AST Theory
May 14, 2026 · View on GitHub
AST = Audience-State-Transfer.
PPT is not an information container. PPT is an audience state-transfer artifact.
Humanize PPT exists because AI-generated decks often fail before visual rendering starts: raw material is passed directly into a slide generator, so the model's reasoning noise, explanatory habits, summary voice, and overstuffed structure leak into the final slides.
A — Audience
Ask first:
- Who is listening?
- What do they already know?
- What do they not realize yet?
- What do they resist?
- Why would they keep listening?
- What judgment or action should they make after the deck?
S — State
A deck should change an audience state.
initial_state: "Before seeing the deck"
desired_state: "After seeing the deck"
core_tension: "The main cognitive resistance"
state_shift: "The transition the deck must create"
T — Transfer
Transfer is not a table of contents. It is the audience's psychological path.
transfer_path:
- role: hook
purpose: grab attention
- role: conflict
purpose: break the old frame
- role: method
purpose: explain the new frame
- role: proof
purpose: build trust with evidence / demo
- role: takeaway
purpose: leave a portable judgment
Experience sources
AST abstracts lessons from several operating styles:
- Karpathy-style structure: explicit assumptions, small viable structures, and verifiable implementation paths.
- Jobs-style presentation: first-screen clarity, ruthless hierarchy, taste, and subtraction.
- Musk-style propagation: tension, screenshot value, memetic hooks, and public discussion energy.
- Garry Tan / YC Office Hours: real user need, narrow wedge, falsifiable assumptions, and product value.
- Humanizer / humanizer-zh: remove AI writing traces, filler phrases, fake significance, and mechanical rhythm.
- Carl's PPT testing practice: AI PPTs often look finished but fail because they do not know what the presenter is trying to make the audience feel, understand, and do.
Acceptance test
A Humanize PPT outline is valid only if it answers:
- Who is the audience?
- What is the initial state?
- What is the desired state?
- What is the core tension?
- How does each slide move the state forward?
- Which renderer / adapter should complete the job?