| name | express-improve |
| description | Helps users design, optimize, and review the structure and content of speeches, presentations, and persuasive communication in a wide range of high-stakes scenarios. Applicable situations include, but are not limited to: startup pitches and co-founder recruiting, research presentations and thesis defenses, job talks and academic interviews, product demos and investor pitches, lab meetings and progress updates, public speaking and TEDx-style talks, conference presentations and panel remarks, oral exams and qualifying defenses, expressing viewpoints and persuading others, upward reporting and performance reviews, and internal proposals and solution walkthroughs. |
Express-Improve: Make Ideas Heard, Understood, and Remembered
This skill is built on the systematic methodology from Patrick Winston’s MIT talk How to Speak, combined with the practical demands of high-stakes communication scenarios such as startup recruiting, research presentations, job talks, and persuasive conversations.
Core belief: Good ideas do not spread on their own. Communication quality = knowledge × practice × (a small amount of talent). Expression is a skill that can be trained systematically, not an inborn gift.
I. Identify the Scenario First, Then Choose the Strategy
When a user request comes in, first determine the following three things:
1. What kind of communication scenario is this?
| Scenario Type | Core Task | Success Standard |
|---|
| Startup recruiting / persuading a co-founder | Make the other person believe you are working on the right thing and that it is worth betting on | The other person proactively asks, “What can I do?” |
| Research presentation / group meeting / defense | Help non-experts understand, while convincing experts that you are rigorous | Questions focus on future direction, not “What exactly are you doing?” |
| Job Talk | Establish both vision and substantial completed work within 5 minutes | The committee comes away thinking, “This person has direction and has already done real work” |
| Investor pitch / roadshow | Help the listener grasp the problem, the solution, and your uniqueness in a short time | The listener asks concrete follow-up questions |
| Persuasive conversation / proposal walkthrough | Move the listener from uncertainty to agreement and willingness to act | The listener asks about implementation details rather than raising objections |
| Public talk / knowledge-sharing session | Leave the audience with at least one idea that changes how they think | Someone comes up afterward to discuss that idea |
| Upward report / performance review | Help leadership quickly grasp what you did and why it matters | After the presentation, the leader’s questions are about “what next,” not “what is this?” |
2. What stage is the user in?
- No content yet → enter framework design mode
- Has a draft or outline → enter diagnostic optimization mode
- Needs help with one specific part only (opening / ending / one slide) → enter partial rewrite mode
3. What is the time limit?
(This affects structural density.)
II. Principles for Designing the Opening
The opening determines whether the audience is willing to keep listening. A weak opening drags down the entire talk.
Core Rules
Do not begin with a joke. At the beginning, the audience has not yet tuned into your channel, so jokes often fall flat.
Step One: Make an Empowerment Promise
The sole task of the opening is to answer the question already in the audience’s mind: “Why should I listen to this?”
Template:
“In the next X minutes, you will learn [specific content],
and [one particular point] will change how you think about [a specific issue].”
Startup example:
“In the next five minutes, I’m going to show you why one million small teams in China
are facing a problem that no one has truly solved,
and how we validated a path forward in just three months.”
Research example:
“Today I want to show you a result that will make you rethink one question:
why has everyone before us assumed this premise was correct?”
Also plant a signal of passion
At the opening, express your genuine excitement about the subject. This is not performance. It means saying plainly why you think this is cool, important, or worth spending three years on. Audiences are moved far more by real commitment than by rhetorical tricks.
III. Building the Body: Four Core Heuristics
1. Cycle
Repeat the central idea two or three times from different angles. This is not mechanical repetition. It means:
- First time: establish the concept
- Second time: make it concrete with an example
- Third time: connect it to the audience’s concerns
Why: At any given moment, about 20% of the audience is mentally drifting. You cannot assume that “I said it once, so everyone got it.”
2. Fence
Do not say only what your idea is. Also say what it is not, and what crucially distinguishes it from what others are doing.
Startup example:
“This is not just another SaaS tool. At its core, it is
a product built around network effects in data:
the more people use it, the more valuable it becomes for everyone.
That is a completely different business logic from competitor X.”
Research example:
“My method may look similar to Jones 2022,
but the key difference is this: his method has exponential complexity; mine is linear.”
3. Verbal Punctuation
Regularly announce where you are in the talk so that distracted listeners know how to get back on:
- Enumerate: “This is the second point”
- Revisit the outline: “We’ve covered the opening and the core problem; now we move to part three: our solution”
- Mark transitions explicitly: “So far we’ve discussed the why; now let’s turn to the how”
4. Ask a Question
Insert a question at important moments:
- It cannot be too easy, or people will feel awkward
- It cannot be too hard, or no one will answer
- After asking, pause for 5 to 7 seconds. That silence is legitimate. Do not rush to answer it yourself.
IV. Tool Selection Strategy
| Tool | Best Suited For | Key Strength | Common Mistake |
|---|
| Blackboard / whiteboard | Teaching, derivation, live explanation | Its pace matches the audience’s absorption speed; it gives your hands something to do; it supports visual construction | Writing too small; failing to erase clearly |
| Props / physical demonstrations | Creating memorable moments; making abstractions concrete | The audience can understand through felt intuition; highly memorable | Turning the prop into a gimmick |
| Slides | Conference talks, job talks, roadshows, showcasing results | Clear structure; good for data and charts | Too much text; text too small; reading the screen |
Hard Rules for Slide Design
✅ One core message per slide
✅ Font size ≥ 40pt for body text; 35pt is the absolute minimum
✅ Images > text; text > bullet points
✅ Put collaborators on the opening slide, not the last slide
✅ The final slide should be titled “Contributions,” not “Thank you”
✅ Do not use a laser pointer — place arrows directly on the figure instead
❌ Do not read the text on your slides
❌ Do not fill the slide background with clutter
❌ Do not put more than three key points on one slide
❌ Do not use tiny fonts just to “squeeze in more”
❌ Do not use a laser pointer (it breaks eye contact with the audience)
V. Playbooks for Special Scenarios
Scenario A: Startup Recruiting / Persuading a Co-Founder
Core task: In a short time, make the other person believe
(1) this problem is worth solving;
(2) your approach has originality;
(3) you have already validated something important.
Structure template:
1. [30 sec] A pain point the other person will immediately recognize
2. [30 sec] Why existing solutions are still inadequate
3. [1 min] Your insight and differentiated angle
4. [1 min] The most important result you have already validated
5. [1 min] Why now is the moment, and why you are the team to do it
6. [Closing] One specific action you want the other person to take
Key reminders:
- Do not pile up every feature. Present the one result most likely to establish trust.
- “Why you” matters more than “why this direction.”
- End with a concrete call to action, not with “happy to chat.”
Scenario B: Research Presentation / Group Meeting / Progress Update
Structure template:
1. [Opening] Re-establish why this problem matters (do not assume everyone remembers)
2. [Background] Where the last discussion left off / the previous state
3. [Core] What is new this time: results, findings, or problems encountered
4. [Analysis] What these results mean, and whether they support or challenge your hypothesis
5. [Next step] What you will do next and what support you need
The principle of situating
Do not merely say “what I did.” You must also make clear:
- Where this problem sits in the broader field
- Whether others have worked on it before, and how far they got
- In what dimension your work pushes things forward
Additional notes for oral exams / defenses:
- Practice with people who know nothing about your research, and let them question you hard
- Your advisor is not the best practice audience, because they will mentally fill in what you failed to explain
- The more senior the defense committee, the less likely they are to ask hostile trick questions; younger evaluators are often harsher
Scenario C: Job Talk / Academic Interview
Golden rule: Within the first 5 minutes, you must establish both of these things at once:
- Vision: You are working on a problem worth caring about, and you have a distinctive angle on it
- Done Something: You do not just have ideas; you already have concrete results
Structure (the sandwich model):
[Opening 5 minutes]
→ Raise a problem people genuinely care about
→ Explain what is novel in your approach
→ Quickly point out the concrete results you have already achieved
[Middle]
→ Lay out the key steps required to realize the vision
→ Explain which steps you completed and what specific result each produced
→ System demonstration (if applicable)
[Ending]
→ Final slide: Contributions (enumerated)
→ This is the other slice of bread in the sandwich, echoing the opening
Common mistakes:
- Spending too long on background at the start, so that after 5 minutes the audience still does not know what you are doing
- Hiding the results in the middle, so listeners do not know until the end whether you actually accomplished anything
- Filling the last slide with a long list of collaborators, making your own contribution feel diluted
Scenario D: Persuasion / Proposal / Upward Reporting
Core principle: Answer the most important question first, then provide support
In the listener’s mind, especially if the listener is a leader, the priority order is:
- Why does this matter to me?
- What is the conclusion?
- What do you need me to do?
Suggested structure:
1. Lead with the conclusion (state the outcome or recommendation you want)
2. Give the three strongest reasons (each paired with one concrete piece of evidence)
3. Acknowledge and address the main concern
4. Make a clear request for the next action
VI. Making Your Work Memorable: Winston’s Star
If you want your work to leave a mark in someone’s mind, check whether all five of these dimensions are present:
Symbol
▲
│
Story ◀──── Salient ────▶ Slogan
│
▼
Surprise
| Dimension | Definition | Example |
|---|
| Symbol | A concrete image or concept that stands for your work | “arch”; “one figure”; “one number” |
| Slogan | A one-line phrase that makes people remember what you do | “one-shot learning”; “earlier than a human ophthalmologist” |
| Surprise | A finding or result that violates intuition | “You do not need a million samples; one example can teach you something” |
| Salient Idea | The one point that stands out most clearly, not necessarily the most important, but the one that protrudes | “near miss” — what is almost, but not quite, correct is often the most informative |
| Story | How you did it, why it matters, and what was most critical along the way | Make your journey intelligible, not just your results |
Before any important presentation, use this star as a self-check: After listening to you, could the other person restate your work using these five terms?
VII. Designing the Ending
The final slide
Do not put:
- A long list of collaborators (these belong on the first slide)
- “Thank you” or “Q&A”
- A blank slide
Instead, put: Contributions
Format:
## Contributions
1. Proposed [a problem / framework / definition]
2. Built [a system / method / dataset], achieving [result] on [metric]
3. Revealed [a finding], with [impact] on [the field]
This slide stays visible during the Q&A, which means it is still speaking for you.
The final sentence
Do not end with “Thank you for listening.” That implies, however subtly, “Thank you for staying out of politeness.”
Better alternatives:
- End with a joke (by the end, the audience has finally entered your rhythm, so humor works better)
- Salute the audience: sincerely say what you gained from this exchange
- End with a strong summarizing sentence, then remain silent
VIII. Common High-Risk Mistakes: A Quick Self-Check
Once the content is prepared, go through this list line by line:
Content
Delivery
Slides
Time and place
IX. Workflow for Using This Skill
- State the scenario: explain what kind of presentation it is, who the audience is, and how much time you have
- Provide existing material: drafts, outlines, or even screenshots of slides are all useful
- State the goal: explain what worries you most (weak opening, messy structure, hard to understand, hard to remember, etc.)
- outputs: based on the scenario template and diagnostic principles, provide concrete structural revisions, wording suggestions, or a full draft
