presentations on S Anand

The Curator's Dilemma - VizChitra 2026

Sat, 11 Jul 2026 11:56:37 +0530

Last week at VizChitra, I ran a “Dialogue” session. A new format for me.
I usually speak 80% in my workshops.
In this dialog, I spoke 20%.
The group discussed.

PART A
I showed 6 charts and said, “Pick the best.”
Then I shared the audience & purpose and asked:
“For THIS audience and purpose, will you publish, fix, or kill it?”
INSIGHT: almost no one said, “Ship”. That’s good – these were all drafts.

I also asked, “How will you verify it?”
No one knew.

PART B
I rotated the groups and asked them to critique the other groups’ critique.
That’s when I learnt that even internally, groups were divided.
It didn’t come out when they could only share ONE verdict.
It DID come out when discussing across two groups.
That was my biggest surprise - AHA moment.
Independent verification matters!

I asked them to share their final verdict:
“Which chart will you put your name against?”

The top-voted chart chart was 100% AI generated.
It was, also, something the Times of India literally published.

My takeaways:

We don’t really know how to verify charts.
We don’t agree on whether a given chart fits a given purpose.
We don’t often see disagreement, especially since we often ask for ONE opinion.

Full story: https://sanand0.github.io/talks/2026-07-04-vizchitra-dialog-curators-dilemma/

PS: This post is 100% human generated. I mean, I wrote every word, without even consulting an LLM. Pangram declared it 100% human-generated. Still, I can’t shake that feeling… that it smells AI-generated. Does this have a name?

Speaking unprepared

Tue, 07 Apr 2026 13:14:13 +0800

I deliver about 3-5 talks a month and usually prepare for them.

Thanks to AI (but even otherwise), I have a steady stream of new content. So, I just to assemble the story.

For example, in my TEDx Whitefield talk “Prisoners of Birth”, I shared the impact of name, gender, lineage, place, and time of birth. I didn’t execute any new analysis. I just cherry-picked disparate analyses into a theme. (Took me three days to plan, though.)

But repetition bores me. I’ve known this since 2002 when I tried coaching CAT students. Conventional teaching isn’t for me. And talks get boring too.

So, often, I ad-lib. Impromptu. Deliberately under-prepared. (Panels are good practice for this. I must try improv next.)

This has a surprising benefit. In the pressure of the moment, inspiration can strike.

For example, yesterday, at an internal Straive fireside chat, I went completely off-script and answered a question on the chat about the benefits of Gemini.

The sort of things that you can do with Gemini is amazing. And what I do … is to poke … what can I do here? What’s in here? What’s in here? Recently, I saw that it can create music. And some of this music is stunning!

Now, I had to think of something interesting to do with the music. Despite weeks of exploring Lyria, I never found a use for it, let alone “stunning”. Now I had 3 seconds.

Then, inspiration struck!

Now, we had music at the beginning of this Zoom call. Why does that need to be something that is unpersonalized?

I went to ChatGPT and asked it to create a personalized jingle for the talk I was delivering, providing it my name, title, talk topic, etc. - something the L&D team could replace their stock jingle with. And it was pretty good, actually.

That’s a novel and a useful idea - something many people can use!

It’s times like this that I really enjoy delivering talks. This is why I live-code. This is why I pause for questions. This is why I’m thrilled when I have no clue of the answer. Pretending to be know is fun and all that, but the kick of discovery on stage – that’s something else!

Using browser tabs as slides

Sat, 14 Feb 2026 17:12:20 +0800

My last two presentations used browser tabs as slides.

For my talk last week titled Your Chotu Is Smarter Than You Think, I planned to show a series of examples. I loaded them all in a browser window as tabs like this:

Once loaded, I can press Ctrl+PgDn to move to the next - just like I’d press the right arrow key in a slide deck. I can also use the mouse to click on the tab if I want to jump around.

But there’s one advantage I missed from slides. I can add title slides, section dividers, etc.

Since web pages are so versatile, I vibe-coded a slide tool by roughly saying:

Give me a single page “slide” tool
Let me edit the title, subtitle, fonts, colors, backgrounds, etc. via a (barely visible) button on the top right
Store this in the URL so I can bookmark and share it

(Note: I had 5 min on 4% battery and my laptop couldn’t connect to the Internet. This was voice vibe-coded on the web and the PR accepted without review.)

That let me create a far richer presentation.

The “slides” allow me to add structure and remind the audience and me about the key points.

For all the bad press PowerPoint receives, I don’t think presentations are a bad format. But today, there are so many more ways of presenting that using slideshow software seems a bit outdated.

UPDATE 18 Jun 2026: My talks in the last few months use this technique almost without exception.

AfterSlides: Write Slides After Talks

Sat, 20 Sep 2025 10:47:56 +0000

25 years ago, Mr. Krishnan (IAS) amused us with anecdotes of bureaucrats writing meeting minutes before the meeting.

This week, I flipped that. I wrote slides after the talk. I call them AfterSlides.

Why. I ran a couple of Ask-Me-Anything (AMA) sessions where the audience set the agenda. I learned their interests. They got answers. No slides prepared.

How. I okayed recording with the organizers, recorded on my phone, transcribed with Gemini, and asked ChatGPT to generate the AfterSlides.

Here are a few AfterSlides from AMAs at:

GALE: https://sanand0.github.io/talks/2025-09-15-llm-ama-gale/
BI Worldwide: https://sanand0.github.io/talks/2025-09-18-llm-ama-bi-worldwide/

I see potential for this in:

Education. Turn teacher lectures into AfterSlides students can revise. Add a quiz.
Management. Turn reviews into AfterSlides for reference and action.
Sales. Turn pitches into Afterslides for clients to revisit
Research. Turn interviews into evidence AfterSlides with citations.
Support. Turn calls into policy/FAQ slides your team can search.

There’s derived content that LLMs can generate.

Quizzes. Non-trivial questions to test understanding.
Errata. Fact-check, cite sources, add confidence.
Counterpoints. Add credible alternative views with references.
Feedback. What worked, what to improve.

Prompts I used.

Transcription: https://github.com/sanand0/prompts/blob/main/transcribe-talk.md
AfterSlides: https://github.com/sanand0/prompts/blob/main/afterslides.md

If you try this and find new uses for AfterSlides, please share!

Command Line Slideshows in Bash

Wed, 26 Feb 2025 02:29:42 +0000

At PyConf Hyderabad, I spoke about uv. It's a package manager for Python.

I usually mix live demos into my narrative. So, rather than present with something static like PowerPoint (or Google Slides), I usually use:

Front-end: Custom HTML mixed with RevealJS and CodePen. Observable is a good, too.
Python: Jupyter Notebooks. Marimo is good, too.
Others: Markdown and VS Code for most other things, e.g. SQL.

For this talk, I needed to run commands on the shell. I evaluated:

VS Code + Terminal. Split screen is good. But slides in VS code were not obvious.
Web App. Write a web shell with xterm.js and node-pty and embed it in RevealJS. But it's too much work.
Web terminals: WeTTY, ttyd, GoTTY, etc. But they struggle on Windows. I'd need WSL or Docker.
Asciinema. But it's not interactive.

So I got ChatGPT to write me an app:

Write a modern, compact Python program that parses a Markdown file and renders it section-by-section colorfully on the terminal.

A "section" is any text beginning with a heading until the next heading.

- uv run talk.py script.md should parse script.md and render the first section.
- Running uv run talk.py should render the next section. And so on.
- If no further sections are found, it should say so and end.

When rendering on the terminal,

- Headings should be very prominent. Highlight H1, H2 and H3 in decreasing order of prominence. Rest can be rendered normally
- **Bold** should be prominent. _Italics_ should be mildly emphasized.
- Code blocks and
  code fences
  should be colored distinctly.
- [Links](...) should be colored distinctly but the URLs can be ignored.

Use inline script dependencies. I think using rich and markdown2 would apt but you can decide.

An aside. These days, it's easier to create small tools than search for something that exists.

The code it wrote works like this.

Write a Markdown file that has my "slides". I used this README.md.
Run slide.py README.md. It shows the first section ("slide") in README.md, colored and highlighted, and exits.
I can run any other commands on my shell, e.g. uv run --with pandas,ipython ipython, and show how it works.
Run slide.py again. It clears the screen and shows the next slide.

This allowed me a new kind of workflow, where the shell itself is the slides layer.

Comments

Voice coding is the new live coding - S Anand 21 Sep 2025 4:48 pm (pingback): […] In Feb 2025 at PyConf Hyderabad, I tried a new slide format: command-line slideshows in bash. […]

Clone any voice with a 15-second sample

Thu, 24 Oct 2024 01:36:21 +0000

It's surprisingly easy to clone a voice using F5-TTS: "A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching".

Here's a clip of me, saying:

I think Taylor Swift is the best singer. I've attended every one of her concerts and in fact, I've even proposed to her once. Don't tell anyone.

(Which is ironic since I didn't know who she was until this year and I still haven't seen or heard her.)

You'll notice that my voice is a bit monotic. That's because I trained it on a segment of my talk that's monotonic.

Here's the code. You can run this on Google Colab for free.

A few things to keep in mind when preparing the audio.

Keep the input to just under 15 seconds. That's the optimal length
For expressive output, use an input with a broad range of voice emotions
When using unusual words (e.g. LLM), including the word in your sample helps
Transcribe input.txt manually to get it right, though Whisper is fine to clone in bulk. (But then, who are you and what are you doing?)
Sometimes, each chunk of audio generated has a second of audio from the original interspersed. I don't know why. Maybe a second of silence at the end helps
Keep punctuation simple in the generated text. For example, avoid hyphens like "This is obvious - don't try it." Use "This is obvious, don't try it." instead.

This has a number of uses I can think of (er... ChatGPT can think of), but the ones I find most interesting are:

Author-narrated audio books. I'm sure this is coming soon, if it's not already there.
Personalized IVR. Why should my IVR speak in some other robot's voice? Let's use mine. (This has some prank potential.)
Annotated presentations. I'm too lazy to speak. Typing is easier. This lets me create, for example, slide decks with my voice, but with editing made super-easy. I just change the text and the audio changes.

Things I Learned - 23 Jun 2024

Sun, 23 Jun 2024 00:00:00 +0000

This week, I learned:

Luma Labs Dream Machine generated videos. It’s free and is of reasonable quality. Update: 6 Jun 2025. Costs $10/month
LLM DataHub has LLM training datasets, regularly updated
From Dan Becker on running a workshop
- Answer questions at the end, not in parallel in a chat, to avoid distraction
- Have fewer words in slides when presenting. It’s less distracting
Morgan Housel Shane Parrish podcast
- Risk is what stops you from achieving YOUR goals. What’s risky for me may not be risky for you
- The lesson from compounding is that you want to optimize for duration, not return. That’s what does the heavy lifting. Survival, consistency, long term - these matter. The performance does NOT matter.