Hi, there! Happy New Year 2026!
Welcome to the first ever edition of Work in Beta.
In this edition, we have:
3 research-backed, crazily simple tips to win the AI prompting game
3 videos to learn everything you need to know about LLMs
11 tools that form my daily AI stack
THE ‘HOW-TO’ PLAYBOOK
3 Insane Tips to Win the AI Prompting Game
Image Credits: Nano Banana Pro / Work in Beta
Have you gone through a scenario where your prompts in ChatGPT (or Gemini or Perplexity) have resulted in a grammatically perfect, impressively long and yet somehow hollow and meaningless responses.
What if I told you that by making just THREE simple changes, you can 10x the quality of a GPT’s response to your prompts. Let’s dive in!
First: Give AI a role
One sentence. That's it. ‘You're a freelance consultant who works with local cafes’ or ‘You are a market research analyst specializing in Indian FMCG.’
This pushes the AI from a generic helpful assistant to something with a point of view.
Second: Specify your format
Tell it what you want back. ‘Keep the draft email under 120 words’ or ‘Start with a 10-bullet summary, then a table of recommendations…’.
Without this, you get whatever the AI defaults to - which can end up too long, too rigid, or too eager to impress.
Third: Ask AI to question you first (my favourite)
It helps you be more explicit. ‘Before you write this up, you can ask me up to 5 clarifying questions to have complete knowledge of what is needed’
This is the meta-move - letting the AI surface what's missing. It reduces your cognitive load while improving the result.
The anatomy of a prompt that actually works
In my experience, a good prompt has got these four parts (or at least a flavour of each):
Role: Who is the AI acting to be?
Task: What exactly do you need done?
Context: What does the AI need to know about your situation?
Format: What should the output look like?
I have spent the last year testing and perfecting my prompts - making subtle changes to see response quality. And, I am sharing all of it with you in extensive detail here in this GUIDE.
THE ‘WHAT-TO’ PLAYBOOK
3 Youtube Videos by Andrej Karpathy to Learn LLMs
Image Credits: Nano Banana Pro / Work in Beta
Andrej Karpathy is one of the better known names in the world of AI. A PhD in AI from Stanford under Fei-Fei Li, founding member of Open AI and director at Tesla, he teaches on Youtube (1Mn+ followers) among other things. Of all the videos he has posted, 3 are of incredible value to anyone who wants to understand the foundation of what is happening under the hood of LLMs. However, that needs 7 hours of your time.
So, I have summarised these videos in to a 13-minute condensed podcast which you can listen here.
PS: I have done this on NotebookLM (which I will deep dive in another edition).
However, if you're willing to put in 7 hours, which I would say is 100% worth it, here are the 3 videos (follow the order for most value):
THE TOOL STACK SECTION
11 Tools that Form My Daily Use AI Stack
Image Credits: Nano Banana Pro / Work in Beta
Here is my go-to daily AI tool stack to get work done:
Claude: My default writing and thinking partner (helps create the best drafts)
ChatGPT: My primary LLM tool for work and research
Gemini (including Nano Banana): I use it for images and guided learning; now using for decks and infographics too
Wispr Flow: My keyboard replacement using voice as primary input
Granola: My meeting synthesis tool
Gamma: My default tool to create decks (usually first drafts)
Cursor: My primary IDE to manage day to day work
Antigravity: My secondary IDE for when Cursor credits run out
Lovable: My default prototyping tool to see / build end to end flows
MagicPattern: My default design tool for apps I am building
NotebookLM: My default multi-modal research tool
THE OUTRO
That is all for today.
Thanks for taking 5 minutes to read through this newsletter. Please do let me know what did you think of it especially on how useful was it for you.
What did you think of today’s newsletter?
You can also reply to this newsletter email if you want to share something specific with me. Have a great week ahead and see you again next week!
-PD