Advantage Gap
Advantage Gap
Nathaniel Whittemore's crystallization (June 2026): the gap in value extracted from AI between power users and casual users is widening sharply — and OpenAI's ChatGPT super-app overhaul is best read as a UX bridge intended to drag the median user closer to power-user behavior patterns. The same logic applies to Anthropic's Claude Code surface stack and to Codex Plugins (Role-Specific) / Codex Sites.
The data point
Sarah Friar (OpenAI CFO, public interview, June 2026):
| Tier | Turns / day |
|---|---|
| Free | ~7 |
| Plus (first paid tier) | ~15 (≈2× free) |
| Plus ($20) | ~3× free |
| Pro | ~11× free |
But the gap isn't only quantity — power users are using AI differently. They're not asking 11× as many trivia questions; they're running agents, loops, and parallel tasks.
Three cohorts, three widening gaps
| Cohort | Primary mode | Trajectory |
|---|---|---|
| Vanguard (lab internal: Cherny, Steinberger, Ryberg) | Loops as Core Primitive — designing the loop, not the prompt | Pulling further ahead fast |
| Power user (Plus / Pro / Claude Code daily) | Agents — multi-turn, parallel Codex tasks | Compounding |
| Casual user (free / occasional) | Chat — questions and one-shots | Linear gains only |
Inflection point: ~November 2025 → January 2026, when coding tools opened up to non-developers and the agent paradigm went mainstream.
Why the UX overhaul matters
Whittemore's reading of the ChatGPT redesign (vs the investor-class reading that it's IPO theater):
- The labs have seen the engagement data — they know the gap is compounding.
- Pure pricing won't close it — casual users won't suddenly run loops because tokens are cheap.
- The way you close it is by embedding power-user behaviors into the default surface: surface coding, surface external apps, push agents to the foreground, make Sites trivial to spin up. Lead the behavior through the UI.
Whittemore: "I think it's reductive to think that [revenue] is the primary motivation… as opposed to getting more people to experience the insane power and opportunity that comes with actually being able to run agents at scale."
Cross-references
- Loops as Core Primitive — what the vanguard is doing.
- Tasks to Responsibilities Shift — the structural shift that makes the gap compound.
- Task Imagination — uneven distribution of this skill is part of why the gap widens.
- Token Maxing — the seat-based → usage-based pricing shift is the business model expression of the same gap.
- Cognitive Offloading / Productive Resistance / Is AI Making Us Dumber (Charlie Gedeon, TEDxSherbrooke) — the other asymmetric outcome: casual users de-skilling while power users compound. The same gap, viewed from the cognitive-effects side.
- Disposable Software / Codex Sites — the casual-user UI bridge.
- ChatGPT Super App — the canonical example.
Connects to your work
This is the vault-grade headline argument for why an enterprise-IT AI upskilling program isn't optional — without intervention, the variance inside a team will widen, not narrow. Pair with:
- Unlocking 10X in Domain Masters as AI Gets Better — the gap as opportunity for the domain-deep folks.
- Designing IT Roles for an AI Era (Talent Strategy POV) — the role redesign that operationalizes "everyone closer to power-user mode."
- DRAG for AI Upskilling at Manila IT Site — the rollout plan.
Sources
- The Way We Use AI is Changing — coinage + the engagement-multiple data