Agile in spirit
Short feedback loops, inspect & adapt, cooperation over blame — the same family of ideas as Agile, applied to people and models together.
-
Why Cascade IDE borrows a cockpit attention model
Signal overload, context switches, and a deliberate hierarchy of zones — PFD, MFD, EICAS — so the editor stays primary and the agent stays observable.
-
Knowledge base, trust, and curiosity
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.
-
Attention, friction, and neurodivergence in the IDE
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
Explicit wrap-up and exports are a shared discipline; opaque host compression replaces the verifiable artifact and breaks parity.
-
Roslyn MCP workspace navigation: what it gives an agent
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.