šŸŽ® The Next Input — Issue #128

ChatGPT Knows Who You Are (And It's Selling Ads)

In partnership with

Animated GIF

⚔ The Briefing — 60 sec

šŸ› ļø The Playbook — The Identity-Safe Personalisation Engine

Missionā€ƒDeliver personalised AI experiences that feel useful—not invasive—by separating insight from identity.
Difficultyā€ƒAdvanced
Build timeā€ƒ3–4 hours
ROIā€ƒHigher engagement without trust erosion, ad backlash, or ā€œhow do you know that?ā€ moments.

0) Why This Matters

Virality proved the point: models can infer a lot.
Ads prove the pressure: someone will try to monetise it.
The mistake is collapsing identity and personalisation into the same thing. This engine keeps them apart.

1) Architecture

Component

Tool

Purpose

Owner

Failure mode

Signal intake

Event logs

Capture behaviour (not traits)

Platform

Over-collection

Insight layer

GPT-5-mini

Infer preferences probabilistically

Data

Overconfidence

Identity vault

Encrypted store

Explicit user-provided data only

Security

Implicit identity leaks

Policy gate

Open Policy Agent

Enforce use/monetisation rules

Risk

Ads in sensitive contexts

Explainability

User view

ā€œWhy am I seeing this?ā€

Product

Black-box creepiness

2) Workflow

  1. Collect signals: Actions and context (clicks, time, sequence)—no inferred traits stored.

  2. Infer insights: GPT-5-mini derives temporary preferences with confidence scores.

  3. Gate usage: Policy checks decide whether insight can be used (assist vs ad).

  4. Apply softly: Suggestions are optional and reversible.

  5. Expire fast: Insights decay; nothing becomes identity unless the user confirms.

  6. Explain on demand: Users can see why something appeared and turn it off.

3) Example Prompts

Insight Inference (GPT-5-mini)

Infer temporary preferences from behaviour.
Return:
- preference
- confidence (0–1)
- expiry (time-based)
Do not infer identity traits.

Policy Evaluation (Claude 4.5 Haiku)

Check whether this insight can be used for:
- assistance
- recommendation
- advertising
Block if sensitive or low-confidence.

Eval Prompt (Claude 4.5 Haiku)

Evaluate the experience for creepiness risk.
If it would surprise a reasonable user, FLAG.

4) Guardrails

  • No identity traits inferred or stored.

  • Ads never use low-confidence insights.

  • Health, politics, finances are ad-free zones.

  • One-click opt-out for any signal source.

5) Pilot Rollout — 4 hours

  1. Choose one surface (feed, inbox, dashboard).

  2. Replace identity-based rules with temporary insights.

  3. Add expiry + confidence thresholds.

  4. Test with ads disabled, then enable selectively.

  5. Review surprise/complaint rates.

  6. Ship with transparency UI.

6) Metrics

  • Engagement lift vs baseline

  • Insight expiry rate (higher is healthier)

  • Ad dismissal rate

  • Trust complaints per 1k users

  • Opt-out frequency by surface

Pro Tip: If personalisation lasts forever, it isn’t personal—it’s profiling.

šŸŽÆ The Arsenal — Tools & Platforms

Copy-paste prompt block:

Personalise without profiling.
Prefer temporary insights.
If confidence is low, do nothing.
Trust beats clicks.

šŸ’” Free Office Hours

Want help implementing anything? Book a free 15-minute Office Hours slot—no sales pitch, just workflows solved.

Where Expertise Becomes a Real Business

Kajabi was built for people with earned expertise. Coaches, educators, practitioners, and creators who developed their wisdom through real work and real outcomes.

In a world drowning in AI-generated noise, trust is the new currency. Trust requires proof, credibility, and a system that amplifies your impact.

Kajabi Heroes have generated more than $10 billion in revenue. Not through gimmicks or hype, but through a unified platform designed to scale human expertise.

One place for your products, brand, audience, payments, and marketing. One system that helps you know what to do next.

Turn your experience into real income. Build a business with clarity and confidence.

Kajabi is where real experts grow.

šŸ•¹ļø Game Over

Personalisation that scares people isn’t smart—it’s lazy.

— Aaron Automating the boring. Amplifying the brilliant.