SubstackGitHubBook a call →
← Back to side projects
Prototype

Go1 Content Intelligence Prototype

Interactive prototype showing how an AI layer (Morgan) makes learning recommendations visible, explainable, and improvable, surfacing role signals, a dismiss feedback loop, and an agent activity log visible in Slack.

Product rationale for account-based workflows

The highest-value bet for Go1 isn't better search — it's making recommendations explainable enough that learners trust them, and improvable enough that the system gets better over time. This prototype makes the intelligence layer visible at the moment it matters: when a learner decides whether to engage or dismiss.

What I built

A fully interactive 4-screen clickable prototype exploring one high-value discovery experience for Go1: what happens when learners can see why content was recommended, give structured feedback, and watch the AI agent improve. Morgan — the intelligence layer — is the central character: she explains her reasoning, logs her actions, and lets learners steer her with editable goals.

Why it matters

Go1's core problem is that recommendation quality can't improve without a feedback loop — and that loop doesn't exist today. The burden of relevance sits with the learner. This prototype shows what it looks like to shift that burden to the system: transparent signals, improvable recommendations, and an agent that acts on behalf of the learner rather than waiting to be asked.

How it works

  1. 1Screen 1: Learner homepage — assigned compliance card, recommendation rail with 'Why?' links, Morgan banner
  2. 2Screen 2: 'Why this was recommended' panel — surfaces role match, team activity, and skill gap signals per card
  3. 3Screen 3: Dismiss feedback — learner picks a reason ('not relevant', 'already know this', 'wrong level') to train Morgan
  4. 4Screen 4: Morgan activity log — shows what the agent did today (calendar block, Slack nudge, rec update) and learner profile with editable goals

What it demonstrates

  • Recommendation explainability: 'Why this was recommended' panel surfacing role match, team activity, and skill gap signals
  • Improvable feedback loop: dismiss flow trains Morgan and closes the intelligence loop
  • Morgan agent activity log: shows the learner what the AI did on their behalf

Trust and control

  • ·Learner can edit their own goals to steer recommendations
  • ·Every recommendation surfaces its signals — no black-box outputs
  • ·Dismiss feedback is explicit and tied to a specific reason

Stack and tools

HTMLCSSJavaScriptVanilla (no framework)

What I'd improve next

  • Admin-facing counterpart: org-level curation with signal transparency and confidence scores
  • Metadata enrichment pipeline to power freshness and quality signals
  • Behavioural signal layer beyond enrolments: completion depth, re-opens, in-content interactions
  • A/B test: does explainability improve completion rates or just satisfaction?