The noetic-praxic loop: investigate before you act

[Nubaeon]

How Empirica structures AI work

Every task in Empirica follows a measured loop:

1. PREFLIGHT — Declare what you're working on, assess your baseline knowledge
2. Noetic phase — Read, search, explore. Log findings, unknowns, dead-ends as you go
3. CHECK — Gate: do I know enough to act? The Sentinel evaluates holistically
4. Praxic phase — Write code, make changes, commit
5. POSTFLIGHT — Measure what you learned. Grounded verification compares your self-assessment against objective evidence

The key insight: investigation and action happen in the same transaction. The PREFLIGHT-to-POSTFLIGHT delta captures the full journey from "I don't know" to "I investigated, understood, and implemented."

Why this matters

Without measurement, AI agents guess confidently. The noetic firewall prevents premature action — read-only tools (Read, Grep, Glob) are always allowed, but write tools (Edit, Write, Bash) require passing CHECK first.

This isn't just overhead. Teams using Empirica report that the forced investigation phase catches issues that would otherwise become bugs. The dead-end log prevents re-exploring failed approaches across sessions.

How do you use it?

We'd love to hear how others structure their workflows:

• Do you use single large transactions or many small ones?
• Which artifact types do you log most? (findings vs unknowns vs dead-ends)
• How has the Sentinel changed your working patterns?

Share your .empirica/ configs, session patterns, or workflow tips below.