Turn the Question Learners Keep Asking Into a Mini-Lesson

The short answer: Every question the cohort keeps asking can become a permanent mini-lesson. Golpo turns the approved answer into a polished visual explanation, with the instructor's selfie, own audio, or authorized voice clone so the library still sounds like them.

Why this is practical now: The answer can live on as a lesson the next learner watches. One instructor records the judgment once. Golpo designs, draws, narrates, and syncs the explanation into a finished video.

A course can build that shelf question by question. Frame editing repairs a detail. The API can generate the next mini-lesson from the tools the team already uses: Codex, Claude Code, Cursor, Copilot, or the Golpo MCP server.

Use the full Golpo production system

Golpo is built to be the world's best AI whiteboard explainer video maker. Give it a topic, a prompt, a script, an audio recording, or a source document. It scripts, plans the scenes, designs the visuals, narrates, times and synchronizes every frame, and returns a polished video. One person can run that system across a whole course.

Recurring questions are ideal for creator-led scale: the instructor records once, uses an authorized voice clone for approved future lessons, or appears from a selfie while Golpo turns each approved answer into a visual mini-lesson.

• Start with an idea or source. Golpo can draft the script and plan the scenes from a topic, a document, or your outline. Script Mode can also speak the exact words you already trust.
• Keep the creator present. With AI Presenter, a course creator, founder, professor, or YouTuber can upload a selfie and appear beside the whiteboard. One person can lead a large lesson library without recording every visual sequence on camera.
• Use the right voice. Own Narration turns uploaded or live-recorded audio into the visual lesson. An authorized Voice Clone gives recurring lessons a recognizable voice at scale. Consent and pronunciation review remain required.
• Repair instead of restarting. The frame-by-frame editor can replace text, images, references, or an individual visual when the QA pass finds a local problem.
• Automate at scale. The Golpo API can be called from Codex, Claude Code, or through the Golpo MCP server in Cursor, VS Code with GitHub Copilot, Claude Code, Claude Desktop, and Visual Studio. The course platform can generate, poll, review, and publish without a person moving files between tools.

Support deflection is a different log

Help-center articles to support videos serve customers who are stuck in a product, and success there is often “they did not need a ticket.” This page serves learners inside a course. The mini-lesson points back at a canonical lesson when one exists. It does not pretend to be the whole unit. If the accounting video already answers the revenue question, the log links to that lesson instead of filming a rival explanation. If the rounding video already exists, float questions that are actually about round() go to the programming spoke. The 0.1 + 0.2 case stays here because that spoke deliberately left it out.

From inbox to item

1. Log the question in the learner’s words, plus the lesson they were on, the date, and a normalized topic tag a person has confirmed. Automatic tags can propose. They do not merge topics by themselves.
2. Set a threshold in advance. This guide’s working example is three independent asks in a term, or one ask from a teaching assistant who reports a roomful. The number is a course policy, not a research result. Publish the threshold so staff do not invent it after seeing a question they want to film.
3. Search the library. The QA agent, or a person, looks for a live lesson with the same objective. A hit means “link the existing file,” not “draft a synonym.” The log stores the lesson ID that answered it.
4. Draft only on a miss. The creator agent writes a one-minute script that answers the question and names the fuller lesson it depends on. It does not open a new theory.
5. Instructor approval. A named instructor accepts the answer. The QA agent recomputes numbers and runs the duplicate search. Approval is still the instructor’s. Uncertain matches, where a lesson almost answers the question, wait.
6. Render and place. Golpo renders the approved script. The mini-lesson appears under the unit it clarifies, and the log entry changes from open to answered, with the new lesson ID.

“We collected $1,200 in January. Why isn’t revenue $1,200?”

Under the accounting spoke’s constructed assumption — one obligation, satisfied evenly over 12 months — January revenue is 1,200 / 12 = 100. The remaining contract liability is 1,100. Cash can be 1,200 and revenue can be 100 at the same time because the service is still owed. The mini-lesson says that once, shows the two numbers, and points to the full entry lesson for the five-step context and the monthly journal line. It does not re-teach the entire revenue model. The QA agent divides 1,200 by 12 and checks that the script did not say January revenue is 1,200. A duplicate search should find the accounting lesson and link it. The mini-lesson is allowed because the question is the sticking point, and the full lesson is the citation. Two competing “what is revenue” videos would fail the duplicate check.

“The test failed on 0.1 + 0.2.”

On a typical CPython 3 build, the expression 0.1 + 0.2 evaluates to 0.30000000000000004, so an assertion of exact equality to 0.3 fails. The course re-runs that expression on the pinned interpreter and stores the repr() it prints. The mini-lesson shows that one result. It points the repair at math.isclose or at decimal.Decimal, and it names which repair the course’s tests actually use. It does not also teach round-half-to-even. That is the programming spoke’s lesson. The QA agent’s duplicate check fails the draft if the script spends its minute on round(0.5). The creator agent drafts only after that search is clean, and only after the printed repr is in the record.

Both questions are constructed illustrations of the log, not exports of a real inbox. Private student text needs the course’s own privacy rules before it is spoken aloud in a video other students will watch. The safe pattern is to generalize the question so the video does not quote a graded submission. The wrong-step spoke is where a specific submission is examined, inside that attempt’s access scope.

Own Narration: OFF · Script Mode: ON · Duration: 1 minute · Orientation: Horizontal (16:9) · Style: Professional Clean · Color: ON · Language: English · Voice: Female 2 · Pace: Normal · Music: None

Learners keep asking the same question. We collected twelve hundred dollars in January. Why isn't revenue twelve hundred dollars? The course already teaches the annual plan. One obligation is satisfied evenly over twelve months. January revenue is one hundred dollars. Twelve hundred minus one hundred leaves eleven hundred still in contract liability. The cash is in the account. The work is not finished. One hundred times twelve returns to twelve hundred by the end of the year. This mini lesson answers the question people actually ask, then points to the full accounting lesson for the five steps and the monthly entry. The question had already crossed the course's threshold. The new video opens that lesson. It does not replace it. One hundred is January. Twelve hundred is the year. Point to the full lesson after one hundred and eleven hundred are both clear. The question is the title of this minute.

The question on screen: twelve hundred was collected in January. January revenue is one hundred. Eleven hundred remains a contract liability. End by pointing to the full accounting lesson. Allowed labels: Collected $1,200 in January; January revenue $100; Contract liability $1,100; 100 x 12 = $1,200; Opens the full accounting lesson. Do not add a sample stamp, a demo label, or any claim that the lesson is fictional. Show only the labels listed. If a number appears, it must match the narration exactly. No extra dates, prices, percentages, logos, or quotations.

Answer the question, then point onward. Do not reteach every step of the full lesson.

Own Narration: OFF · Script Mode: ON · Duration: 1 minute · Orientation: Horizontal (16:9) · Style: Professional Clean · Color: ON · Language: English · Voice: Male 1 · Pace: Normal · Music: None

Learners keep asking why the test failed on zero point one plus zero point two. In Python, that sum evaluates to zero point three zero zero zero zero zero zero zero zero zero zero zero zero zero zero zero four. It does not evaluate to zero point three exactly. The extra digits are the binary fraction. Point the assertion at math dot isclose, or at Decimal, when the course needs a decimal result. Print the long value once so the learner can see it. This mini lesson answers that repeated question and points to the rounding lesson for ties. The rounding lesson is where zero point five belongs. The instructor approves this mini lesson before it is published to the course. Show the long value in full. Do not round it to zero point three. The long value stays exact on the screen. Zero point three would hide the failure.

Show zero point one plus zero point two, then the full long result. Do not replace it with zero point three. Point the assertion toward math.isclose or Decimal, and point the viewer to the rounding lesson for zero point five. Allowed labels: 0.1 + 0.2 = 0.30000000000000004; math.isclose; Decimal; Points to the rounding lesson. Do not add a sample stamp, a demo label, or any claim that the lesson is fictional. Show only the labels listed. If a number appears, it must match the narration exactly. No extra dates, prices, percentages, logos, or quotations.

Read the long decimal exactly. Do not round it while speaking.

What the log should refuse

• A question asked once, by one person, below the threshold, unless an instructor overrides and the override is stored.
• A new video when a live lesson already states the answer. Link it. Record the link.
• A script that changes the accounting assumption or the Python pin without saying so.
• A render that speaks a private submission verbatim.
• An answer the instructor has not accepted. The creator agent’s draft is not acceptance.

Risks

Logs reward volume. A threshold set after the fact will always be exactly the count of the question someone wants to produce. Set it first. A QA agent that searches titles and misses a lesson filed under “deferred revenue” will allow a duplicate. The search has to include the objective sentence and the topic tag, not only the title. Uncertain search results go to the instructor with both lesson IDs visible.

Mini-lessons also rot. When the full accounting assumption changes, the January mini-lesson changes with it or retires with it. The version habit on the update spoke applies. A mini-lesson that outlives its parent lesson becomes a secret curriculum.

A constructed log that crosses the threshold

Picture a staff view, not a real inbox. The topic tag “January cash versus revenue” collects three notes in a term. Note one, from a learner on the accounting unit: they collected the annual cash and expected the income statement to show the same number. Note two, a week later, asks why the liability still shows most of the cash. Note three, from a teaching assistant, says the same confusion showed up in workshop. The threshold this course published in advance is three. The duplicate search finds the full accounting lesson, where the $1,200 entry and the $100 January revenue are already taught. The log links that lesson on each note. The instructor still approves a mini-lesson because the notes show people finished the parent video and kept the question. The mini-lesson’s script is the short answer: January revenue is $100, the remaining obligation is $1,100, and the parent lesson has the entries. The QA agent checks the division and checks that the mini-lesson does not introduce a second obligation the parent lesson does not use.

The float question follows the same gate. Three notes say a test failed because 0.1 plus 0.2 was not 0.3. The search finds the rounding lesson and rejects it as the answer, because that lesson is about ties to even, not about binary fractions. The miss is real. The creator agent may draft only after someone re-runs the expression on the pinned interpreter and pastes the repr into the row. The instructor approves the repair the course’s other tests use. If the suite uses math.isclose, the mini-lesson says that. If the suite uses Decimal, the mini-lesson says that. Saying both “or” without a course choice leaves the learner with a new question. One repair, named, plus the printed bad equality, is the whole minute.

Below-threshold notes stay on the log and do not generate. Staff can still answer the person with a link. A single eloquent question is not a mandate to film. Overrides are allowed, and they store the name of the instructor who lowered the threshold for that topic. An override with no name is indistinguishable from skipping the rule. At the end of the term the log is reviewed. Answered items whose parent lesson was retired reopen. Unanswered items below the threshold expire or carry forward under the same tag, by a policy the course writes down once. The log is a production queue with brakes, not a scrapbook of every confusion the platform can detect.

Automate it: review gates and publishing with the Golpo API

For recurring-question mini-lessons, the whole path from source to published lesson can run as one pipeline on the course platform. An agent in Codex with the Golpo Codex Skill, in Claude Code with the Golpo Claude Skill, or in Cursor, VS Code with GitHub Copilot, Claude Code, Claude Desktop, or Visual Studio through the Golpo MCP server can call the Golpo API, wait for the render, and hand the file to the platform's review steps. Production stops being the limit on how many checked lessons a course can offer.

1. Attach the reference first. Store the instructor-approved answer and the source it cites with the lesson before any script is drafted.
2. Review the script against the reference before rendering. If a solution or source exists, the platform's QA agent checks the draft against it: it re-solves, re-adds, or re-reads, then compares. A mismatch or an uncertain result stops the job before Golpo renders anything.
3. Generate with Golpo. The agent sends the approved script in Script Mode, with the course's locked style, voice, Video Instructions, and Voice Instructions.
4. Transcribe what was actually said. Run speech-to-text on the finished video's audio. Learners hear the transcript, not the draft.
5. Run two independent transcript reviews. Make two separate LLM calls, separate from the creator agent and the QA agent. Each gets only the transcript and the reference, with no access to the draft or to the other reviewer's verdict. Each checks every spoken number, term, and claim, then returns pass or fail and quotes the line it objects to. The lesson moves on only when both pass. A failure or a disagreement goes to an instructor.
6. Check the frames. Pull a frame every few seconds and confirm every number and label on screen matches the approved script.
7. Publish. The pipeline posts the video to the course platform or LMS through its API as a draft, then releases it under the course's own approval rule.

FAQ

Can these lessons be generated and published automatically?

Yes, on the course platform's side. Using the Golpo MCP server or the Golpo Claude Skill, a pipeline can check the script against the instructor-approved answer and the source it cites, generate the video, run two independent transcript reviews, and post the lesson to the platform as a draft. Anything that fails a check goes to an instructor.

When does a question earn a video?

After it crosses a threshold the course set in advance, a search finds no live lesson that already answers it, and a named instructor approves the reply. The working example threshold is three independent asks in a term. That number is policy, not a measured optimum.

Why isn’t January revenue $1,200 in the example?

The constructed course assumes one obligation satisfied evenly over 12 months. 1,200 / 12 = 100 of revenue in January, and 1,100 remains as a contract liability. The mini-lesson points to the full accounting lesson instead of replacing it.

What should the Python mini-lesson show?

The pinned interpreter’s value of 0.1 + 0.2, which on typical CPython 3 is 0.30000000000000004, and the course’s chosen repair, math.isclose or Decimal. Re-run it and store the printout. Leave round-half-to-even on the programming page.

Who searches for duplicates?

The platform’s QA agent searches, and a person reviews uncertain hits. A hit links the existing lesson. A miss allows a draft. Golpo is not in the search. It renders after approval.

Can the video quote a student’s email?

Not by default. Generalize the question. A specific wrong line belongs to the wrong-step workflow, with that attempt’s access rules. The mini-lesson is for the cohort.

Is this a help-center deflection program?

No. Help-center videos answer product support and are judged partly by tickets avoided. This log answers learners inside a course and is judged by whether the reply matches the canonical lesson and survived a check.

What happens when the parent lesson changes?

The mini-lesson is revised or retired with the parent. An answered log entry stores the lesson ID and the version. A stale ID is an open question again, not a permanent checkmark.

Related guides

• Checked lesson videos — approval and watch-back still apply to a one-minute reply.
• Help center to support videos — customer support, not the learner log.
• SaaS accounting lessons — the parent lesson for the January revenue question.
• Programming explainers — the round-half-to-even parent, kept off this float question.
• The learner’s wrong step — a specific submission, not a generalized question.
• Training updates — retire the mini-lesson when the rule it cites changes.

Count the repeated question against a threshold you set first, link the lesson that already answers it, and draft a mini-lesson only when that search misses.