DOGFOOD-013Internal record
Multi-tool continuity
Same workspace across tools; proposals, not Decisions.
Limitation
Continuity lives in files. A tool may fail auth; that does not prove a product.
Memory and governance · Experimental
Systems evolve.
The why deserves to stay.
DARXI organizes decisions, evidence and constraints to preserve the context of AI-assisted work. Models propose. People decide.Decisions, evidence and context for AI-assisted work.
Experimental development. Usefulness beyond the lab remains unproven.
Context connects. Authority remains human.Continue the investigation
01 / The problem
The result remains. What about the reasoning?
A decision loses value when we can no longer explain where it came from, which alternatives we rejected or where it stopped being valid.
- 01
The tool changes.
The conversation is left behind.
- 02
The person changes.
The reasoning is no longer at hand.
- 03
The evidence changes.
We need to know what to revisit.
02 / Inside a decision
Not just what was decided. The reason, too.
Explore how a rejected proposal is preserved, with its reasons and open questions.
Editorial example · based on the laboratory
A proposal is rejected. What remains?
This example illustrates the shape of an internal case of human rejection. It simplifies the content and does not reproduce an original record.
Proposal, no authority01 / 04
The system proposes a change.
A proposal opens a possibility and explains its reasons. Here we follow the shape of that journey without reproducing the internal details of the change.
What is preserved
The original proposal, even when it is later rejected.
Origin
System proposal
Evidence with limits02 / 04
Disagreement contains information, too.
An architecture question passed through several tools and produced independent proposals. The case preserved the positions with their provenance.
Evidence limit
One question, one session. The positions were closer than the headline suggests.
Scope
Internal experience
Human review03 / 04
A person can reject the proposal.
The public account records a human rejection. Review separates what the system proposes from what a person accepts as a recorded decision.
The distinction
Receiving a proposal does not mean approving it. Evidence informs judgment; it does not replace it.
Authority
Human
Rejection preserved04 / 04
Rejected for now. Preserved for later.
The original proposal, its reasons and open questions remain available alongside the human rejection. A decision does not erase the path that led to it.
What remains open
Whether the preserved disagreement is useful to an external team has not been measured.
Case outcome
Rejected for now
Evidence accompanies the proposal and review; it is not an automatic approval.
Read the case in EvolutionContext that travels together
- Original proposal
- Evidence and limits
- Human authority
- Rejected alternatives
- Open questions
- History of changes
03 / The mechanism
Each step has a role. None decides for all.
A proposal opens a possibility. Evidence lets us examine it. A person reviews and records a decision that can approve, reject or narrow what was proposed.
Proposal
A possibility
Evidence
What supports it
Review
Human judgment
Decision
A recorded judgment
- Proposal ≠ Decision
- Evidence ≠ Truth
- AI ≠ Authority
04 / From the laboratory
An open question. Work you can examine.
The same records and limitations you will find in the lab, brought together here as a starting point.
Question under investigation
Can the reason behind a decision be preserved when tools and models change?
What remains to be established
Whether this holds outside this lab. External evidence: none.
DOGFOOD-017Internal record
Drift detection
Detector reports Decision↔code mismatch.
Limitation
Detection only — no automatic remediation.
External human evidence: N=0 in the published content. These records describe internal experiences.
05 / What we learned
We keep what we stopped believing.
Revising an idea is part of the work. The history preserves the earlier proposal, what contradicted it and the judgment that followed.
Explore the evolutionWe believed
The value was in producing a single answer.
We hold today
There is also value in preserving disagreement and its provenance.
Whether that disagreement helps another person later remains unmeasured.
06 / Your next question
Choose your depth.
Start with what you need to understand.