The Top 1% Build AI Presentations Differently. Here's How.
About this video
Jeff Su argues typical AI presentations look polished but are thin on substance: text models like ChatGPT, Gemini and Claude produce sharp arguments but not pretty slides, while tools like Gamma and Beautiful.ai produce gorgeous visuals with generic content. He demonstrates a Gamma workflow aimed at analytically sharp and visually polished slides, using a flawed outline pitching Google’s CEO on partnering with Apple. He pastes text in, uses three dashes to separate slides as Gamma recommends, chooses ‘preserve this exact text’ to keep his strong titles, picks a custom theme, and leaves additional instructions blank because he says they tend to conflict with Gamma’s system prompts. Most time is spent not on the first draft but on iterations from stakeholder feedback, which he says takes 80% of the time. Using the Gamma agent he trims dense body text, moves a risk slide after the executive summary, web-searches for a market-share table and chart, verifies and replaces an unsourced 400 million monthly active user statistic with OpenAI’s officially reported weekly figure plus a footnote, converts text to a waterfall chart totaling roughly 3.25 billion smartphones, makes a third column visually dominant, combines two slides with a three-talking-points limit, and translates a duplicated deck to Simplified Chinese, which he rates 70-75% there.
Key takeaways
- Creator’s test for good titles: someone reading only the titles in sequence should understand the full storyline without body text or context.
- In Gamma, paste an outline separated by three dashes, choose ‘preserve this exact text’ to keep titles, and leave additional instructions blank to avoid conflicting with system prompts.
- Use the Gamma agent for iteration: trim body text, move slides, web-search for new data as tables/charts, and verify unsourced statistics by selecting the text so it appears as agent context.
- When combining slides, add precise constraints like ’three talking points maximum’ and follow up to remove images and optimize spacing — vague prompts produce a crammed result.
- Duplicate a deck before translating it, since translation replaces the original; creator suggests downloading both versions as PDFs and using Gemini as a bilingual translator to recommend naturalness fixes.
Episode details: published March 3, 2026 · 12:53 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are Jeff Su’s own claims, shared for education — they are not typical or guaranteed.
← Previous: 4 Skills I’m Learning that AI Can’t Replace (backed by data) · Next: NotebookLM Changed Completely: Here’s What Matters (in 2026) →
All Jeff Su videos · Jeff Su profile · Jeff Su (official site)
