Skip to content
AletheIA Docs
Esc
↑↓navigate↵open⌘Jpreview
On this page

External references — execution patterns

External sources evaluated while designing the Execution Pattern Governance layer. This is the only place where vendor-specific workflow terminology appears: these terms are external references, never canonical AletheIA language (the canonical terms are execution pattern, execution vehicle, looping model, orchestration contract, pattern selection, loop state, objective gate, human-led workflow).

Primary source

Anthropic — “A harness for every task: dynamic workflows in Claude Code”

URL: https://claude.com/blog/a-harness-for-every-task-dynamic-workflows-in-claude-code

How it informed the layer:

  • the idea that orchestration can move out of the model and into a harness;
  • named execution shapes — classify-and-act, fan-out-and-synthesize, adversarial verification, generate-and-filter, tournament, loop-until-done — that the pattern library abstracts;
  • failure modes worth designing against: agentic laziness, self-preferential bias, goal drift;
  • the cost observation that complex workflows consume more tokens and are not needed for every task — the root of the proportionality principle;
  • the observation that workflows are useful beyond engineering.

Decision: used as an external reference only. “Dynamic workflow” is not internal canonical language; the abstraction is Execution Pattern Governance.

Secondary source

Paweł Huryn — “Claude Dynamic Workflows (not only) for PMs: The Ultimate Guide”

URL: https://x.com/PawelHuryn/status/2064079508689358857

How it informed the layer:

  • translating execution shapes to product, research, and PM work;
  • the interviews → opportunities → score → prototypes example behind the interview synthesis example;
  • the principle that code can coordinate while the model judges.

Decision: used as an applied reference. The example’s numbers are not treated as general benchmarks; vendor-specific naming is not adopted.

Tertiary source

awesome-agent-loops

URL: https://github.com/serenakeyitan/awesome-agent-loops

How it informed the layer:

  • the distinction between interval loops, goal-until-condition loops, and scheduled loops;
  • the layering heuristic “timer outside, condition inside, skill innermost”.

Decision: used as a practical inspiration library; its specific commands are not canonical contract language.

Evaluated, not incorporated

OpenVera

URL: https://github.com/Reef123/OpenVera

Not incorporated as a structuring source in this round — insufficient consistent verification.

Fazm

URL: https://fazm.ai/cc

May be evaluated in a separate round for session persistence, chat forking, handoff, and agent UI topics.

Was this page helpful?