research

field notes on research.

• 2026-07 sleepcast: bedtime audio with an inverted rubric

  a bedtime audio product whose generation rubric inverts every content incentive: repetition is good, drift is good, resolution is bad, and a completed play is a failure signal.

• 2026-05 techno-alexithymia: de-skilling introspection

  delegating emotional self-naming to a language model is the same de-skilling mechanism as offloading judgment to a coding agent, applied to the most interior faculty there is.

• 2026-04 agents.md effectiveness

  evaluating repository-level context files for coding agents.

• 2026-04 enterprise agent design

  bcg framework for production-grade agents.

• 2026-04 formal verification with agents

  property-based testing and specification generation.

• 2026-04 persona anchors

  character-based instruction patterns for consistent agent behavior.

• 2026-04 reward engineering for coding agents

  why coding agents optimize the rubric more than the prompt.

• 2026-04 reward hacking in coding agents

  how poorly designed metrics produce plausible but unstable code.

• 2026-04 context is a budget

  more tokens do not create more truth; useful context is what changes the next decision.

• 2026-04 protocol before personality

  shared artifacts and exit conditions beat charisma, cosplay, and fake seniority.

• 2026-04 specs as shared reality

  agents do not share understanding. they share binding artifacts, versions, and tests.

• 2026-04 what is prompting: operational constraints

  prompting is a constrained pipeline where tokenization, max tokens, and control tokens form the operational boundaries that determine agent behavior.

workflows

field notes on workflows.

• 2026-04 agent psychology

  understanding how agents reason and respond to instructions.

• 2026-04 ci automation with ai agents

  integrating ai agents with continuous integration workflows.

• 2026-04 instant project sync

  synchronize ticket queues against contract dates and repository evidence, with requirement reconstruction when docs are missing.

• 2026-04 preference toml

  use rlhf-shaped semantics in a simple config dsl for agent evaluation loops.

• 2026-04 project ai philosophy

  a written position for bounded, evidence-backed ai use in delivery.

• 2026-04 prompt patterns

  common prompt structures for reliable agent behavior.

• 2026-04 psay agent notifications

  local voice and notification center loops for agent completion, handoff, and operator recall.

• 2026-04 reward rubric dsl

  a small machine-readable format for scoring coding agent output.

• 2026-04 claude code skills stack

  a practical three-layer stack for claude code: decision, context, and execution.

• 2026-04 modular workflow stack

  right tool, right job, right time: layered orchestration, typed agent delegation, human-in-the-loop gates, parallelism, and loops with real halting conditions.

• 2026-03 anubis in front of github pages

  how to put anubis in front of a github pages site with a real reverse proxy and no proxy-loop nonsense.