v1.5.0

Generate Becomes a Real Study Workspace

Generate, uploads, lesson activities, companion memory, shared study, and sellable learning data all grew up together. The page is no longer just a prompt box with ambitions.

April 12, 202612 min read
Workspace release

Uploads became context. Activities became evidence. Memory became product.

This release connected Generate, FYP, companion memory, lesson activities, uploads, shared study, and sellable learning data into a deeper learning loop.

Uploads

The upload pipeline got serious

Learners do not only study from pristine typed prompts. They bring PDFs, media, YouTube links, notes, screenshots, and classroom debris. So the upload system had to become more than a file picker with optimism.

v1.5 added transcription, YouTube transcript caching, upload planning, source-backed prompting, OCR batching, transcript shortening, and richer generation paths. The system now has a much better chance of understanding the source before generating from it, which is generally preferable to hallucinating confidently. Bold stance.

  • Media transcription and Transcript API usage tracking.
  • YouTube transcript cache and source text shortening.
  • Upload needs-plan detection before expensive generation.
  • Source-backed lesson, quiz, and flashcard generation.
  • OCR batching, execution tests, trusted origin checks, and resumable upload support.
Learning loop

Activities and memory became first-class signals

Lesson activities, interactive assessment sessions, companion memory, scratchpad evidence, and unified learning events all landed in the same release window.

The result is a stronger feedback loop: lessons can produce practice, practice can produce evidence, evidence can shape the path, and the companion layer can carry useful context forward. Finally, a loop that is not just a loading spinner with ambition.

Lesson activities

Interactive quiz and assessment infrastructure lets practice create structured learning evidence.

Companion memory

The app can preserve learner-relevant context and expose it through companion surfaces.

Path bridges

Generate outputs can connect back into FYP so useful artifacts do not die at the edge of the chat.

Product data

Sellable learning data moved from idea to workflow

The release productized aggregate learning insights with commercial positioning, market enrichment, packaging, release manifests, and admin review workflow.

This does not mean selling private student notes. It means building aggregate, governed learning signals that can become a real business asset while preserving user boundaries. Less creepy data swamp, more disciplined product system. Revolutionary, apparently.

Admin workflow
new

Admin workbench and review-state helpers were added for insight packaging.

Governance
manifest

Release manifests and immutable snapshots make packaged insights auditable.

Evidence
grounded

Insight interpretation, aggregation, and buyer framing gained tests and jobs.

Cleanup

The old placement era started leaving the building

The placement quiz remnants were removed and onboarding continued shifting toward Learning DNA, real activity, and live learner evidence.

That matters because a one-time placement test is a pretty thin way to understand a learner. Lernex is moving toward something more durable: observe what happens, update the model, and keep adapting.

Release takeaway

Generate became the center of a richer evidence loop.

v1.5 is the release where uploads, activities, memory, sharing, and product intelligence started behaving like one system instead of a handful of impressive but lonely parts.

These updates are live product work, not decorative changelog confetti. The goal stays the same: Lernex should learn the learner and make the next step easier to find.

Start learning