Overview
Praxist is an evidence-governed control plane for persistent, computer-executable research. The series follows a runnable task from its scientific contract and one-unit canary through peer generations, evaluation, evidence classification, boundary commits, and inherited agendas. It then examines diversity mechanisms, retention, memory, scheduling, budgets, runtime normalization, prompt cost, operation, replay, and independent audit. The central boundary remains stable: the task owns the objective, evaluator, constraints, and scientific meaning; Praxist coordinates execution and preserves provenance. These records can make research more durable and inspectable, but they cannot validate a metric, guarantee improvement, prove generalization, or replace independent scientific judgment.
Learning outcomes
- Praxist coordinates the research process, while the task project owns the objective, evaluator, constraints, and scientific meaning.
- Evidence guides inheritance only when task-defined maturity, retention, and generation-boundary rules support that use.
- DIG, QD, retention lanes, memory, and graph guidance can broaden search and continuity without proving a candidate is better.
- Scheduling, budgets, runtime normalization, replay, and tests strengthen process integrity but cannot validate a metric or guarantee generalization.
Keywords
Preview every chapter · Unlock full text, audio, and video from chapter 4
PlansFAQ
What does Praxist control in a research run?
Praxist controls the reusable process around a task: peer sessions, generations, scheduling, runtimes, budgets, evidence transport, durable state, replay, and synthesis. The task project still owns the objective, evaluator, constraints, and scientific meaning.
What does a Praxist generation commit?
A generation commits reconciled evidence and the next agenda at an explicit boundary, not merely when files or processes appear finished. Results after the cutoff can remain visible as validation signals but do not rewrite the closed generation.
Can Praxist prove that a research result is scientifically valid?
No. Praxist can make the task, measurements, lifecycle, provenance, and replay checks inspectable, but it cannot show that the metric is a good proxy, the evaluator is sound, or the result generalizes; independent validation remains necessary.
When is Praxist a good fit for research?
Praxist is a stronger fit when a project has a runnable baseline, a public evaluator, a measurable objective with known direction, explicit constraints, and an uncertain path worth exploring in parallel. It is weaker for one-off answers or work without a credible evaluator.