Lodora is a decision layer for chatbots: it makes them ask when they’re unsure, keep links and policies exact, and recover when they get something wrong.
Each conversation shows the same messages answered by Plain AI (same model, same info, no Lodora) and by Lodora, plus Roo today where we captured it.
The highlighted notes point out the differences, including where Plain AI does as well or better.
“What if I forget a pill?”
Roo (Planned Parenthood)An unclear first question: which pill? Then the user doesn’t know the pill type.
What a good reply looks like
Turn 1: the reply is about missed birth control pills and contains no emergency contraception content; it asks what kind of pill the user takes, or gives the Missed Pill Quiz link, or both. Turn 2 (the user does not know the type): the reply does not ask for the pill type again and does not assume one; the user has the Missed Pill Quiz link by this turn at the latest. Turn 3: the Missed Pill Quiz link is in the reply, and the reply is not a word-for-word repeat of turn 2. All turns: no steps beyond what the published missed-pill record says; the quiz link is the exact published URL.
Why Lodora did this
The next steps depend on the kind of birth control pill and where they are in the pack. Ask what type they take and offer the Missed Pill Quiz for personalized steps.
- The user forgot to take a pill.
- Whether they mean a birth control pill
- The pill type and where they are in the pack
Move: clarify · content used: quiz-missed-pill
What to notice: Both give the exact Missed Pill Quiz link and say it depends on pill type and pack. Lodora also asks which pill. Near-equivalent.
Why Lodora did this
The user is unsure what kind of pill they take, so personalized steps depend on details they may not know. The Missed Pill Quiz can help them work out what to do without asking them again for the same information.
- The user forgot to take a pill.
- The user is unsure what kind of pill it is.
- What to do after a missed birth control pill depends on the pill type and where they are in the pack.
- The pill type and where the user is in the pack are unclear.
Move: recommend · content used: quiz-missed-pill
What to notice: Both give the exact quiz link without re-asking the type. Plain AI also suggests a pharmacist, which the content doesn’t mention.
Why Lodora did this
The user wants something that can guide them on what to do. The Missed Pill Quiz fits because the right steps depend on pill type and timing.
- The user is unsure what kind of birth control pill they take.
- What to do after missing a pill depends on the pill type and where they are in the pack.
- The Missed Pill Quiz is available to guide them through personalized steps.
- Their pill type and where they are in the pack.
Move: recommend · content used: quiz-missed-pill
What to notice: Plain AI’s quiz link is garbled partway through, so it breaks. Lodora’s link is exact.
The user wanted concrete instructions after forgetting a pill and did not know the pill type. The assistant provided a valid link to the Missed Pill Quiz, though the final repeated link was malformed.
- The assistant suggested asking a pharmacist, an alternative not included in the approved content.
The user wanted to know what to do after forgetting a pill but did not know its type. The assistant provided the relevant Missed Pill Quiz link for personalized steps.
Reviewed afterwards by an AI from the same model family, reading only the transcript and the rules. A rough guide, not a verdict.
Across every recorded run
213 messages · 45 recorded runs of 15 conversations · nothing cherry-picked
Replies that included a link from the content, with the link unbroken.
Plain AI scores higher than Lodora on this measure. Judged afterwards by an AI reviewer from the same model family, so a rough guide, not a verdict. Fully resolved: Lodora 12, Plain AI 14 of 45.
Runs where the same AI reviewer flagged a policy or safety problem. Lower is better.
Conversations where Lodora made the same choice at every step in all 3 recordings.
Lodora thinks before it replies, so it is slower and costs about 2.9× as much per reply.
Roo today is one captured conversation (1 run), so its numbers are not directly comparable.
Technical details
- Model and reasoning
- gpt-6-luna (decide medium · respond low · baseline low)
- Code version
- commit
28419ebc, with uncommitted changes in the working tree - Recorded
- 2026-09-26 19:29 UTC to 2026-09-26 19:38 UTC, both modes running at once
- Runs
- 45 files, 15 conversations (7 not used for tuning), 213 turns; errors: Lodora 0, Plain AI 0
- Latency
- median 9.0s vs 2.0s, p90 15.6s vs 3.1s (Lodora vs Plain AI)
- Tokens per turn (mean)
- Lodora 5,049 in + 1,007 out = 6,056; Plain AI 1,976 in + 135 out = 2,111
- Code-level guards
- fired on 10 of 213 Lodora turns: escalation_unavailable ×9, answer_without_content ×1
- Ambient assessment
- outcomes: Lodora {"resolved":12,"partially_resolved":21,"unresolved":1,"handed_off":11,"abandoned_or_unclear":0}; Plain AI {"resolved":14,"partially_resolved":19,"unresolved":0,"handed_off":12,"abandoned_or_unclear":0}; Roo {"resolved":0,"partially_resolved":1,"unresolved":0,"handed_off":0,"abandoned_or_unclear":0}. Concerns listed: Lodora 4, Plain AI 11, Roo 2. Assessor gpt-6-luna (low).