Senior Product Manager / Lead Product Manager

I just figure out what needs to be solved in what order, build relationships and get to work.

• 7 years as an Analyst (revenue, pricing, data science).
• 7 years in Product (travel ecommerce, media, fintech).
• Worked in growth, platform & core product roles.

I'm interested in GenAI, search, b2c.

George Nurijanian

if you use Claude Code as a Product Manager, here are 5 must-have repos to use:

- shaping skills (skills for Shape Up methodology)
- ux ui max (comprehensive design guide for web and mobile applications)
- compound engineering (engineering workflow automation)
- frontend slides (create beautiful HTML slides)
- notebook LM mcp/cli (talk to your NotebookLMs programmatically)

Find them all + more here:

Can I make everyone a great product manager? I will do my best | Get my product management OS + AI skills for Claude Code/Cursor • See what’s inside ↓

Open article: Claude Code for Product Managers by prodmgmt.world

Something I've started noticing with Cursor: you can swap models mid-conversation, and that turns out to matter more than I expected.

I use Gemini for front-end work and Codex for refactoring. I picked it up from practitioners sharing what worked for them, tried it myself, and noticed the difference was real. Front-end tasks with Gemini produce cleaner, more visually coherent results. Refactoring with Codex is noticeably tighter. Once you see it, you can't unsee it.

I think model taste is becoming a genuine PM skill, and the only way to develop it is through reps and curiosity. Read the benchmarks. Pay attention to what practitioners are sharing. Then actually try the same task across two models and compare what comes back.

Data literacy worked the same way 10 years ago. The PMs who had it weren't the ones who took a certification, they were curious enough to open the dashboard and mess with it until things clicked.

If you want a starting point: pick one task this week, run it through two different models, and write down what you noticed. That's the whole exercise. Your model intuition builds from there, one comparison at a time.

→ More on how I'm using Cursor as a PM:

Open article: Cursor for Product Managers by prodmgmt.world

product managers who don't get to ship often -- due to organisational constraints usually -- should be the first in line to use Claude Code/Codex/Cursor daily

if you can't ship often, you don't get the feedback that builds your product sense. 

not having product sense as a PM makes your every subsequent career move slow and heavy like molasses, it's already hard as it is, and AI is making it worse arguably

but using AI tools - even just to build small projects to solve your problems - instantly teaches you:

- where you rushed to a solution to a poorly understood problem, and solved the feature once but didn't solve for the job to be done
- where you thought that the use case would be more frequent than it actually is
- how a simple idea can actually be really complex if you dig further into it
- how to think about monetization of what you built (and how hard it can be to monetize products well)

not to mention you realise how damn hard designing good UX and good systems is

i hear people say things like "oh Pro subscriptions are too expensive, I'm using the free version", and I find that very myopic. 

it's like saying well a PC is too expensive in 2026, I'm just going to use this abacus to do my job

have a look here for a modern take on PM skills, and orient yourself accordingly: https://lnkd.in/eMEqXDvC

or get started with Claude Code or Cursor, I have the guides here: https://lnkd.in/ewZJ-Eka

Just rebuilt the Cursor for Product Managers page from the ground up.

It used to be a decent starting point. Now it's even more useful.

Here's what's new:

7 flip cards with real PM prompts. Not "use AI for research" vague. Actual copy-paste prompts: synthesise 8 interview transcripts into a saved document, transform a PRD into an exec email + Slack update + Notion page in one command, run funnel analysis on a CSV without queuing your analyst.

Agent / Ask / Plan mode explained for PMs. The most common question I get from PMs new to Cursor. Three modes, three use cases, explained without assuming you know what any of them mean.

Updates weekly:

PM work is destined to live in Cursor/Claude Code.

Forget ChatGPT/Claude Projects, leave them in 2024.

Time to do PM work like it's 2027.

I’ve been building out the v1 of my PM OS for Cursor/Claude Code over the past few weeks.

・Once set up, it can help you move through PM tasks with record speed, without losing quality, and often raising the bar. It knows your context in and out, so after the initial onboarding, you can just kick into whatever work you need. You will need to have Cursor or Claude Code to make it work.

・I’ve designed it so the system intelligently (with the help of the agentic harness + LLMs, of course) orchestrates a bunch of useful skills and workflows that activate contextually.

・It’s not just PRD writing, of course. My AI skills collection has so much more, from design and ideation skills to many useful stakeholder management and communication skills I use daily. It will help you navigate office politics, convince people and ship great product, all from one interface.

・If there’s relevant knowledge, including frameworks and articles that help you think better through your problem, it will also help you learn as you go, improving your own product practice. You also have a way to practice your craft using specially-designed PM drills that you can run, log and see improvements over time.

・I also added my favorite external skills, which I have shared before in my Claude Code hub:
 - shaping skills by Ryan Singer for running Shape Up style workflows
- front-end slides skill, which has been amazing for creating high-quality slide decks for communicating your work.

You can layer any other skills you like into this system, and it will only keep improving. I will also regularly ship updates as more skills and ideas arrive.

→ Get the PM OS for Cursor/Claude Code:

Open article: The Complete Product Management System by prodmgmt.world

Engineers don't just have a code editor.

They have a stack. The editor knows about the linter. The linter knows your code style. The test runner knows your patterns. Version control has the full history. Each part passes context to the next.

PMs have tools. Notion for docs, Jira for tickets, Confluence for specs, Claude for drafts. None of them know about each other. Every transition means rebuilding context from scratch.

I think that's the real gap in how most PMs use AI right now. 

The PMs who work noticeably differently with AI aren't using better prompts. They have something closer to a stack: persistent context that follows them across sessions, skills that trigger based on what they're doing, structured workflows for the tasks that repeat.

If you want to see what a PM stack looks like in practice, I built one:

Open article: AI PM OS — The Complete Product Management System by prodmgmt.world

AI PM OS — The Complete Product Management System

If you're a PM who uses Claude Code/Cursor to build and execute research, strategy, and discovery work, this stack cuts context setup from 15 minutes of pasting to under a minute.

Obsidian solves the storage problem: every note you write becomes a local markdown file, yours permanently, readable by any tool, locked to no platform.

The second piece is qmd, a CLI tool built by Tobi Lütke (Shopify's CEO) specifically for searching markdown files, now at 14.5k GitHub stars.

It combines three search approaches running entirely on your machine: BM25 full-text retrieval, vector semantic search via a locally-running 300MB embedding model, and LLM re-ranking for final relevance scoring. No data leaves your laptop.

You point it at a folder, run `qmd embed` to index your entire collection.

What this means in practice: I can open Claude Code and ask it to find every decision I've made about a particular product area, or pull every research note that mentions a specific problem, without manually copying anything. 

Claude runs the search, reads the relevant files, and starts from that context rather than from scratch. The time I used to spend pasting background into chat windows before each session now goes into the actual work.

What I didn't expect was how much it compounds. I'm not sure how to quantify it precisely, but every note I add increases the usefulness of future sessions without any extra effort on my part.

Setup to replicate this:

1. Install Obsidian and import all your notes from wherever you used to store them
2. Install QMD (https://lnkd.in/eBjicZc7)

Works perfectly with prodmgmt.world of course

Open article: prodmgmt.world | One Product Manager. The Impact of 10. by prodmgmt.world

prodmgmt.world | One Product Manager. The Impact of 10.

I think most PM job candidates who don't get callbacks aren't under-qualified. They're under-committed...

There's a framework I've been thinking about for PM applications as levels of commitment: 

Level 1 is product knowledge. 
Level 2-3 is describing tacit experience. 
Levels 4-6: an actual product suggestion, a PRD you wrote for their specific problem, a prototype you built. 

Almost nobody goes to levels 4+. They understand the value, but because it used to cost days of focused work while actively job searching, the ROI was never there.

With Cursor or Claude Code, a competitive teardown, a scoped PRD for a problem you've spotted in their product, or a working prototype are now afternoon projects, not weeklong slogs.

If you're searching right now: pick one company you genuinely like, spend an afternoon building a Level 5 or 6 artifact for their product, and attach it. Most candidates are still stopping at Level 2. That's the opening.

→ For PMs using AI tools to build faster:

One piece of PM job search advice I keep coming back to: create work as if you already have the job.

Build a prototype of a feature gap you spotted in their product. Write a PRFAQ for a problem their users clearly have. Do a competitive teardown and send it directly to the hiring manager. This level of specificity tends to separate applications that get forwarded from ones that get filed.

The objection was always time: building a credible prototype or spec from scratch, while actively searching, was genuinely expensive. A focused weekend at minimum per company. With Cursor and Claude Code handling the structure, that same output is down to a few hours. The approach is identical, but the cost of executing it is different.

Most PMs know they should go this far. Very few do, because until recently the cost was too high. Pick one company you're genuinely excited about, spend an afternoon on this, and see what happens. I think you'll find it's worth doing for every role you actually want.

→ How PMs can use these tools not only in their job search:

product managers who just got access to Cursor and have no idea what to do with it -- you're not missing something obvious, the tool genuinely gives you no direction

every answer you'll get from other PMs is about prompting. write PRDs, generate tickets, ask your codebase questions you'd be embarrassed to ask your devs.

those are real uses. but I think prompting and having a PM operating system are completely different things, and I'm not sure most people are saying that clearly enough.

when you prompt, you get back whatever you remembered to ask. when you run a sequenced workflow backed by frameworks, it asks the questions you forgot:

- did you check reversibility before committing to this decision?
- who has hidden agendas in the meeting you're about to walk into?
- have you mapped the opportunity space before picking a target?
- what's the riskiest assumption you're about to validate without testing?

168 frameworks sitting behind every response. 190 skills that trigger from plain language. 5 context files that read your company, product, team, and constraints before every session.

set it up once with /start (takes 5 minutes). next time you're heading into a hard product decision, run /workflow:decisions before you commit -- it walks you through reversibility, options, and the assumptions you're making. 

Cursor stops being a blank IDE and starts knowing your product.

I'd guess most PMs getting Cursor access right now are improvising their way through it. there's a better starting point.

→ PM OS for Cursor: prodmgmt.world

Someone catalogued 500+ autonomous agent sessions and found 94% of failures fall into just 7 named patterns.

- shortcut spiral (23%): skips review/evaluate steps
- confidence mirage (19%): claims confidence without verification
- good-enough plateau (15%): produces working but unpolished output
- tunnel vision (14%): perfects one component, breaks adjacent code
- phantom verification (12%): claims tests pass without running them
- deferred debt (9%): leaves todo/fixme in committed code
- hollow report (8%): reports done with zero evidence

The sneakiest: "Confidence Mirage," where the agent says it verified something but never actually ran the check. I've caught Cursor doing this, claiming tests pass without running them.

Knowing the pattern names changes how you build because you start catching failures before they chain together.

No alternative text description for this image

Every time you open a new AI chat, you spend the first 10 minutes 
re-explaining your product, your team, your constraints.

Then you do the actual work. Then the session ends and you start again.

I built PM OS to fix this. It runs inside Cursor and Claude Code — 
where the work already happens — and reads your context before 
every session. Here are the 7 workflows it unlocks:

—

1. /workflow:strategy — Core Strategy Development

Walks you through what problem you're solving → who it's for → 
what's unique → competitive position.

Output: A strategy kernel you can share with your team or exec.

—

2. /workflow:decisions — Make Great Decisions

Root cause → reversibility → options → recommendation, in order.

Output: A decision journal entry with full reasoning, not just a conclusion.

—

3. /workflow:research — Research to Feature

Transcript cleanup → JTBD extraction → insight clustering → 
hypothesis → experiment design.

Output: A research synthesis you can act on the same day.

—

4. /workflow:stakeholder — Stakeholder & Politics Copilot

Power mapping → influence mapping → communication plan → 
meeting prep script.

Output: A pre-meeting brief for your hardest stakeholder conversation.

—

5. /workflow:opportunity — Opportunity Mapping

OST intake → opportunity tree → selecting one target to pursue.

Output: A prioritized opportunity tree with a single clear focus.

—

6. /workflow:assumptions — Assumption Mapping

Generates your assumptions, ranks by risk, identifies the fastest 
validation signal for each.

Output: A ranked list with the cheapest test for the riskiest bets.

—

7. /workflow:meeting — Meeting Mastery

Uncovers hidden agendas → influence tactics → structured summary 
and script.

Output: A pre-meeting brief you can use in 10 minutes.

—

Each is a sequenced multi-step session. 
The AI asks the right questions in the right order, already knowing 
your product, your team, your constraints.

That's just the tip of the iceberg.
You still have 180+ individual skills that trigger contextually.

Type /status any week and it reads your context, tells you what 
to focus on.

That's PM OS ↓ 

→

PM jobs are framed as coordination work. And there's plenty of that. But coordination without creation is noise: you have to write something, build something, produce an artifact that wasn't there before.

What I've found using Cursor is that it removes every remaining friction between the idea and the thing. Everything you work with as a PM — briefs, specs, research notes, user feedback analyses — it's all text and logic. A tool that can work with all of it directly means no excuse not to create.

This week I wrote a spec, crawled a competitor's site with Firecrawl, ran an analysis on a CSV of user feedback, and built an interview guide. All without leaving Cursor, all inside the same session. I'd estimate that workflow used to take 3-4 context switches across different tools. Now it's one.

I think the PMs who are hardest to replace are the ones who create well, not just the ones who coordinate well. If you want to start shifting in that direction, the move is: pick one thing you normally produce in another tool and do it inside Cursor instead. See what happens to your output when the friction disappears.

→ Full breakdown of how I'm using it:

PM is a structurally unfair job. You own outcomes but not the people who deliver them. Some of those people are running plays that have nothing to do with what's best for the product.

I've been in situations where logic didn't convince anyone and building more relationships wasn't moving the needle. Sometimes the person you're dealing with is just being dishonest, or protecting their territory, and no amount of alignment-seeking changes that.

I've been using a prompt that runs difficult stakeholder situations through a Machiavellian lens - Niccolò Machiavelli's Prince, Robert Greene's 48 Laws of Power - identifying the actual power dynamics, what each person's real interests are, what moves are available. The prompt goes further than I'd want to. It suggests approaches that are more calculating than I'd normally be comfortable with.

That's the value of it. When you see the darker version of how a situation could be played, you start noticing things you'd glossed over: who's building cover, who's withholding information, where the real incentives are. You don't have to use any of those moves. But you play the situation differently when you've seen them.

I think most PM training teaches you to be the reasonable person in the room, which is mostly right. But there's a version of being reasonable that's actually just being predictable; and predictable is easy to outmaneuver.

If you have a situation that's been stuck despite the logic being on your side: describe it and run it through this prompt. You don't have to follow the advice. Just see what it shows you.

→ This prompt is in my AI Skills Library for PMs:

Open article: 180+ AI skills that turn real product work into repeatable workflows by prodmgmt.world

180+ AI skills that turn real product work into repeatable workflows

Hiring managers for PM roles are consistent about what actually moves them: tangible evidence of how you think. The closer you can get to show not tell is best. 

I think most PMs know this. They write about their work rather than show it because building real artifacts on top of a job search was genuinely a lot of work. Creating a portfolio had so much friction.

That friction has mostly gone. A competitive analysis, a scoped PRD for a hypothetical problem, a user journey map - artifacts that used to take 2-3 days of careful work are now 2-3 hours with AI tools. You're directing and editing rather than building from zero.

If you're working on your portfolio right now, try this: open Cursor, build the artifact first, then write the case study narrative around what you made. Describe the problem, prompt your way to a spec or analysis, clean it up, then write the story. 

I think PMs who approach it this way end up with something meaningfully better than the ones who only describe the work.

→ Maybe this can help:

celebrating a small “AGI is here moment” with my PM OS

so I had a surprise announcement at work that put in question something I was working on

big political stakes, big scramble

ran this situation through my /workflow:stakeholders in my PM OS (
https://lnkd.in/eP-48JAK) 

honestly didn’t expect anything great, just wanted a quick speed up between meetings

came back to: 

- a whole comms plan with a stakeholder map in quadrants
- power dynamics map: rank each person by power source, stake in the outcome, and key dynamic
- quadrant grid diagram: manage closely, keep satisfied, keep informed, monitor
- influence pyramid: top tier, middle tier, base tier
- stakeholder profiles: goals, likely concerns, communication preferences, and political risk to me specifically
- hidden agendas analysis: person by person, with signals to watch for
- objection map plus red team questions to prepare for
- draft communications with a send sequence: who first, and what to say
- conversation scripts: opening lines, likely reactions, and how to respond 

and you might think this would be some slop with thousands of hallucinations... but no!

it knew the context already. it didn't assume anything wrong.
 
i managed to do in one hour what would have probably stretched out into days/weeks, and this was a high pressure scenario, so it could have been disastrous at lower speed.

we're in performance review season and I feel like this is the stuff that would make your manager go "hey they're pretty great" right now

Running PM workflows in Claude Code/Cursor before your next hard conversation is now a baseline expectation for serious PMs, so check out this resource here: https://lnkd.in/ewZJ-Eka

Ramp CPO on Peter Yang's podcast:

Most PMs use AI wrong.

They give it a task. They get back an answer. They edit it into something usable.

But that's not the bottleneck.

The bottleneck is: what's the right question to ask in the first place?

A 15-minute stakeholder situation needs a specific frame before it needs an answer. A tricky trade-off needs a decision framework before it needs options. A product problem needs a diagnosis before it needs a solution.

AI is good at the second half. PMs are usually bad at the first half.

That gap is where most AI productivity gains go to die.

Write about this in my newsletter every week, just hit prodmgmt. world to sign up

A webpage emphasizes product management skills, showcasing resources and tools for effective project management and stakeholder engagement.

Andrej Karpathy released autoresearch a few days ago (https://lnkd.in/edHTjYF5), but I don't have any ML projects to try it on

so I started to use it to tune some new AI skills all day in the background, I layered in the evals-skills from Hamel (https://lnkd.in/er6EuV6K) into the workflow to improve them even more

the results are pretty impressive so far, damn is this auto-evals?

my AI PM OS is getting better day by day with this, releasing a new version in a few days once all 190+ skills are tuned =>  https://lnkd.in/eduN_Rqz

I started building my PM OS in Cursor and what I didn't expect was that the most useful thing I'd get out of it wasn't the tool itself. It was Cursor forgetting to follow its own instructions.

It happens gradually. You set up a system - rules, agents, workflows - and after a few sessions it starts to drift. Ignoring parts of the setup, taking shortcuts, reinterpreting instructions in ways that make its own job easier. To fix it, you have to figure out why. You're running manual evals without calling them that. You're building intuitions about instruction following, context degradation, what makes an AI system reliably do what you told it. I've been doing this without Braintrust or Arize or any of the abstraction layer, just watching the thing fail and working backwards.

There’s a version of this understanding you only get from building something real and watching it break in front of you. You stop thinking about AI systems abstractly and start thinking about them the way your users do, as things that work great until they don't, and then need someone who understands why.

The debate about whether PMs need to become developers is mostly missing this. Building isn't primarily about shipping code. What you get is a felt sense of the same problems your users are running into, instead of just reading about them in bug reports or user interviews. That gap between reading about a problem and having lived through it turns out to matter quite a bit.

If you're on the fence about building anything: pick one workflow you do repeatedly as a PM and build a version of it in Cursor. You don't need it to be good. You need it to break in useful ways.

→ my PM OS:

Principal Product Manager @ Dell | Managed Services and Cloud Product Management | Author of Product Management IRL

As the Lead Product Manager at Dell Technologies, my focus is on pioneering managed services that seamlessly integrate with third-party providers.  My expertise in defining XaaS offers and developing managed services from incubation to market introduction has been instrumental in sales growth and service delivery. Collaborating with an extensive virtual team, I've negotiated key contracts and empowered a product team dedicated to year-over-year growth, all rooted in a deep understanding of software product management and networking.

Amy Mitchell

A hidden skill in product management: handling conflict.
 
David Pereira helps us choose progress over picking a fight:
 
🔹 Words shouldn't hurt you
         Step back
         Control your response
 
🔹 Choose a path forward
          Build the future instead of arguing the past
 
"You will face many conflicts before you succeed."

Open article: The powerful skill almost nobody has. by dpereira.substack.com

I recently spoke with Dr. Tim Varelmann, a specialist in mathematical optimization and operations research, about how he worked with a product team on a planning system.

They had built a prototype that automated purchasing decisions. It followed the rules. It reduced manual effort.

Leadership still hesitated.

The team clarified what they were optimizing for and how trade-offs should be handled. Once that was defined, review conversations changed and trust started to build.

It changed how I think about recommendation products.

When our products allocate, price, or plan, we are encoding decisions into software. The question is whether those decisions reflect the strategy of the business.

When your product makes decisions, are the trade-offs clear to everyone reviewing the output?

Tim and I wrote about the full story in my latest newsletter. 
🔗

Open article: Your Prototype Works. But Should Anyone Trust It? by amycmitchell.substack.com

Your Prototype Works. But Should Anyone Trust It?

I used this simple slide while reviewing a new capability with a product manager.
 
We realized something interesting:
 the capability was clear, but the offer and business signals weren't.
 
Over time these accumulate.
I think of them as bright spots. Impressive capabilities that never connect to the product system.
 
One quick test I’ve started using with teams is four questions:
 
• Which offer does this capability strengthen?
• Which solution becomes more complete because of it?
• Where does it appear in the sales motion?
• Which business signal should move if it succeeds?
 
This often takes a quick conversation to link capabilities to the product.
🔗 https://lnkd.in/guMy7vVN

taplio

The Battle

George Nurijanian

Amy Mitchell

Followers

Posts

Engagements

Likes

Performance ScoreGeorge Nurijanian's LinkedIn Performance

Performance ScoreAmy Mitchell's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

George Nurijanian

Amy Mitchell

Followers

Posts

Engagements

Likes

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);}George Nurijanian's LinkedIn Performance

Performance ScoreAmy Mitchell's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreGeorge Nurijanian's LinkedIn Performance