Turn training videos into learning that keeps moving.
Give learners a focused video experience with progress, contextual guidance, knowledge checks and a clear next step.
Interactive demo · No signup required
AI Foundations for Teams · Week 3 · Understanding LLMs
A 45-second preview of the complete flow — watch, guide, check, next action. Interactive demo · simulated playback
Watching is the start of the journey — not the finish
This page demonstrates the Watch stage. In the product, every stage is a lesson in the same course — completing them is what moves progress.
Not a media library — a queue with a next action
AI Foundations for Teams · Week 3 · Lesson 2 of 4
Understanding Large Language Models
Quick check Concept demo
During prediction, what does the model actually choose at each step?
In production, knowledge checks are quiz lessons between videos — this in-player version is a concept we're testing.
Lesson complete
Understanding Large Language Models · 8:20 watched
A simulated player stands in for the video so the demo loads instantly. In the product, uploaded videos play in the browser's own player with captions from an uploaded caption file, and an interactive transcript below. Progress, resume, captions & viewing analytics: production functionality
What this lesson covers
How large language models represent text as tokens, why context windows matter, how prediction actually works, and where these models fall short.
- Explain tokens and context in plain language
- Describe next-token prediction
- Name two practical limitations
Up next in Week 3: Weekly Knowledge Check · 4 questions · ~3 min
Chapters
In the product, the interactive transcript plays this role: the current line is highlighted, and clicking any line jumps there. Production functionality
Notes Concept demo
Timestamped notes are a concept we're testing — they aren't stored in the product yet, and these live only in this browser tab.
Ask AI Mentor Interactive demo
Responses here are canned for the demo. In the product, Sublime Intelligence answers from the learner's real lesson context — "Explain this more simply" and "Quiz me" are its actual prompts. "Ask about a moment" is keyed to the chapter at that timestamp, not real video understanding.
Lesson complete ✓
Up next: Weekly Knowledge Check
4 questions · about 3 minutes. Key concepts from this lesson: tokens & context, next-token prediction, model limitations.
The same watching, seen from every role
Key question: What should I do next?
- Current video with live progress and resume from where they left off
- This week's queue: video → knowledge check → apply activity
- AI Mentor a click away, inside the lesson
- Completions counting toward the course and its certificate
Lesson: Understanding LLMs · Week 3 Demo data
Lesson insights
- Average watch progress78%
- Most revisited chapterModel prediction (04:10)
- Most skipped section06:35 – 07:10
In production, per-learner watch %, rewatches and skipped sections come from real per-second viewing telemetry. Production functionality
Needs attention
- Maya R.Video 0% · check not started · active nowOn track
- Aisha K.Video 84% · check incomplete · last activity 2 days
- Noah T.Video 32% · last activity 4 days Review
Nudges are a staff action in the product — nothing contacts a learner automatically. "Prepare Nudge" here is simulated.
Across the workspace Demo data
Program comparison
- AI Foundations for Teams81%
- Security Essentials74%
- Manager Onboarding68%
A completed video shouldn't be the end of the learning signal
- Watch
- Engage
- Check
- Signal
- Guide
- Progress
Adding a training video takes a form, not a pipeline
Add Training Video Interactive demo
- Uploading video…
- Upload complete
- Preparing lesson…
- Video ready
- Generating captions… Concept demo
- Captions ready
In production you upload a video (or link one) and attach a caption file — captions are not auto-generated today, so that step is a concept we're testing.
AI-assisted lesson setup Interactive demo
Ask Sublime Intelligence to draft the supporting material around a video — always as a draft for review, never auto-published.
Large language models turn text into tokens, hold them in a context window, and generate output by predicting the most likely next token — powerful, but bounded by context limits and confident-sounding mistakes.
Tokens · Context window · Next-token prediction · Hallucination
00:00 What is an LLM? · 02:15 Tokens and context · 04:10 Model prediction · 06:35 Limitations
"What does the model choose at each step of generation?" · "Which limitation explains a confidently wrong answer?"
The product's AI course builder drafts course structure and quiz questions for review today; this per-video flow is the same principle, simulated for the demo.
See where learning is moving — and where it slows down
Video completion, week by week
Maya's engagement timeline Demo data
- Started lesson
- Asked AI Mentor about tokens
- Revisited "Model prediction"
- Completed lesson
- Passed knowledge check
Figures on this screen are demo data. The production analytics behind them: per-lesson started/completed counts, average watch %, rewatches and skipped sections from real viewing telemetry, and quiz pass rates.
What happens when a learner stops halfway?
Noah T. Behind
- Video progress32%
- Last activity4 days ago
- Weekly progressBehind cohort pace
Noah's learning signals
- Video32% · last watched 04:10
- Knowledge checkNot started
- AssignmentNot started
- Cohort paceBehind
Noah stopped during Model prediction and hasn't continued.
- Resume point prepared — 04:10, Model prediction ✓
- Mentor message prepared for review ✓
- Learning queue updated ✓
This flow is simulated. In the product, resume points and the recovery list of open activities are automatic; reminders and mentor messages are prepared and sent by staff, never automatically.
The whole journey, start to signal
Transcript
- 0:02 The admin opens the course's training videos and adds a new one.
- 0:08 Title, details — and the MP4 itself.
- 0:14 Save — the video lesson is ready to publish.
- 0:20 Assigning is one click — pick who should watch.
- 0:26 The course shows exactly how many people are assigned.
- 0:32 Teachers are assigned the same way — separately from students.
- 0:38 The teacher opens their assigned training video…
- 0:45 It plays in the real player — the teacher's own progress is tracked as they watch.
- 1:02 The student opens the same video — it resumes right where she left off, at 3:51.
- 1:10 She skips ahead — the player records the jump as a skipped section.
- 1:17 Her own analytics count only what actually played — never the skipped parts.
- 1:29 Now the admin. Course progress shows every video and how the class is doing.
- 1:37 One video, the whole class: who completed it, who is partway, who never started.
- 1:46 And one learner, second by second: watched, skipped, not reached — with every skipped section listed.
Where guided video learning fits
- Corporate trainingRole skills delivered as weekly video + check + apply loops, with completion visible to managers.
- ComplianceMandatory videos with real viewing records, knowledge checks and exportable completion status.
- CertificationVideo lessons that count toward course completion and an automatically issued certificate.
- BootcampsCohort weeks where the video is the setup for live sessions and project work.
- AcademiesStructured programmes where instructors see who watched, who's stuck and who needs a hand.