LLenny's Podcast
← All episodes
Tomer Cohen (LinkedIn CPO)04 December 2025

Why LinkedIn is turning PMs into AI-powered "full stack builders”

5Frameworks
19Insights

Frameworks in this episode

Strategy2 steps

Benefit = (Volume × Quality) / Time

Measure AI-building impact as experimentation volume times quality, divided by time to launch

Read →
Mindset2 steps

Don't Wait for the Re-Org

If you're waiting for a formal declaration to build differently, you're already too late

Read →
Innovation2 steps

Golden Samples Over Full Corpus

Curate a small set of gold examples for your AI, don't dump your whole knowledge base on it

Read →
Leadership2 steps

The Five Traits to Keep, Automate the Rest

Protect where humans shine — vision, empathy, communication, creativity, judgment — and automate everything else

Read →
Innovation4 steps

The Full-Stack Builder Model

Empower one builder to take an idea to market end-to-end, regardless of role or team

Read →

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Myth Buster· 2

Myth Buster17:00

No AI Tool Works Off the Shelf on a Real Codebase

One of LinkedIn's biggest learnings: you cannot bring a third-party AI tool and have it just work on your stack. Whether it's coding agents (Cursor, Devin, Copilot, Windsurf) or design tools (Figma), you have to build a customization layer and work almost in alpha mode with those companies to adapt their product to your codebase, design system, and unique context.

  • You can't drop a third-party tool onto the LinkedIn stack and have it work — 'it never works'
  • LinkedIn works in near-alpha mode with vendors to adjust their stack to LinkedIn's
  • Applies to coding agents (Copilot, Cursor, Windsurf) and design tools (Figma, Subframe, Magic Patterns)
  • Vendors first need to learn how to work with LinkedIn's design systems and codebase context

“you can't just go and bring a third-party tool and have it work on the LinkedIn stack. In fact, that's one of our biggest learnings,…”

Tomer Cohen · 17:30
#ai-tools#engineering#customization#vendors
Myth Buster49:00

Not Everyone Should Be a Full-Stack Builder

Tomer pushes back on the idea that the whole org should become full-stack builders. Some people see themselves in specialization, and that has a real place and role — there are system builders who empower full-stack builders and specialists too. But he does think companies need far fewer specialized people than in the past.

  • Some people do not want to be full-stack builders, and that's completely okay
  • Specialization has a real place and role in the organization
  • There are 'system builders' who empower full-stack builders, plus specialists
  • But we don't need as many specialized people as we did in the past

“Some people do not want to be full stack builders. And that's completely okay. Some people see themself in specialization, and I think specialization has…”

Tomer Cohen · 49:00

“I don't think we need as many specialized people as we did in the past.”

Tomer Cohen · 49:30
#specialization#org-design#full-stack-builder#talent

Hot Take· 5

Hot Take13:30

Why the Winning Org Looks Like Navy SEALs in Small Pods

Tomer's mental model for the future organization is Navy SEALs: cross-trained across areas, specializing in the mission rather than a function, operating in small nimble pods that can be assembled quickly. LinkedIn now assembles small full-stack pods that tackle a problem for a quarter, then reassemble.

  • Navy SEALs are cross-trained across areas; what they specialize in is the mission
  • Small pods can be assembled and reassembled quickly for velocity and focus
  • LinkedIn moved from large teams to pods that tackle a problem for ~a quarter, then reassemble
  • Result: organizations that are more nimble, adaptive, and resilient, matching the pace of change to the pace of response

“the analogy I have in mind is kind of Navy SEALs, you know, you come to training you they're all kind of learning they're cross-trained…”

Tomer Cohen · 14:00
#organizational-design#teams#pods#agility
Hot Take38:00

AI Makes Your Top Talent Even More Amazing

Answering the long-running question of whether AI lifts weak performers or strong ones, Tomer says at LinkedIn it's the top talent using these tools the most and getting the most value. He attributes it to top talent's innate drive to keep improving their craft and stay at the cutting edge of how they build.

  • The question: does AI make non-amazing people amazing, or amazing people even more amazing?
  • At LinkedIn, top talent are the heaviest users and give the best feedback
  • Top talent has a tendency to continuously try to get better at their craft
  • They're willing to spend the time, and their output quality and speed have jumped

“top talent has this tendency of continuously trying to get better at their craft and this innate need to be at the cutting edge of…”

Tomer Cohen · 38:30
#ai#talent#performance#adoption
Hot Take38:30

Giving People the Tools Isn't Enough — Change Management Is Everything

Tomer's strong take on why AI rollouts fail: shipping agents and expecting adoption doesn't work. Only about 5% of cutting-edge talent adopts new tools on their own; the vast majority needs deliberate change management — incentives, motivation, and visible examples of others succeeding. He compares it to LinkedIn's desktop-to-mobile shift.

  • It's not enough to give people the tools — you must build incentives, motivation, and examples
  • Roughly 5% of talent has a bias for change and adopts new tools eagerly; the majority needs change management
  • 'I see a lot of companies roll out their agents and just expecting companies to adopt. It doesn't work this way.'
  • Culture is by far the biggest and most important thing to get right — and the same playbook worked in the desktop-to-mobile transition

“It's not enough to give them the tools to use it. You have to build incentives programs uh the motivation, the examples to how you…”

Tomer Cohen · 39:00

“I see a lot of companies roll out their agents and just expecting companies to adopt. It doesn't work this way.”

Tomer Cohen · 39:00
#change-management#ai-adoption#culture#leadership
Hot Take45:30

If You're Waiting for a Re-Org to Build Differently, You're Waiting Too Long

Tomer says he cares about mindset, not titles — calling someone a full-stack builder means nothing if their mindset hasn't changed. His biggest message is permission to not wait: don't wait for a formal re-org or declaration; grab an outside tool, bring it in, and prove you're a full-stack builder in mindset before anything else.

  • 'I could care less about your title. I care about how you work.'
  • The goal is a full-stack mindset — believing you can do the whole thing — not a job label
  • If you're waiting for a formal re-org or declaration to build differently, you're waiting too long
  • Permission to go: grab a tool from the outside, bring it in, and show the examples yourself

“if you're looking for a formal re- org or declaration to start building differently, you're waiting too long.”

Tomer Cohen · 46:00

“it's not about I could care less about your title. I care about how you work.”

Tomer Cohen · 45:30
#mindset#full-stack-builder#leadership#initiative
Hot Take1:03:00

Tomer's Motto: Becoming Is Better Than Being

Tomer shares a life motto tied to a growth mindset: 'becoming is better than being.' It maps onto the full-stack builder mode — you're always in progress and iteration mode, in love with the journey rather than reaching a fixed state, measuring yourself by how much you grew year over year.

  • Growth mindset is 'like a second religion' at his home
  • 'Becoming is better than being' — always in progress and iteration mode
  • Fall in love with the journey and process, not reaching a state
  • Measure the delta: how much did version-2026 you grow versus version-2025 you

“there's like a phrase there that is becoming is better than being, which I think ties into the FSB mode a little bit, which is…”

Tomer Cohen · 1:03:30
#mindset#growth#philosophy#motto

Explainer· 3

Explainer05:30

70% of the Skills Your Job Needs Will Change by 2030

Tomer cites LinkedIn's unique data on the world of work: by 2030, the skills required to do your current job will change by 70%. He argues change is now happening faster than organizations can respond, which forces a first-principles rethink of how products get built.

  • By 2030 (4 years out), 70% of the skills needed to do your current job will change
  • The change is uneven — nurses see less impact, some jobs see 90-95% impact
  • 'The time constant of change is far greater than the time constant of response' — change outpaces our ability to respond
  • Whether or not you want to change your job, your job is changing

“When we look at like the skills required to do your job, by 2030, which is literally 4 years from now. Sounds a long time,…”

Tomer Cohen · 06:00

“we're entering this phase where the time constant of change is far greater than the time constant of response. This basically means that change is…”

Tomer Cohen · 05:30
#future-of-work#ai#skills#linkedin
Explainer06:30

70% of the Fastest-Growing Jobs Weren't on the List a Year Ago

Beyond skills changing, Tomer shares that the most in-demand jobs today are growing by more than 70% over last year's fastest-growing jobs, and 70% of today's fastest-growing jobs weren't even on the list a year ago. Many didn't exist a decade or two back.

  • The fastest-growing, most in-demand jobs are growing north of 70% from last year's fastest-growing job
  • 70% of today's fastest-growing jobs were not on the list a year ago
  • Many of these jobs didn't exist a decade or two ago
  • Organizations need a new iteration of skills to thrive

“the fastest-growing jobs right now, the most in-demand jobs in the market, are growing by north of 70% from last year's fastest-growing job.”

Tomer Cohen · 06:30
#future-of-work#jobs#labor-market#linkedin
Explainer07:00

How Simple Work Became a Massively Complex Process

Tomer explains how the basic builder loop (research, spec, design, code, launch, iterate) got bloated at scale. Each step spawned sub-steps — researching a problem became pulling from 10-15 sources; reviews multiplied into design, privacy, and security reviews — and process complexity turned into organizational complexity and micro-specialization.

  • Researching a problem expanded to 10-15 sources of information before it feels done
  • Reviews multiplied: design reviews, privacy reviews, security reviews and more
  • Each sub-step has a valid reason to exist, but together they require multiple teams, codebases, and sprints for a small feature
  • Process complexity became organizational complexity, leading to micro-specialization (interaction, animation, content design, research)
  • 'The work itself is not complex, but the process we made very complex'

“really, the work itself is not complex, but the process we made very complex.”

Tomer Cohen · 08:00

“you're like, "Oh my god, this is why it takes to build a small feature multiple teams, multiple code bases, multiple sprints, just to get…”

Tomer Cohen · 08:00
#product-development#organizational-design#complexity#process

Story· 4

Story19:30

The Trust Agent That Caught Holes Humans Missed on Open to Work

LinkedIn's head of trust built an internal trust agent that reviews a spec and flags vulnerabilities and harm vectors. When they ran the years-old 'Open to Work' spec through it, the agent surfaced not only the trust issues the team had originally identified but also holes they didn't catch until much later — like bad actors targeting vulnerable job seekers with scams.

  • The head of craft for each area builds their own agent; the head of trust built the trust agent
  • Walk a spec through the trust agent and it flags your vulnerabilities and potential harm vectors
  • 'Open to Work' is a great signal but introduces a trust vector: job seekers are more vulnerable to scams
  • Run retroactively, the agent found both the originally-known issues and holes the team missed until later

“we run we walked that spec from a couple of years ago through the trust agent. Not only was it able to find all the…”

Tomer Cohen · 20:30

“the head of craft for every area is building their own agent.”

Tomer Cohen · 20:00
#ai-agents#trust-and-safety#product#case-study
Story26:30

A Maintenance Agent Now Fixes ~50% of Broken Builds

On the code-to-launch side, LinkedIn built a maintenance agent that fixes failed builds automatically — Tomer says it's close to 50% of those builds — plus a QA agent. Instead of engineers hopping on a broken build, they can finish their coffee before dealing with it.

  • LinkedIn splits the lifecycle into 'idea to design' and 'code to launch'
  • The code-to-launch side got the most early attention because coding was already accelerated by IDE tools
  • A maintenance agent handles failed builds — close to 50% of them
  • A QA agent complements it; the idea-to-design side got investment only after this program launched

“In fact, I think we're close to 50% of all those builds being done by the maintenance agent and a QA agent.”

Tomer Cohen · 27:00

“still go and finish your coffee before you have to go and and redo the build again.”

Tomer Cohen · 27:00
#ai-agents#engineering#builds#automation
Story42:30

A User Researcher Used the Tools to Become a Growth PM

To show what full-stack building unlocks, Tomer tells of a user researcher who took an open growth PM role that had gone unfilled. She said 'I feel I can do it,' used all the AI tools, and got the job — a career jump you don't normally see. It also illustrates that it's the easiest time ever to transition between product roles.

  • A growth PM role had been open for a while
  • A user researcher said she could do it, used all the tools, and became a growth PM
  • This is not the usual career path from research to growth PM
  • Highlighting real openings and success stories like this moves people to try it

“This is a user researcher becoming a growth PM, not usually the career path you see. But, she was excited about the area. She used…”

Tomer Cohen · 43:00
#careers#full-stack-builder#case-study#mobility
Story1:04:00

Leaving LinkedIn After 14 Years — and Loving the Product Like a Baby

Tomer reflects on departing LinkedIn after 14 years. He first got hooked hearing Reid Hoffman speak at Stanford in 2008 about professional communities online, joined a few years later, and found the mission aligned with his purpose. He says the best trait for a builder is passion for the product itself — feeling the pain when users complain — likening building to raising a baby.

  • Tomer is leaving LinkedIn after 14 years, having joined before the Microsoft acquisition
  • He first learned about LinkedIn deeply at a 2008 Stanford lecture where Reid Hoffman spoke on professional communities
  • He feels proud and grateful; much of his learning came from tough situations and hard calls
  • A top trait for a builder is deep passion for the product — caring about it, not the job — 'like raising a baby'

“one of the best traits for a builder is to become very passionate with what they're building. Really care. Not not about the job. It's…”

Tomer Cohen · 1:06:00
#linkedin#career#builder-mindset#reflection

Tool· 2

Tool56:30

Tomer's Book Trio: Why Nations Fail, Outlive, The Beginning of Infinity

Tomer recommends a diverse trio of books. Why Nations Fail (Nobel-winning authors) argues nations succeed or fail based on extractive vs inclusive institutions, not culture or resources. Outlive by Peter Attia lays out 'medicine 3.0' and personalized health optimization. The Beginning of Infinity by David Deutsch — a hard but powerful read — argues real progress only comes after clear explanations of cause and effect.

  • Why Nations Fail: success comes down to extractive vs inclusive institutions, not culture/resources/religion; ties to LinkedIn's idea of opportunity
  • Outlive (Peter Attia): 'medicine 3.0' and personalized medicine; categories to optimize health across fitness, diet, and major health factors
  • The Beginning of Infinity (David Deutsch): a hard read, but powerful on cause-and-effect and good explanations enabling breakthroughs
  • Breakthroughs only come once you have a clear understanding of why things happen

“It kind of falls into two camps. Are they extractive or inclusive institutions? Can people participate broadly and opportunities shared or they are institutions that…”

Tomer Cohen · 57:30
#books#recommendations#reading#lightning-round
Tool1:01:00

The Product He Wants: An AI Friend Button on Your Steering Wheel

As a product-thinking wish, Tomer describes wanting a button on his steering wheel that invokes his AI friend to ride shotgun. He's used ChatGPT voice mode for over two years. Lenny notes Teslas already do this — hold the right wheel and Grok appears — and Tomer asks Rivian to catch up.

  • Tomer has used ChatGPT voice mode for well over 2 years, having conversations in the car
  • His wish: a steering-wheel button that invokes an AI friend in the passenger seat to eliminate friction
  • Teslas already do this — holding the right wheel brings up Grok, which connects to your account
  • He'd prefer being able to switch between different AIs rather than being locked to Grok

“I would love to have on my steering wheel a button that invokes my AI friend that can sit next to me in the passenger…”

Tomer Cohen · 1:01:30
#ai#product-ideas#voice#cars

Takeaway· 3

Takeaway12:30

The Five Traits Builders Own — Everything Else Gets Automated

Tomer names where he wants human builders to spend their time: vision, empathy, communication, creativity, and judgment. He rates judgment (taste-making, high-quality decisions in ambiguity) as the most important, and says everything outside these five traits he is working hard to automate.

  • The five human traits: vision, empathy, communication, creativity, judgment
  • Judgment — making high-quality decisions in complex, ambiguous situations — is the most important
  • AI is not yet great at next-level creativity; humans still lead there
  • 'Everything else I'm working really hard to automate. Really really hard.'

“ultimately what I think is the most important trait for a builder is judgment. That's, you know, some people call it test making but it's…”

Tomer Cohen · 13:30
#ai#product-craft#judgment#automation
Takeaway28:00

Feed AI the Right Data, Not All the Data

A major learning: giving an agent access to your entire drive and telling it to reason over everything fails miserably and hallucinates. AI does a poor job weighting importance and tying old specs to actual success. The work is curating golden examples and choosing the specific context and knowledge base — Tomer ties it back to hand-filtering good vs bad posts when rebuilding the LinkedIn feed over a decade ago.

  • 'Just reasoning over your entire knowledge base does not work' — it hallucinates like crazy
  • AI poorly weights the importance of past data and doesn't tie specs to success
  • The real work is curating 'gold examples' / golden examples and picking the right context
  • Most people wrongly focus on connectivity and integration instead of what you feed the model
  • When rebuilding the LinkedIn feed 10+ years ago, Tomer spent weeks hand-filtering good vs bad professional posts

“it's not great to just give it access to your drive and say reason all over this knowledge base. It actually does a very poor…”

Tomer Cohen · 28:30

“the first and most important part was fitting in the right data, not all the data.”

Tomer Cohen · 29:30
#ai#data-curation#knowledge-base#learnings
Takeaway53:30

You Can Build a Startup in a Week — You Can't Transform a Big Org That Fast

Tomer counters the 'a startup built this in a week' hype: yes, starting from scratch with no legacy is genuinely not hard. Transforming a large organization is different — be impatient about the goal and hold high ambition, but be thoughtful and patient about how you bring it to life. Without upfront investment in platform and customized tools, the outcome won't be successful.

  • Startups can build fast because there's no legacy code, structure, or knowledge to work around
  • For a large org: be impatient about the goal, but patient and thoughtful about execution
  • If you don't invest in your platform, a successful transformation isn't realistic
  • If you don't customize the tools, you just get vanilla, generic agents from the outside
  • Expecting 2x productivity in a week is the wrong mental model

“Yes, you can build our startup in a week right now if you start from scratch. It's actually not hard. But when you are trying…”

Tomer Cohen · 54:00

“if you don't invest in customizing the tools for you, then you're just going to get vanilla generic agents from the outside.”

Tomer Cohen · 54:30
#transformation#platform#investment#leadership