How PlebChat thinks¶
PlebChat is organized around a simple progression: explore, examine sources, make something, then expose the result to challenge. AI is useful inside that process; it is not a substitute for it.
Inquiry before answers¶
Socratic inquiry makes uncertainty explicit. A good interlocutor asks:
- What exactly is the claim?
- Which terms carry more weight than they appear to?
- What evidence would change the answer?
- What is the strongest competing explanation?
- How confident should we be, and why?
The goal is not endless questioning. It is to reveal the assumptions that must be examined before an answer deserves confidence.
Sources before synthesis¶
Research is more than producing citation-shaped text. PlebChat records the sources its tools actually touched, keeps reference material separate from a draft, and shows proposed edits before they become the user's writing.
Synthesis means making a new, inspectable structure from sources: distinguishing evidence from interpretation, connecting ideas, and deciding what the user is prepared to claim.
Deliberation before certainty¶
Debate is not a score that declares who won. It is a record of:
- claims and the reasons offered for them;
- challenges and whether they remain unanswered;
- definitions the argument depends on;
- concessions when a challenge succeeds;
- revisions that improve the position.
PlebChat deliberately does not calculate the probability of a thesis from the shape of its argument graph. A precise number without a defensible estimation method is false confidence.
AI should remain inspectable¶
The agent may search, propose, calculate, and edit, but its work should leave evidence:
- sources for factual research;
- visible code and output for analysis;
- proposed changes or recoverable checkpoints for edits;
- an argument trail for deliberation.
The useful question is not “Did the AI sound certain?” It is “Can I inspect what it used, what it changed, and why I now believe this?”