Aakash Gupta Profile picture
Apr 30, 2023 20 tweets 9 min read Read on X
The last 45 days have been the biggest ever in AI.

All of this happened:

GPT-4
Auto-GPT
Midjourney v5
Runway #Gen2
Viral Drake Songs
ChatGPT in Robots

Here are the key takeaways from a crazy month and a half in AI 🧵
March 14: GPT-4

We were already collectively losing our minds over GPT-3.5. Then, GPT-4 dropped 🤯

@skirano created pong the very same day.

March 15: Midjourney v5

After having our minds blown by GPT-4's code and text, the very next day @midjourney v5 dropped.

The photorealism was unreal from the get-go. @DotCSV created the iconic sad Darth Vader drinking a Coke. Image
@juliewdesign_ showed us its full power a few days later.

March 20: Runway Gen2

After having our minds blown by text and image, video was next.

@runwayml released the next generation of its excellent text-to-video model.

@ChristianF369 showed us just how powerful the tool can be.

March 30: AutoGPT

We thought there wasn’t going to be another modality to wow us until AutoGPT dropped.

It uses GPT to write itself tasks.

And days later @SigGravitas added code execution.

Autonomous AI became a reality.

April 4: Heart on My Sleeve Released

After text, code, image, and video, what was left? Audio!

We completed the circle on all forms of creative media when TikTok user ghostwriter977 spawned the viral song of the year.

And then he did it again with a Bad Bunny x Rihanna drop.

The music labels have been reporting videos 24 hours a day to take them down.

April 17: DINOv2

We didn't just see advancements in generative AI. We also saw advancements in computer vision.

Meta released mind-blowing transformers to identify moving objects.

April 20: WebGPT

19-year-old genius @willdepue released a way to run chatGPT natively in your browser.

@SullyOmarr explained how this tech opens up the door for a massive play by Apple.
April 25: GPT Robots

Just when we thought the innovation had to slow down, Boston Dynamics put ChatGPT into their robot dog Spot.

Compelling AI in our homes looks closer than ever.

Bonus: #NeRFs took off

Although not released in the last month and a half, #NeRFs powered by @LumaLabsAI did take off.

Artists like @karenxcheng showed us just how powerful the tech is.

I thought we just needed to take a step back in and look at it all.

We are in a Cambrian explosion of AI advancement.

Language
Imagery
Music
Video
Code

Everything's moving at exponential speed - and accelerating.

What a time to be alive.
Want to go the layer deeper - The Midjourney story? The Runway story?

Check out the newsletter:

aakashg.com/subscribe-emai…
This thread took half of my Saturday to write.

If you enjoyed, follow me for regular updates: @aakashg0

And retweet the first tweet for others to see:

Here's all of that in an image: Image

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Aakash Gupta

Aakash Gupta Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @aakashg0

Jul 29
If I had to learn AI PM again, I would start here:

($2,000 course totally free) Image
1️⃣ AI PM Foundations

@marilynika: How to become an AI PM - youtube.com/watch?v=GQd54g…

Sam Kawsarani: 15+ Years PM in Era of AI - tinyurl.com/2txs9mab

@_amankhan: Prototyping → Observability → Evals - youtube.com/watch?v=Ej4pBD…
2️⃣ AI Strategy

@tomocchino: Behind v0 (Top AI Prototyping Tool) - open.spotify.com/episode/2wAfwJ…

Scott Jones: 21 Harsh Truths about PM in AI - tinyurl.com/8zcw26rs

@PawelHuryn: AI basics to expert - youtube.com/watch?v=IfW1FM…
Read 10 tweets
Jul 23
You want to code with AI.

But you don't know what tool to use.

Here's your guide on when to use which AI coding tools:

(+ free resources to learn each) Image
1: Cursor

They surpassed $500M ARR for a reason. Your developer probably uses it. It's great for pros - but it's also great for non-technical folks now, as its agent can build things just via natural language.

Free guide:
2: Windsurf

Despite losing most of its technical staff to Google, Windsurf still exists. Now a part of Cognition (makers of Devin), it's great for non-technical people to deal with large code bases.

Free tutorial:
Read 10 tweets
Jul 17
There's a new PM interview in town.

But most PMs don't know it yet.

Last week, a PM messaged me in panic:

"I just got destroyed in a Google interview. They wanted me to prototype something with AI tools. I'd never even heard of 'vibe coding' before."

Here's what happening:
This isn't isolated. I'm coaching more PMs who are blindsided by this new interview format every week.

Here's what's happening behind the scenes:

The nature of the PM role is undergoing its biggest shift in years.

We're now moving to the era of AI PMs.
Companies are looking for PMs who can:

1. Build AI features (understanding evals & tech details)
2. Use AI to focus on high-leverage tasks
3. Prototype with AI tools like pros

And their interviews are changing to match.

At Google, this is only in India (for now).
Read 7 tweets
Jul 16
>50% of PM interviews now ask metrics questions.

But not all metrics questions are the same.

5 categories + 16 sub-categories to know: Image
𝗦𝘂𝗰𝗰𝗲𝘀𝘀 𝗠𝗲𝘁𝗿𝗶𝗰𝘀 (𝟮𝟭% 𝗼𝗳 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀)

Three major flavors here:

Product launch success: Measuring launching GPT-5.

Feature success: Measuring a mobile payment feature.

Business model success: What would you look at if we launched a marketplace?
𝗗𝗶𝗮𝗴𝗻𝗼𝘀𝘁𝗶𝗰𝘀 (𝟮𝟮% 𝗼𝗳 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀)

Investigating issues when metrics drop.

Sudden drops: IG Stories dropped 5% overnight. Approach?

Gradual declines: Monthly retention declining 6 mos. Why?

Unexpected changes: Feature 30% → 60% engagement. WDYD?
Read 7 tweets
Jul 14
Context engineering is the new prompt engineering.

Here's everything you need to know: Image
1. Etymology

It's unclear who coined the term, but two folks have been particularly important in its rise:

Andrej Karpathy and Dex Horthy.
2. What It Is

It's all about thinking beyond the prompt - thinking about ALL of the tools you have to drive success.

5 are the most important to understand:

a. RAG
b. Memory
c. State/ History
d. Prompt Engineering
e. Structured Outputs
Read 8 tweets
Jul 11
Which is it: use LLMs to improve the prompt, or is that over-engineering?

By now, we've all seen a 1000 conflicting prompt guides.

So, I wanted to understand:

• What do actual studies say?
• What do experts at OpenAI, Anthropic, & Google say?

Here are the answers: Image
I spent the past month in Google Scholar, figuring it out.

I firmed up the learnings with Miqdad Jaffer at OpenAI.

Some of my favorite takeaways from the research:
1. It's not just revenue, but cost

You have to realize that APIs charge by number of input and output tokens.

An engineered prompt can deliver the same quality with 76% cost reduction.

We're talking $3,000 daily vs $706 daily for 100k calls.
Read 10 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(