Product Growth

Author of 1 newsletter deep dive and 2 podcast episodes a week on how to succeed as a PM, product leader, or aspiring PM.

Lead a team of folks.

I'm the product guy - making sure everything hits our ever increasing quality bar.

Here's a sample of recent content:

Course 1: Mastering the PM Career

Unit 1: The Key Outputs

Product requirements documents
Roadmaps: Advanced techniques
The feature results writeup
The product review bible

Unit 2: Hard Skills

Advanced discovery techniques
How to identify and validate product risk
Advanced customer interview techniques
How to ramp a product team without strategy

Unit 3: Soft Skills

How to say no
Handling conflict as a PM
How to bridge the divide between stakeholders and users

Unit 4: Career Skills

The PM Career Ladder: Your Unofficial Guide
How to get a raise in this market
Why politics isn’t a dirty word

Course 2: Mastering Growth

Unit 1: Lessons From The Greats

How Miro grows
How Canva grows
How Klarna grows
How did Intel do so Badly?
Everything about Midjourney
What you need to know about Temu

Unit 2: Systematizing Our Learnings

Build a detailed growth model
Gaming growth playbook
When to A/B test
Impact Sizing
Growth loops
How to build a flywheel
How to use Code Interpreter

Unit 3: Success in Specific Product Teams

Information Architecture
Product-led sales
Design Systems
Product Copy
Onboarding
Expansion
Activation

Course 3: Product Leadership

Unit 1: How to set up your product team

How to set up your product team
Do you really need a North Star Metric
Pitch a growth role to senior leadership
How founders think differently from PMs

Unit 2: How to scale your team

Billion-Dollar Execution
How to scale a product team from 0 to 15
Reference checks can supercharge your hiring

Unit 3: How to be a great leader

7 currencies of a product leader
The 7 failure patterns of managers
How to Make Your Ideas Stick
Product leadership in India

Newsletter Author, Podcast Host

Apollo.io

Product leader when the company became a unicorn ($750M->$2.5B valuation).

I led several product teams, including the growth team, pricing & packaging, activation, expansion, information architecture, acquisition, and engagement.

Raised at a unicorn valuation while I was there.

Ended up leaving to focus full-time on my newsletter.

VP of Product

Affirm

Member of the senior product leadership team. Led growth team when company revenue grew from $300M to $1.2B/year.

Head of Growth Product

Epic Games

Led data driven growth for Fortnite during some it's best performing years, including the Travis Scott Concert and launch of "Chapter 2."

Lead Product Manager

thredUP

Built and led the team's product growth team when it raised its series D.

We worked on the following key initiatives:

- Pricing optimization: we re-architected the pricing system to optimize for conversion, new customers, and stockthrough levels, driving industry-leading inventory turns
- Conversion rate optimization: we drove 2.5x in first visit to purchase conversion via tools like outfits, dynamic merchandising, ad campaign customization, and the aforementioned pricing system
- Referrals channel: we grew this channel back when it actually worked, with things like $10-$10 and also a famous $0.99 sales
- Paid advertising growth: we built an ad-technology system to handle budgeting and bids cross-channel across Facebook, Pinterest, and Yahoo ads

Overall, I helped the company eventually go public by setting the foundations for its growth, growth marketing, and growth product teams.

Director of Growth

Rap to Beats

Developed iOS app Rap to Beats & scaled to 300K monthly active users (MAU). Led team.

We shipped several game changing features:
- In-app purchases for premier DJ beats
- A leaderboard to showcase your raps by country
- Social sharing to Instagram and Facebook

With premier DJs regularly using the app to sell their work, we were able to scale one of the first consumer apps for rap on the iPhone.

Co-Founder/ Head of Product

Scoutforce Werving & Selectie

ScoutForce was a SaaS for College Athletic Recruiting Teams. 

I joined as one of the first employees, and helped create the initial product: in the cloud recruiting software to enable recruiting teams. I did in depth discovery, wireframing, and iteration on the early features. 

This built my initial skillset in product management.

Product Manager

The Product Growth Guy 🚀 | Helping you land your next job + succeed in your career | Newsletter Writer and Podcast Host

TL;DR: I write a newsletter for product managers, product leaders, and aspiring product managers. I focus on actionable and deep tips to improve your career.

The Longer Version:

After a long career in PM - rising to VP of Product at a Unicorn - I have made the shift to writing about it.

In particular, I write about product growth. That is, the art of growing via your product. 

This means I also talk a lot about product-led growth (PLG), the SaaS B2B motion.

My newsletter doesn't just focus on growth though. 

I also break down technology, product management, product leadership, and how to get a PM job. 

It's all the things that make for a great use of your learning & development budget. 

Or even personal spend. 

The math is pretty great. 
↳ $150 for an annual subscription
↳ $211K average PM salary in the US
↳ 0-10% average range of raises offered

If the newsletter helps you get a 0.07% better raise, it pays itself. 
That's 1/142th of your raise range.

It's a pretty smart investment for every level of PM.

http://www.news.aakashg.com

This is what a Senior AI PM at Google's Github looks like.

It's wild:

Take a look:  https://lnkd.in/e7Jb8kRe

He's made 484 contributions in the last year (>1/day). 

New-age AI PMs don't just write PRDs. They ship:

1. Prototypes
2. Frameworks
3. Perspectives on AI
4. Open source contributions

The bar for technical fluency for AI PMs continues to rise as PMs like Shubham Saboo raise the standard. 

AI PMs don't just have a theoretical understanding of concepts, but a practical knowledge of AI engineering.

Take a look at Shubham's top repos, for inspiration:

1. Collection of LLM Apps: https://lnkd.in/e-Zesb7u
2. MCP Playbook for AI agents: https://lnkd.in/eR5KHvww

These are practical resources that he's collected himself. They aren't necessarily at-scale production apps. They're PM friendly projects. 

He's also built a more-technical open source Python toolkit for building AI agents: https://lnkd.in/eCsAzPtR

When you put these types of work products together, jobs come to you. This is how you break into AI PM. 

If you look at his LinkedIn, most of his career was in dev rel. But a Github speaks louder than titles.

This is what all aspiring AI PMs should do:

1. Start with Prototyping: https://lnkd.in/eJujDhBV
2. Then pick up Cursor: https://lnkd.in/d2pcXD7R
3. Build with Claude Code: https://lnkd.in/eUyPEAma
4. Ship to Github: https://lnkd.in/eVFVT_Yc
5. Create a Portfolio: https://lnkd.in/eT3RvQiM
6. Build a Job Search System: https://lnkd.in/efbppEDj

Then you too can become an AI PM at Google like Shubham. He's created the roadmap. Now follow.

📌 Want my complete guide to the AI PM job search? Comment 'AI job search' + DM Me. Must do both.

➕ Follow Aakash Gupta for daily AI PM tips. 

Proof of work > titles

Aakash Gupta

The AI PM Guy 🚀 | Helping you land your next job + succeed in your career

You don't need a CS degree or MBA to become an AI PM

If I had to learn it from scratch, I would do this:

1. Understand what is an AI PM

All PMs have to be AI-powered, but only ~20% of PMs build AI products and features. 

🔗 Guide: https://lnkd.in/e3akMRm8
🔗 Foundations: https://lnkd.in/e6zyYugs
🔗 Break in: https://lnkd.in/efEvq-Mu
🔗 Course: https://lnkd.in/eqFzK6tr

2. Master prompt engineering

Sounds basic, but is the basis of great AI products. 

🔗 Core techniques: https://lnkd.in/d_qYCBT7
🔗 ChatGPT guide: https://lnkd.in/eyfUi8ZN
🔗 Prompt library: https://lnkd.in/eHU88esH

3. Learn context engineering

This is the moat for modern AI products.

🔗 Context engineering: https://lnkd.in/eUUPMmJK
🔗 RAG comparison: https://lnkd.in/ebfnDUmi
🔗 Customer intelligence: https://lnkd.in/evCwj4rU
🔗 Building AI: https://lnkd.in/eDGmsvZ5

4. Learn AI Prototyping

A prototype replaces 1000 meetings.

🔗 Guide: https://lnkd.in/eDVWcaPz
🔗 Interviews: https://lnkd.in/e66DrW-h
🔗 Production: https://lnkd.in/eNB4mQY7

5. Learn agents
   
They are the next frontier of AI products.

🔗 Practical: https://lnkd.in/eyamsW7a
🔗 Fundamentals: https://lnkd.in/eeey5Cxr
🔗 Copilot: https://lnkd.in/eAZZ8MSb
🔗 MCP tutorial: https://lnkd.in/dVp93fRE

6. Master AI testing
   
This is the new critical PM skill. 

🔗 Evals: https://lnkd.in/eGbzWMxf
🔗 Infra: https://lnkd.in/eKhmKYNp
🔗 LLM judges: https://lnkd.in/ez3stJRm
🔗 Observability: https://lnkd.in/e3eQBdMp
🔗 Experimentation: https://lnkd.in/e86mpjGR

7. Ship to production
   
You must go from toy to scaled.

🔗 AI PRDs: https://lnkd.in/eMu59p_z
🔗 Product strategy: https://lnkd.in/egemMhMF
🔗 Roadmaps: https://lnkd.in/eEuNcj-m
🔗 PRDs: https://lnkd.in/ef5Yde54

8. Land the job
   
OpenAI pays $569K, Google $557K, Anthropic $549K

🔗 Job search: https://lnkd.in/efbppEDj
🔗 Transition: https://lnkd.in/efqv4qUc
🔗 Product sense: https://lnkd.in/ekWty_bw
🔗 OpenAI: https://lnkd.in/ebR_Auss
🔗 Interview prep: https://lnkd.in/eZZ_vvQc

This is the roadmap I wish existed when I started.

Talks about #growth, #fintech, #startups, #technology, and #productmanagement

AI + Product Management 🚀 | Helping you land your next job + succeed in your career

Shirounda Rice

Linear hit a $1.25B valuation with just 2 PMs.

The way they build product is from the future:

They're contrarian.

They do almost no "growth product" work:

• No A/B tests
• No work on activation
• No growth dashboards
• Zero focus on retention tactics

They also don't have a design system. 

Instead, at the core of the way Linear builds product is "The Linear Method." Many founders I talk to have been religiously copying it. There's three keys, as I see it:

1. Directness
They don't do user stories, they speak clearly

2. No Busy Work
They don't do sprint planning, or OKRs for ICs

3. Momentum
They don't have cycles, scope aggressively, ship fast

Their founders, and CEO Karri Saarinen have operated this way from day 1. They set a very intentional culture. 

Beneath the method, there's tons more texture... 

One of the most unique processes they have is the roast. Before every launch, the entire team spends 30 mins breaking every feature with a massive doc. 

Another one is their zero bugs policy. I heard from multiple people of bugs being solved in hours, not days.

They also don't have a PM on every project. With 2 PMs, small little teams of designers and engineers really think like product people in the work that they do. 

The results speak for themselves.

$1.25B valuation with just 100 employees.
145% net revenue retention. 
No retention team needed.

They've put on a masterclass in building a unique product culture, with lots to learn for others out there.

Want to go deeper?

• My Deep Dive on Linear: https://lnkd.in/e_R9ruDG
• Head of Product Nan Yu: https://lnkd.in/e3WDcETz
• AI Agents Demo: https://lnkd.in/etXnfc2C

📌 Want my breakdown of their PLG iceberg? Comment 'Linear Iceberg' + DM Me and I'll reply to your DM.

I can't help but think most teams will move this direction.

Linear just hit a $1.25B valuation with 100 employees.

The way they work is singularly unique:

I spent 6 months studying them.

I spoke with CEO Karri Saarinen, COO Cristina Cordova, Head of Product Nan Yu, and Head of Design Conor Muirhead: https://lnkd.in/e_R9ruDG

That's the longest I've ever studied any company. 

Here's what I learned about efficiency:

1️⃣ Profitable > growth at all costs

Linear has been profitable since 2021, three years after founding. They've chosen to grow the company profitably from before it was cool. This instills discipline.

2️⃣ Opinionated by design

Their product has a POV. It's purpose-built with one really good way of doing things. Speed is a feature of the tool, which makes task management more pleasant.

3️⃣ Product-led growth perfection

Across all 7 layers of the PLG motion, Linear is really humming. From its organic word of mouth to its 145%+ NRR. There's very little growth hacks here though. 

4️⃣ Quality at the core

One of the most important things at Linear's core is its zero-bug policy. When I spoke with customers, they literally had anecdotes of bugs being solved in hours.

5️⃣ Figuring out AI conservatively

Linear hasn't gone full-bare everything AI into the product. They've been more measure. Integrations with coding agents, yes. Full AI makeover, no. 

6️⃣ A unique product philosophy

Linear lives the "linear method." Designers and engineers think like PMs. PMs don't project manage every feature, they focus strategically in key areas.

I can't help but thinking this is the playbook for companies of the future. The ZIRP era of burning VC cash is getting rarer and rarer. 

3 resources to learn more:

1. Deep dive on their product: https://lnkd.in/e_R9ruDG
2. Coding agent demo: https://lnkd.in/etXnfc2C
3. Interview with Nan: https://lnkd.in/eCaC6M6k

📌 Want my take on their culture? Comment 'Linear PM culture' + DM me and I'll respond with the graphic. 

Are you going to adopt some of Linear's principles?

Want to pass PM Interviews at OpenAI, xAI, Meta, Google, Airbnb, Dropbox, Palantir?👇 

1. Understand these important concepts:

- Business: P&L, metrics trees
- Technical: APIs, scalability, performance
- Roadmapping: now/next/later, alignment
- AI/ML: prompt eng, RAG, fine-tuning, evals
- Experimentation: when to A/B test, p-values
- Prioritization: RICE, value vs effort, opp cost
- Discovery: customer interviews, user research
- Metrics: north star, guardrail, leading v lagging
- Product strategy: future trends, big tech choices

2. Practice these case interview questions:

- Product sense: https://lnkd.in/eTqEFtPM
- AI product sense: https://lnkd.in/ekWty_bw
- System design: https://lnkd.in/etTmzNzz
- Product design: https://lnkd.in/encHB-k7
- Product execution: https://lnkd.in/e8t8zS9q
- Product estimation: https://lnkd.in/eZvjP4RR
- Product strategy: https://lnkd.in/eqzpahmW
- Success metrics: https://lnkd.in/efdj_QYT
- Vibe coding: https://lnkd.in/e66DrW-h

3. Practice these behavioral questions:

- Tell me about yourself: https://lnkd.in/eYmRYF2Y
- Why want to work here: https://lnkd.in/eeuRsR9M
- What's your current pay: https://lnkd.in/gS-CKH9P
- Improve app speed: https://lnkd.in/dG9EFhiR
- Greatest weakness: https://lnkd.in/ePQ7dBtr
- Were you laid off: https://lnkd.in/e2iTP3xs
- Leadership style: https://lnkd.in/eCPdQVp2
- What level targeting: https://lnkd.in/e2YQ8tK2

4. Study these company specific processes:

- Microsoft:  https://lnkd.in/eXWWYefH
- Amazon: https://lnkd.in/eA_WxXHQ
- Apple: https://lnkd.in/eXWWYefH
- Google: https://lnkd.in/ezAHYXhV
- OpenAI: https://lnkd.in/ebR_Auss
- Meta: https://lnkd.in/evHQ_w-E

Image: Yangshun Tay

———

♻ Repost to help others in your network
📕 Save the post so you can find it in future
📌 Comment 'Interview guide' + DM me for all the links
💡 Follow me Aakash Gupta and my company account Product Growth for 2x daily AI PM insights

You want to use AI agents. But you don't know where.

Steal my top 21 AI agent use cases:

For this week's deep dive, I set out to answer the question: "what are all the PM use cases of agents?"

After surveying the landscape, this is what I found:

🔗 https://lnkd.in/eyamsW7a

COMMUNICATION & DOCUMENTATION

1. Email & Slack drafting
2. Meeting intelligence 
3. PRD generation 
4. Release notes automation 
5. Executive reporting 
6. Changelog maintenance 

AI can do 80% of the draft writing for you, if you enable your agents with the right context and iterate on the system prompts. Freeing you up for high leverage work. 

RESEARCH & INTELLIGENCE

1. Competitor monitoring
2. User feedback synthesis
3. Interview analysis
4. Market trend identification
   
Agents track what's happening in your market 24/7. They can surface patterns from customers, pull insights from research, and flag competitor moves automatically.

DATA & ANALYTICS

1. Automated reporting
2. Experiment interpretation
3. Funnel analysis
4. Cohort insights
5. Metric anomaly alerts
6. Usage pattern detection
   
Instead of pulling manually, agents can run your weekly reports, flag anomalies, interpret A/B tests, and find user behaviors. You get to focus on insights, not grunt work.

GTM & MARKETING ENABLEMENT

1. Sales battle cards
2. Landing page copy
3. Social media posts
4. Case study drafts
5. Product demo scripts
   
Agents can turn specs into sales collateral, landing pages, and demo scripts. Sales stays armed with fresh intel, and you don't become the GTM bottleneck.

AI agents help solve the age-old PM problem: "how in the heck do I find for all the things that are my responsibility?" Which is why all PMs should adopt them. 

Want to build these yourself? I have 5 resources for you:

1. Step-by-step guide: https://lnkd.in/eyamsW7a
2. Agents primer: https://lnkd.in/eeey5Cxr
3. Agents demo: https://lnkd.in/eekC6bwT
4. Relay CEO tutorial: https://lnkd.in/eXDEhupa
5. Lindy CEO tutorial: https://lnkd.in/e7_YSFUw

📌 Want me to send you my mind map of agents tools? Comment 'agent tool mind map' + DM me (do both).

AI agents are making some PMs much more productive.

Don't be left out.

I'm seeing a new PM interview everywhere.

Here's your guide to AI product sense:

Everyone from Figma to Meta is asking this new interview type. So it's not really narrow yet. 

I break down the types of questions I've heard candidates are asked in this round to 7 categories:

1. AI Product Improvement
 
Here, they're testing: "can you improve existing products within realistic constraints?"
 
Examples: 
↳ Claude's code generation has a 40% thumbs-down rate from non-technical users. You have 2 months and 5 engineers. What would you build? 
↳ ChatGPT has a 32% hallucination rate on factual questions. You have 3 months. How do you reduce this without access to the research team?
 
2. AI Product Design
 
Here, they're testing: "can you design specific AI experiences and interaction patterns?"
 
Examples: 
↳ Design a conversational UI for an AI image editor that combines generation and editing. 
↳ Create an interaction model for an AI code review tool that works in VS Code.
 
3. AI Product Growth
 
Here, they're testing: "can you hit aggressive growth targets despite AI constraints?"
 
Examples: 
↳ Claude has 50M MAU. Get to 100M in 6 months without increasing the research team. 
↳ Your free AI product has great retention but 99% never convert. How do you 2x daily usage in 1 quarter?
 
4. AI Product Strategy
 
Here, they're testing: "can you think at company level with multi-year horizons?"
 
Examples: 
↳ You're CEO of Anthropic. What should your product strategy be for the next 5 years? 
↳ OpenAI has ChatGPT, DALL-E, Sora as separate products. Should they consolidate?
 
5. AI Product Launch
 
Here, they're testing: "can you think 0-to-1 with appropriate scoping?"
 
Examples: 
↳ OpenAI wants an AI scheduling assistant. You have 8 months and a team of 10. What do you build? 
↳ Design an AI-powered fraud detection product for e-commerce that could challenge Google Shopping.
 
6. AI Safety & Ethics
 
Here, they're testing: "can you build responsibly without breaking the product?"
 
Examples: 
↳ Users are jailbreaking ChatGPT to write abusive content. Stop this without breaking legitimate use. 
↳ An AI product you built is showing bias toward certain demographics. How do you fix it at the product layer?
 
7. AI Metrics & Pricing
 
Here, they're testing: "can you measure AI products and build sustainable business models?"
 
Examples: 
↳ Design a metrics framework for an AI code assistant. What's the north star? 
↳ Create a freemium model for Claude where 98% of users stay free but you're still profitable.
 
Want to prep for it? 3 Resources to help:

Interview guide: https://lnkd.in/ekWty_bw
Skill guide: https://lnkd.in/ezbB5CES
Mock: https://lnkd.in/eCXmkWX2

Want me to send you the high-res version of the infographic? Comment 'ai prod sense qs' + DM me.

Repost this to help your friends
Save it for yourself.

This interview will be everywhere soon.

🛠️🤖 How to build AI agents from scratch
(Even if you've never done it before.)
𝗧𝗵𝗲𝘀𝗲 𝗮𝗿𝗲 𝘁𝗵𝗲 𝟴 𝘀𝘁𝗲𝗽𝘀 𝘁𝗼 𝘁𝗮𝗸𝗲, 𝗳𝗿𝗼𝗺 𝗽𝘂𝗿𝗽𝗼𝘀𝗲 𝘁𝗼 𝗨𝗜.

𝙁𝙞𝙧𝙨𝙩, 𝙬𝙝𝙖𝙩 𝙞𝙨 𝙖𝙣 𝘼𝙄 𝙖𝙜𝙚𝙣𝙩?

An AI Agent is an LLM that:

• Reasons on its own
• Act (via calling tools)
• Observes & improve itself

This is the core "ReAct" framework Google released all the way back in 2022. It's now a reality. 

---

𝙃𝙚𝙧𝙚'𝙨 𝙝𝙤𝙬 𝙩𝙝𝙞𝙨 𝙩𝙧𝙖𝙣𝙨𝙡𝙖𝙩𝙚𝙨 𝙞𝙣𝙩𝙤 𝟴 𝙠𝙚𝙮 𝙨𝙩𝙚𝙥𝙨:

1. Define purpose & scope

Over-scoping is the killer of agents. You want a single clear use case with success criteria, and constraints. 

2. Build system prompt

This is often the most important step. You want a really good prompt, or the whole thing falls over. 

3. Choose LLM

Don't always go for the best. Balance speed, multi-modality, and what each LLM is best for. 

4. Integrate tools

This is what makes your agent an agent, MCP, APIs, other agents as tools. Create your architecture.

5. Add memory

Don't add memory unless needed. But for support agents that need conversation history, or research agents that build on past findings? Game-changer.

6. Orchestrate things

You want a control flow over routes, workflows, message queues and your agent-to-agent (A2A) coordination.

7. Build a user interface

This is the flashy part we want to jump to. But only it do it now. Figure out how users will interact. 

8. Add testing & evals 

This is the most important step to create a continuous improvement loop. Systematically analyze errors.

Here's guides to help with each step:

1. Purpose ➝‬ https://lnkd.in/eekC6bwT
2. System prompt ➝‬ https://lnkd.in/d_qYCBT7
3. Choose LLM ➝‬ https://lnkd.in/eqRgsMZx
4. Integrate tools ➝‬ https://lnkd.in/ezNsTfbD
5. Add memory ➝‬ https://lnkd.in/eeey5Cxr
6. Orchestrate things ➝‬ https://lnkd.in/eKhmKYNp
7. Build a UI ➝‬ https://lnkd.in/epRC64ku
8. Evals ➝‬ https://lnkd.in/eGbzWMxf

---

𝙏𝙚𝙘𝙝 𝙨𝙩𝙖𝙘𝙠:

Now, what tools should you use? 

For general assistants: ChatGPT, Claude, Perplexity 
For coding projects: Cursor, Windsurf, Claude Code 
For business automation: Lindy, Relay app, n8n 
For complex workflows: LangGraph, CrewAI, LlamaIndex

The best stack is the one that gets you shipping fast.

---

Agents are one of the most powerful skills for PMs and engineers to learn right now. 

I have been studying them deeply.
 These are my core learnings.

📌 Want my 8-layer agent architecture? Comment 'agent architecture' + DM me.

What agents have you been building?

A $20B gorilla is coming for the AI prototyping space.

It's Figma (Make):

They just launched in May, went GA in July, and now I'm hearing tons of teams pick it up. 

So I've been testing it: https://lnkd.in/eXA3b_KE

Here's what I learned:

1. Figma benefits from its design roots
2. Being in the same place is a huge win
3. Code in prototyping is mostly commoditized

𝗢𝗡𝗘 - 𝗙𝗶𝗴𝗺𝗮 𝗯𝗲𝗻𝗲𝗳𝗶𝘁𝘀 𝗳𝗿𝗼𝗺 𝗶𝘁𝘀 𝗱𝗲𝘀𝗶𝗴𝗻 𝗿𝗼𝗼𝘁𝘀

The product is made very tastefully. It feels buttery to use. And multiplayer is built in. That makes a difference. 

It's also in the same tool your company is already paying for, so adding some seats is pretty easy. 

Especially if your company did with FigJam earlier. 

𝗧𝗪𝗢 - 𝗕𝗲𝗶𝗻𝗴 𝗶𝗻 𝘁𝗵𝗲 𝘀𝗮𝗺𝗲 𝗽𝗹𝗮𝗰𝗲 𝗶𝘀 𝗮 𝗵𝘂𝗴𝗲 𝘄𝗶𝗻

Figma is where your team has already built your design system. Figma Make can plug right into that. 

This makes it easy to prototype in a way that will "feel" native - to internal stakeholders and users. 

It reduces the delta between AI prototype and "prototype usable for insightful user testing."

𝗧𝗛𝗥𝗘𝗘 - 𝗖𝗼𝗱𝗲 𝗶𝗻 𝗽𝗿𝗼𝘁𝗼𝘁𝘆𝗽𝗶𝗻𝗴 𝗶𝘀 𝗰𝗼𝗺𝗺𝗼𝗱𝗶𝘁𝗶𝘇𝗲𝗱

All of these players have one model underlying them all: Claude. Anthropic's coding capabilities are elite. 

Sure, they all have a mixture of models and their own proprierary stuff, too. But Claude is the code workhorse.

That makes all the design touches in Figma critical.

Try it yourself (free): https://lnkd.in/eXA3b_KE

And here's more resources to do it well:

1. Full guide: https://lnkd.in/eJujDhBV
2. Full tutorial: https://lnkd.in/eEGy9Dri
3. How to go to production: https://lnkd.in/eNB4mQY7
4. How to use it in discovery: https://lnkd.in/e_MhNCkd

𝗙𝗜𝗡𝗔𝗟𝗟𝗬 - 𝗪𝗵𝗮𝘁 𝗱𝗼 𝘆𝗼𝘂 𝘁𝗵𝗶𝗻𝗸?

Will Figma take a significant bite out of the AI prototyping market?

My bet is multiple players will survive, but Figma is going to win a significant market share. 

📌 Comment 'figma make tutorial' + DM me for my step by step AI prototyping workflow. 
➕ Follow Aakash Gupta  for daily AI tips. 

#ProductManagement #FigmaPartner #UX

This $4B CPO runs her entire day out of Claude Code + Cursor. And she shared exactly how she does it:

It was a delight to sit down with Rachel Wolan.

She is one of the most AI-forward CPOs I have met, and she dropped all the sauce on her AI workflows.

🎬 Watch Now: https://lnkd.in/es6aQigP

Available Everywhere
Spotify: https://lnkd.in/eyt7agKj
Apple: https://lnkd.in/eAEVwr3u

🏆 Thanks to our sponsor:

Linear: Plan and build product like the best

Use my link! https://lnkd.in/e8rtjVRZ 

✍️ Here were my favorite takeaways:

1. There are two sides to AI product leadership: being a productive AI product leader yourself, and shipping AI native features for your customers. Rachel showed us both in this masterclass.

2. Webflow went from 0% of designers using Cursor to 30% using it weekly after one builder day. That's the power of setting up your org correctly. Champions teach, incentives align, adoption follows.

3. Rachel built an "agentic chief of staff" that analyzes her calendar for delegation opportunities, triages her entire inbox, and queries Snowflake in natural language. She gets answers in seconds without waiting for data scientists.

4. Her calendar agent spots double-bookings, context switching issues, and meetings she should delegate. It recommends what to cut next week. She runs it weekly & shares outputs with her EA.

5. Her email agent removes junk automatically, pins what matters, and drafts replies based on her inbox behavior patterns. It learns what she cares about by watching how she responds.

6. She queries their data warehouse in plain English through Claude Code. "How many sites does this workspace have?" answered instantly. Set up MCP servers for Snowflake & Tableau to enable this.

7. Building custom agents takes minutes. Create a markdown file in your /agents folder, describe what you want, Claude generates it. She built a LinkedIn post generator live on the pod.

8. To enable your team: give them tool access (Cursor, Claude Code), run builder days where everyone demos something new, update career ladders to expect AI usage, and let champions on your team teach others.

9. For shipping AI features, invest heavily in evals. Webflow's code generator broke when they switched models because eval coverage was insufficient. 

10. Choose AI features that leverage your core strengths. Webflow built a code generator using their CMS, design system, and production hosting. That's their differentiation versus every other coding agent.

11. Think distribution first, not last. Answer engine optimization is the new SEO. Feed FAQs to your website, build ChatGPT apps, and make sure product changes get recognized in answer engines from day one.

What's even cooler is she is a paid subscriber to the newsletter! 

Join her: https://lnkd.in/gp2cCv8K

And don't miss the ep for her live demos!

Tag who you think I should have on next 👇

What does a PM building AI features do?

I've been interviewing 2+ weekly for 14 months:

Here's the latest tools, skills, responsibilities, knowledge

𝟭. 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀

1a. Discovery + Research
Finding where AI actually adds value takes research

1b. Execution Delivery
Getting AI into production is harder than it looks

1c. Strategy Planning
You're planning the future of AI at your company

1d. Leadership Impact
The best AI PMs shape company strategy

𝟮. 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲

2a. AI Fundamentals
Understanding what AI can and can't do is critical

2b. Product Patterns
Designing for AI requires different UX patterns

2c. Market GTM
Positioning AI products requires market knowledge

2d. Testing Taxonomy
Testing AI is completely different from testing software

𝟯. 𝗦𝗸𝗶𝗹𝗹𝘀

3a. Context Engineering
Getting AI to understand what you need is half the job

3b. Evaluation Testing
You can't ship without knowing if it actually works

3c. AI Architecture
Choosing the right model & system design matters

3d. Cross-Team Communication
You're the bridge between tech teams & business

𝟰. 𝗧𝗼𝗼𝗹𝘀

4a. Core AI Platforms
Claude, GPT-5, Gemini & NotebookLM are the top tools of the trade

4b. No-Code AI Builders
Zapier, n8n, Lindy, Linear & Replit are the most used

4c. Prototyping/Vibe Coding
Bolt, Cursor, Claude Code & v0 ship prototypes fast

4d. Analytics Monitoring
DataDog, Amplitude & Arize track AI performance

With so many things to do and learn...

PM is really a "do everything" role. It's hard to have work-life balance. Here's resources to stay sane:

𝟭. 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀

1a. Discovery → https://lnkd.in/e9QrMEDw
1b. Execution → https://lnkd.in/eajEfG5R
1c. Strategy → https://lnkd.in/egemMhMF
1d. Leadership Impact → https://lnkd.in/eymXeBmE

𝟮. 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲

2a. Fundamentals → https://lnkd.in/e6zyYugs
2b. Product Patterns → https://lnkd.in/eq8xvCAK
2c. Market GTM → https://lnkd.in/eJj4bx_i
2d. Testing Taxonomy → https://lnkd.in/eKhmKYNp

𝟯. 𝗦𝗸𝗶𝗹𝗹𝘀

3a. Context eng → https://lnkd.in/ebfnDUmi
3b. Prompt eng → https://lnkd.in/d_qYCBT7
3c. Architecture → https://lnkd.in/eeey5Cxr
3d. Communication → https://lnkd.in/eg8anRXa

𝟰. 𝗧𝗼𝗼𝗹𝘀

4a. Claude → https://lnkd.in/eZ6UBbAQ
4b. Prototyping → https://lnkd.in/eJujDhBV
4c. Experimentation → https://lnkd.in/e86mpjGR
4d. No-Code Builders → https://lnkd.in/e_fjvHHN

📌 Want my free 5 day course? Comment EEC + DM me.

Follow Aakash Gupta for daily AI PM tips.

Repost to help others ♻️

Expectations for PMs have transformed in 24 months.

Here's the latest AI PM tech stack (Dec 2025):

Traditional PM skills are table stakes. There's real differentiation from mastering these 8 layers, in order:

(𝘚𝘢𝘷𝘦 + 𝘳𝘦𝘱𝘰𝘴𝘵 𝘵𝘩𝘪𝘴 𝘣𝘦𝘧𝘰𝘳𝘦 𝘢𝘯𝘺𝘵𝘩𝘪𝘯𝘨 𝘦𝘭𝘴𝘦 ♻️)

1. Foundation Models

Without understanding the pros/cons of the different foundational models, you can't make informed build decisions.

- AI Foundations for PMs: https://lnkd.in/e6zyYugs
- Complete Course: https://lnkd.in/eqFzK6tr
- Model Comparison: https://lnkd.in/eZ-q_ZxU

2. Prompt Engineering

Poor prompts cost companies thousands in wasted tokens and inconsistent outputs.

- Guide: https://lnkd.in/d_qYCBT7
- Library: https://lnkd.in/eHU88esH
- Claude Skills: https://lnkd.in/eRidpGSR

3. Context Engineering

RAG with Pinecone, Chroma, or LangChain turns generic models into domain experts for your product.

- Context Eng: https://lnkd.in/eUUPMmJK
- RAG vs Fine-tuning: https://lnkd.in/ebfnDUmi
- DBs: https://lnkd.in/ejgF9-eC

4. AI Prototyping

Bolt, Cursor, and v0 let you ship working demos in hours instead of weeks.

- Top 5: https://lnkd.in/eNB4mQY7
- Bolt: https://lnkd.in/eHzfiXxq
- Guide: https://lnkd.in/eDVWcaPz

5. Testing & Evals

Tools like Braintrust and Arize help you catch hallucinations before customers.

- Evals: https://lnkd.in/eGbzWMxf
- Testing: https://lnkd.in/eKhmKYNp
- LLM Judges: https://lnkd.in/ez3stJRm

6. AI Agents & Workflows

Use tools like LangChain to orchestrate multi-step tasks that replace entire workflows.

- AI Agents Guide: https://lnkd.in/eeey5Cxr
- Practical AI Agents: https://lnkd.in/eyamsW7a
- Agent Patterns: https://lnkd.in/eMu59p_z

7. Observability  

Use tools like LangSmith to show you exactly where your AI burns budget and fails users.

- Observability Guide: https://lnkd.in/e3eQBdMp
- AI Monitoring Best Practices: https://lnkd.in/eKhmKYNp
- Cost Tracking Tutorial: https://lnkd.in/eGbzWMxf

8. Product Strategy (The ultimate level)

These decisions determine if your product survives.

- AI Product Strategy: https://lnkd.in/egemMhMF
- Strategy Playbook: https://lnkd.in/e5F2J6pB
- Product Sense Interview: https://lnkd.in/ekWty_bw

📌 Comment 'tech stack' + DM me for the high-res file

And now go try these tools 🤖

Cursor is the fastest company ever to reach $1B ARR.

And it did so with 0 full-time PMs:

It's wild.

But everyone is taking the wrong lesson from this. 

Cursor scaled without traditional PMs, because they’re building a developer tool for developers. 

The model breaks the second you’re building for non-technical users. 

You see, product decisions require: 

1. Understanding customer jobs
2. Translating between the technical + user needs
3. Prioritizing across conflicting stakeholder demands
4. Maintaining strategic coherence as the team grows

Cursor gets away with fuzzy roadmaps and fluid roles because every person in those concentric circles (staff, beta users, early adopters, enterprises) can evaluate a live prototype and tell you exactly what’s wrong. 

Most companies don’t have that luxury:

• Your developers are disconnected from your users
• Your go-to-market team needs a sales roadmap
• Your stakeholders don’t understand API latency
• Your support team needs documentation
• Your compliance team needs audit trails
• Your users can’t code

The Cursor model works when the product is the development environment. For everything else, you need someone to find what’s: 

a. technically possible
b. what’s commercially viable
c. what customers actually want

 That’s the PM role.

Marty Cagan explained it best: https://lnkd.in/eBRg4tVm

Peter Yang broke this news in an interview with Cursor's head of design: https://lnkd.in/emx9STrT

My take?

Copying Cursor’s structure without Cursor’s context is how you end up with features nobody asked for and engineers burned out from scope creep.

I covered Cursor here: https://lnkd.in/eq8xvCAK

The PM role is changing. But the PM is not dying.

PMs need to bring skills that earn them a right at the table. These are things like:

1. AI PRDs: https://lnkd.in/eMu59p_z
2. AI Prototyping: https://lnkd.in/eJujDhBV
3. AI Product Strategy: https://lnkd.in/egemMhMF
4. AI Product Sense: https://lnkd.in/ekWty_bw
5. AI Testing: https://lnkd.in/eKhmKYNp
6. AI Agents: https://lnkd.in/eeey5Cxr
7. AI Evals: https://lnkd.in/eGbzWMxf

If PMs don't have such skills, they may not earn a seat at the table. But I suspect even Cursor will hire PMs soon. 

📌 Want me to send you my mental map of how the PM role will evolve? Comment 'PM Cursor' + DM me. 

The death of the PM is fun to talk about once a year.

But it's vastly overrated.

Image: Yuchen Jin on X

You're still writing prompts from scratch.

Ctrl+c ctrl+v these instead:

The best PMs know when to use AI - and have ready-made prompt frameworks to prompt well.

Here are 8 proven use cases + frameworks.

(𝘚𝘢𝘷𝘦 𝘢𝘯𝘥 𝘳𝘦𝘱𝘰𝘴𝘵 𝘣𝘦𝘧𝘰𝘳𝘦 𝘢𝘯𝘺𝘵𝘩𝘪𝘯𝘨 𝘦𝘭𝘴𝘦 ♻️) 

—

𝗢𝗡𝗘 - Writing AI Feature Specs

No one should be writing PRDs fully by hand anymore.

FEATS walks you through the key input areas so AI can write a PRD that gets 80% there first try.

Still best used out of a project with context.

🔗 Breakdown → https://lnkd.in/eMu59p_z

—

𝗧𝗪𝗢 - User-Facing AI Documentation

Your users don't know what to expect from AI features.

WHENT helps AI write documentation that resonates.

Create more documentation in less time. 

🔗 80/20 writing with AI → https://lnkd.in/er5E5Buf

—

𝗧𝗛𝗥𝗘𝗘 - Simulating User Conversations

Many underrate this use case:

Test scenarios before you ship.

SCRIPT generates realistic user conversations across emotional states, edge cases, and patterns.

🔗 Customer interviews → https://lnkd.in/eAKu8qDp

—

𝗙𝗢𝗨𝗥 - Analyzing User Feedback at Scale

You can process 200 feedback messages in minutes with AI.

BPAINT structures what to ask for: batch analysis, patterns, impact scores, next actions.

🔗 More on AI intelligence → https://lnkd.in/evCwj4rU

—

𝗙𝗜𝗩𝗘 - Competitive AI Feature Teardowns

Stop manually reverse-engineering competitor features.

CFHGB helps AI analyze their capabilities and surface your differentiation angle.

Works best when you paste in their product docs or help articles.

🔗 AI product sense → https://lnkd.in/ezbB5CES

—

𝗦𝗜𝗫 - Building System Prompts

Production system prompts need more than "you are a helpful assistant."

RCFTEG structures role, context, task, format, examples, and guardrails so AI gives you reliable outputs.

🔗 Claude tutorial → https://lnkd.in/eRidpGSR

—

𝗦𝗘𝗩𝗘𝗡 - Creating Training Examples

You need great training data for AI features.

CFHGBT generates examples with controlled variety.

🔗 More on evals → https://lnkd.in/eGbzWMxf

—

𝗘𝗜𝗚𝗛𝗧 - Explaining AI to Executives

Why spend the cycles doing this yourself?

Use GOALCOM to spend 80% less time on status updates.

Speak "exec" → https://lnkd.in/emUujVyd

—

𝗚𝗢 𝗙𝗨𝗥𝗧𝗛𝗘𝗥 - Build a library

These 8 frameworks are battle-tested.

A complete prompt library should go further.

Steal mine → https://lnkd.in/eHU88esH

- 80+ use cases
- LLM-friendly formatting
- Even more hardened frameworks

I spent 3 weeks + 2 years hardening it.

—

📌 Want these in copy-paste format? Comment 'prompts copy' + DM me. 

Follow Aakash Gupta for daily AI PM resources.

Stop reinventing prompts. Use what works.

Google JUST dropped 70 pgs on context eng and it's 🔥

I summarized it for you:

Stateful, personal AI starts with Context Engineering: assembling the right information at the right time.

There's 7 keys:

—

𝗢𝗡𝗘 - 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 = 𝗦𝗺𝗮𝗿𝘁 𝗔𝘀𝘀𝗲𝗺𝗯𝗹𝘆

Your LLM context window is prime real estate. Context Engineering optimizes what goes in:

→ User intent + conversation history
→ Retrieved facts (RAG for general knowledge)
→ Long-term memory (user preferences, expertise)
→ Tool outputs and grounding data

The right information, assembled at the right moment.

—

𝗧𝗪𝗢 - 𝗦𝗲𝘀𝘀𝗶𝗼𝗻𝘀 = 𝗬𝗼𝘂𝗿 𝗪𝗼𝗿𝗸𝗯𝗲𝗻𝗰𝗵

A session is a single conversation with:

1. Events (user messages, tool calls, observations)
2. State (conversation history, accumulated context)
3. Lifecycle (create, interact, close)

Think: one task, one session. When you close it, the session ends but memories can persist.

—

𝗧𝗛𝗥𝗘𝗘 - 𝗠𝗲𝗺𝗼𝗿𝘆 = 𝗬𝗼𝘂𝗿 𝗙𝗶𝗹𝗶𝗻𝗴 𝗖𝗮𝗯𝗶𝗻𝗲𝘁

Memory persists across sessions for true continuity:

- Declarative (facts, preferences: "I'm vegan")
- Procedural (how you work: "I debug like Sarah does")

RAG retrieves general facts from the internet. Memory captures YOUR specific expertise and preferences.

Big difference.

—

𝗙𝗢𝗨𝗥 - 𝗠𝗲𝗺𝗼𝗿𝘆 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 = 𝗘𝗧𝗟 𝗣𝗶𝗽𝗲𝗹𝗶𝗻𝗲

LLMs themselves drive memory creation:

1. Extract relevant info from session conversations
2. Consolidate with existing memories (dedupe, merge)
3. Load into vector stores for retrieval

An LLM-powered ETL pipeline. Automated intelligence extraction from your interactions.

—

𝗙𝗜𝗩𝗘 - 𝗣𝗿𝗼𝘃𝗲𝗻𝗮𝗻𝗰𝗲 = 𝗧𝗿𝘂𝘀𝘁

Every memory needs metadata for production use:

→ Source (which session created this?)
→ Timestamp (how fresh is this?)
→ Confidence (how certain are we?)

Provenance is critical for debugging, trust, and reliability in production systems.

—

𝗦𝗜𝗫 - 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗲𝘀

Two approaches to memory retrieval:

Proactive: Always include in context (user name, critical preferences)
Reactive: Retrieve on-demand based on query (semantic search, vector similarity)

Smart agents know when to pull vs push. Not everything belongs in every context.

—

𝗦𝗘𝗩𝗘𝗡 - 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗥𝗲𝗮𝗹𝗶𝘁𝘆

Building this at scale requires:

• Privacy: User data isolation, encryption, compliance
• Performance: Caching strategies, batch operations
• Scale: Vector stores for semantic search, not SQL

Google runs this across millions of users. The infrastructure matters.

—

This took me 2 hours.

To go further, you need to learn:

1. RAG vs Finetuning: https://lnkd.in/ebfnDUmi
2. Prompt engineering: https://lnkd.in/d_qYCBT7
3. AI agents: https://lnkd.in/eeey5Cxr

📌 Want me to send you the whole PDF? Comment 'context PDF' + DM me. Repost to cut the line.

Follow Aakash Gupta for daily AI tips.

Context Engineering Masterclass by Google

Uber PM: AI prototyping only goes so far.

Is it possible to solve this?

For most PMs still, eng is still the bottleneck. I completely sympathize with Nimisha! 

And don’t mean to call her out.

I think there are 3 ways teams can solve this:

1. Modernizing code

This one is really for engineering leaders to solve. You want to make the code base easy to vibe code with. 

> Are engineers using Claude Code and Devin? Or are they stuck with Github Copilot?
> Can you build and deploy nearly instantly? Or does it take hours?
> Is your code built with Tailwind and microservices? Or too big for AI context windows?

2. AI experimentation:

You don’t always need to make huge backend changes. 

Set up with an Optimizely or Kameleoon where you can allows non-technical users to vibe experiment in a way that uses your design and code base.

This unlocks front-end testing from the engineering dependency almost entirely, with the exception of a code review by the on-call engineer.

3. AI discovery:

Not everything needs to go to production.

Vibe code then send the prototype to Usertesring or Voicepanel. Get real human feedback in just a few hours. 

Resources:

1. Modernizing code:

a. Meta example: https://lnkd.in/eeqvdbvV
b. Tech guide: https://lnkd.in/eD6e6Wgt

2. AI experimentation 

a. Full guide: https://lnkd.in/e86mpjGR
b. Tech review workflow: https://lnkd.in/etXnfc2C

3. AI discovery 

a. Teresa Torres: https://lnkd.in/e7Q6mMpc
b. Detailed guide: https://lnkd.in/e9QrMEDw

Want to stay up with this new way of working?

Follow Aakash Gupta for daily tips.

OpenAI, Meta & Figma all test the same new interview.

And it's HARD:

It's AI Product Sense - a variant of product sense but not exactly the same. 

Nearly all the top paying companies ask it:

1. OpenAI: $569K TC
2. Google AI: $567K TC
3. Anthropic AI: $549K TC

So if you want to get into the top paying AI PM roles, you need to understand and practice this interview.

Sadly, there are absolutely no mock interviews or guides.

Until now:

1. Free mock interview: https://lnkd.in/eESzCxUJ
2. Full guide: https://lnkd.in/ekWty_bw

I've done all the research for you: spoke with hiring managers and candidates who passed it. 

Here's what you need to know. 

1. AI product sense is not one interview

You can get any of these 7 questions: 

Product improvement: https://lnkd.in/eTqEFtPM
Product growth: https://lnkd.in/eqzpahmW
Product launch: https://lnkd.in/eQR3ePYh
Product design: https://lnkd.in/encHB-k7
Safety & ethics: https://lnkd.in/eZZ_vvQc
Product strategy: https://lnkd.in/dve8it6D
Metrics & pricing: https://lnkd.in/eW-zBN-e

Prepare for only one type and be lost. You need to understand all of these. 

2. The rubric has 5 key categories:

a. Structure & framework
b. AI technical fluency
c. Product sense & judgment
d. Communication & collaboration
e. Execution & metrics 

You want to get a pass in all 5, so demonstrate them. 

3. Let's break down a real question

How would you increase weekly active users of ChatGPT image creation from 175M to 350M in 3 months with only 3 engineers?

You'll want to build a custom framework and execute.

I share my example in the slides and video. 

- - - - 

📌 Want my full AI product sense guide with example answers? Comment 'AI product sense guide' + DM me. 
➕ Follow Aakash Gupta for daily AI PM tips. 

This is one interview to prep, not wing. 

What questions have you heard?

How to Ace the AI Product Sense Interview

AI prototyping has completely changed PM.

PMs are now much closer to the "bare metal."

But most AI prototyping tools have a problem:

You lack the visual editor toolbar of most site builders. This makes edits take way longer than needed.

And it disconnects you from the code. 

I've been advising a little startup in Israel. Their founder, Nadav Abrahami, started Wix 20 years ago. 

They've built the solution. And it's GA today.

They're #1 on ProductHunt: https://lnkd.in/eFwEUSxb
Check them out: https://lnkd.in/eynieiJi
Get 1 month free with code DazlXPG

The purpose of AI prototyping tools is three-fold:

1. Reduce risk → https://lnkd.in/eD3GzVeq
2. Align people → https://lnkd.in/eMu59p_z
3. Find edge cases sooner → https://lnkd.in/eEGy9Dri

Because of their speed, you can explore many more divergent solutions. That's why tools that help you explore better solutions are interesting.

Follow Aakash Gupta for daily AI PM tips.

What's your favorite prototyping tool?

I analyzed 250 AI PM job postings. 

Here's what hiring managers want:

1. AI Product Strategy & Lifecycle (94%)

Define product vision, strategy & roadmap for AI/ML features from 0→1 through scale:

• Quality Envelopes → Define accuracy thresholds
• Cascade Architecture → Right-size models
• SUQS Framework → Track success pillars
• Evals-First → Define success first

AI Product Strategy → https://lnkd.in/egemMhMF
AI Prototype to Production → https://lnkd.in/eNB4mQY7
AI Roadmaps → https://lnkd.in/eEuNcj-m
AI Product Sense → https://lnkd.in/ezbB5CES
How to Build AI → https://lnkd.in/eDGmsvZ5

2. Gen AI Technical Depth (89%)

Deep understanding of LLMs, transformers, RAG, fine-tuning, prompt engineering, multi-agent sys:

• Context Engineering → Structure > perfect prompts
• RAG is Default → Retrieve before fine-tuning
• Multi-Agent Architecture → Orchestrate agents
• Know When NOT to Use AI → When traditional > AI

Neural Networks & LLMs → https://lnkd.in/eACQ8Wsu
RAG vs Fine-Tuning → https://lnkd.in/ebfnDUmi
AI Foundations → https://lnkd.in/e6zyYugs
Context Engineering → https://lnkd.in/e-_7SxVd
AI Evals for PMs → https://lnkd.in/eGbzWMxf

3. Metrics, Evals & AI GTM (84%) 

Define success metrics, build evaluation frameworks & drive go-to-market for AI features:

• Trust Transfer → Build measurable trust
• Dual Metrics → Track tech + product
• Eval Maturity → Automate testing suites
• Probabilistic Mindset → Design for uncertainty

Finding Right Metrics → https://lnkd.in/ea8sWSsS
AI Testing → https://lnkd.in/eKhmKYNp
AI Observability → https://lnkd.in/e3eQBdMp
AI Customer Intelligence → https://lnkd.in/evCwj4rU
Product Launch Playbook → https://lnkd.in/eB7s6umA

4. Cross-Functional AI Leadership (78%) 

Partner with AI eng, data scientists, designers & business teams to ship AI:

• Speak 3 Languages → Translate tech-business-user
• Fast Prototyping → Demo before specs
• Model as Teammate → Understand AI strengths
• AI Product Sense → Intuit AI strategy

WTF is an AI PM → https://lnkd.in/eyDvgmZE
AI PRDs → https://lnkd.in/eMu59p_z
Pairing with Design → https://lnkd.in/e_c8eue5
AI Prototyping → https://lnkd.in/eJujDhBV
AI Discovery → https://lnkd.in/e_MhNCkd

More in the comments.
Repost if you found this helpful ♻️
Follow Aakash Gupta for daily AI PM advice.

Drishti Batra

Amplitude just dropped its biggest update yet.

AI Feedback:

For the modern PM, the key is: How do I synthesize data + user insights to make better decisions? 

There's a few roadblocks in your way:

1. Customer feedback is scattered
2. PMs don't always have all the access
3. It's hard to synthesize across data sets
4. What insights we have aren't traceable back

That's where a tool like AI feedback comes in. It aggregates the customer feedback, gives you access, and uses AI to draw traceable conclusions.

It just launched today: https://lnkd.in/ezm-yme5

Amplitude's AI feedback is a natural result of its acquisition of Kraftful, an AI startup I first wrote about all the wayback in 2023: https://lnkd.in/exsqFuWA

Amplitude is completing the arc I predicted when I interviewed their CEO: becoming an all in one insights platform, not just analytics: https://lnkd.in/esuAeUUK

AI feedback is easily one of the best uses of AI for PMs.  

It just went GA today. Will you try it?

#Analytics  #AmplitudePartner #ProductManagement

taplio

Aakash Gupta's Linkedin Analytics

Aakash Gupta

Performance ScoreAakash Gupta's LinkedIn Performance

What is Aakash talking about?

Who is engaging with Aakash

Aakash Gupta's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?

taplio

.css-6ytdur{background-image:linear-gradient(284deg, rgba(39, 43, 72, 1) 0%, rgba(33, 47, 183, 1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Aakash Gupta's Linkedin Analytics

Aakash Gupta

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Aakash Gupta's LinkedIn Performance

What is Aakash talking about?

Who is engaging with Aakash

Aakash Gupta's Best Posts .css-1emoo84{font-size:var(--chakra-fontSizes-md);font-weight:400;color:var(--chakra-colors-gray-600);}(last 30 days)

Ready to grow your LinkedIn presence?

Aakash Gupta's Linkedin Analytics

Performance ScoreAakash Gupta's LinkedIn Performance

Aakash Gupta's Best Posts (last 30 days)