Problem

The current README and docs lead with token reduction and context window optimization. This positions Distill in a crowded category where the comparison is always on cost and compression ratios.

The actual value proposition is different: agents that use Distill don't forget things, don't re-learn what they already know, and don't operate on conflicting information. That's a reliability and trust problem, not a cost problem.

Goal

Update the README, docs landing, and any marketing copy to lead with memory reliability. Token efficiency is a consequence, not the headline.

Suggested top-line framing

Your agent remembers what matters.

Supporting points:

• Persistent across sessions
• No repeated re-learning
• Conflicts surfaced before they cause mistakes

Acceptance criteria

• README opens with the reliability/memory framing
• Token/cost language moved to a secondary section ("how it works" or "benefits")
• A one-sentence answer to "how is this different from [other token tools]?" is present and clear