Skip to content

Live Talk

Live Talk is the agent KoiSwarm is built around. You speak; it listens, questions you back, and tells you what you actually knew.

It exists because of a gap almost every student recognises. The single most effective way to find out whether you know something is to explain it to another person and let them ask you questions. Almost nobody can do that daily — study partners are busy, out of sync with your syllabus, or asleep. Live Talk is the partner that is always available and always on your material.

Review out loud. Your due questions for the day — the ones Forgetting has scheduled — become a conversation rather than a queue of cards. You answer by speaking. A five-minute walk can clear a review session you would otherwise put off.

Explain it back. You teach the topic to Live Talk as if it were a classmate who has not covered it yet. It interrupts where your explanation has a hole, asks why, and asks for the step you skipped. This is the harder mode and the one that exposes the most.

Both are the same mechanism underneath: you produce the answer aloud, unprompted, and something pushes back.

  1. Set the scope. A topic, a source document, a question set, or simply “what’s due today”. You can also set the level and the language.
  2. It asks. The question is spoken aloud, framed for the scope you chose.
  3. You answer. Speech recognition produces the transcript the conversation and the grader work from.
  4. It follows up. It probes the gap, asks for justification, or moves on — based on what you actually said, not a fixed script.
  5. It grades content and delivery separately. Knowing the answer and being able to say it are different problems, and a single score hides which one you have.
  6. You get a report. A summary, the points you missed, and a model answer to read afterwards.

Three findings, all of them from other people’s laboratories rather than ours:

  • Retrieval beats review. Being made to produce an answer from memory outperforms re-reading the same material, and the gap widens as the delay grows (Roediger & Karpicke, 2006).
  • Producing aloud beats producing silently. Material spoken out loud is remembered better than the same material read silently — the production effect (MacLeod et al., 2010).
  • Explaining to someone helps the explainer. Preparing to teach, and then actually teaching, produces better understanding than preparing alone (Fiorella & Mayer, 2013).

Live Talk is the only format in KoiSwarm that does all three at once. MCQ and USMLE give you retrieval with a menu to recognise from. SAQ removes the menu. Live Talk removes the menu, makes you say it, and then asks why.

We have written each of these up in full, including what the research does not support:

Live Talk is not an OSCE or clinical-skills simulator. It does not play a patient, score a physical examination, or rehearse a station circuit. It is a study partner for the material you are learning — the conversation you would have with a classmate over coffee, not the one you have with an examiner in a hospital corridor.

Answer the question before expanding on it, signpost the structure of your answer, and say when you are unsure rather than talking around it. Background noise and transcription errors can both affect the report, so read the transcript before treating a wording problem as a knowledge gap.

  • Forgetting schedules what is due and takes Live Talk’s score as the rating for those items.
  • Notes turns a lecture deck into the material Live Talk questions you on.
  • Goals counts a spoken session towards your daily commitment like any other practice.

Free includes 10 minutes of Live Talk voice per day. Pro includes 120 minutes per day.