Molly Graham and Lenny Rachitsky discuss how the famous career advice "give away your Legos" should be reinterpreted now that AI is rapidly changing how we work and our professional identities. The core is to grieve change and loss honestly while treating AI not as an unconditional replacement but as a capable yet immature intern who must be managed and supervised. The conversation's conclusion is that you can hand repetitive work to AI, but you must not hand over vision, judgment, responsibility, trust, and your sense of what is good.

1. Where the "Give Away Your Legos" Advice Began

Lenny opens the conversation by noting that we've entered an era where AI is taking people's work and everyone feels pressure to hand their "Legos" to AI. Molly's longstanding advice originally meant that when a company grows quickly, you should hand the projects, teams, and areas of work you built over to others and go find new opportunities. But now that AI has arrived, there are parts of this advice that are hard to apply as is.

"We've hired a new employee. The smartest employee you've ever met. Teach them everything you know. Then, within six months, they'll take your job. Who would want to do that?"

Molly developed this concept from firsthand experience of hypergrowth at Google and Facebook. When she joined Google in 2007, the company as a whole had about 10,000 people, but the organization she belonged to grew from 25 to 125 people in nine months. She later joined Facebook when it had around 500 employees and 80 million users, and when she left five years later, it had 5,500 employees and more than 1 billion users.

Along the way, she saw how tightly people cling to the work and identity they've built. At first, someone builds a blog, a product feature, or part of a system, and gets good at it. Then that work becomes more than just a job—it shapes "who I am." But as the company grows and they're asked, "Now hand this off to someone else and take on the next thing," they naturally resist.

"I'm good at this and I know how to do it, so why should I hand it over? What if this is the only thing I enjoy?"

Molly compares this to kindergarteners and a pile of Legos. The kids may be confused at first, but they have fun each building their own tower. But when someone comes over and tries to take the tower they built, they instinctively push them away and say, "Go away!" Yet having more people to play with doesn't only mean losing your existing tower—it also means you get the chance to build something new and more interesting.

"Giving away your Legos doesn't mean telling someone to build an exact copy of your tower. It means truly handing the future of that tower over to them."

At first she thought this advice only applied to explosively growing startups like Facebook, but after the piece was published it resonated with all kinds of people around the world. From employees at large retail companies to a Nigerian founder who grew their team from two people to four, many people adopted the metaphor as language for describing changes in their own lives. Molly realized it was ultimately about a bigger theme than growth itself: the emotions humans feel when going through change.

2. Principles That Still Hold in the AI Era: Change Is Scary, and Learning Is Survival

Molly says much of the core of her pre-AI Lego advice is still valid. Today almost every company is going through enormous change, whether or not it's growing. So the essence of the Legos idea is not "how to survive at a fast-growing company" but how to be your best self amid change.

Here is how she used to summarize the Legos piece in two sentences:

"In times of rapid growth and change, your most important job is to make yourself unnecessary. That way you're ready to take on whatever comes next."

"Don't worry. It'll be okay in the end."

Lenny points out, however, that in the AI era even the second sentence doesn't sound as simple as it used to, because people have started asking, "Will it really be okay?" Even so, Molly stresses that acknowledging that change itself is scary and hard is the starting point. Feeling anxiety, anger, loss, and exhaustion in the face of change is not a sign that you're broken—it's a normal human reaction.

"Feeling these emotions doesn't mean you're crazy or broken. It just means a lot of change is happening."

In particular, she sees grief and burnout spreading widely across the tech industry right now. The excitement and opportunity AI opens up are real, but at the same time people are losing the ways of working they loved, familiar team structures, and pride in their expertise. If leaders ignore these feelings and say, "The future will be amazing. Just adapt," team members end up feeling even more isolated.

Another principle Molly repeatedly emphasizes is this:

"The future will belong to those who learn, more than to those who know."

In an AI environment, even knowledge you were certain of just six months ago can quickly become outdated. So what matters more than what you know now is what you can learn tomorrow. A company's growth curve is also a graph showing how fast individual roles and capabilities need to change.

She says staying put without changing can feel safe, but it's actually the riskiest choice. If you cling to old ways while the company and environment change quickly, you will inevitably fall behind.

"Standing still looks safe, but amid change it's the least safe thing you can do."

3. Engineers' Sense of Loss and the Breakdown of Team Structures

Lenny says engineering roles in particular have gone through the most drastic change over the past few years. An engineer's job used to be writing code directly, solving problems, and immersing yourself for long stretches to build something. Now, more and more of the job is instructing AI to write code, reviewing what multiple agents produce, fixing errors, and instructing them again.

Lenny describes this as a shift from "rowing the boat" to "steering it." But Molly responds that not everyone wants to steer.

"I think a lot of people right now are saying, 'I don't want to steer. I liked rowing.'"

What's at stake here is not just a matter of adapting to technology but a loss of professional identity. The immersion and joy people felt in the process of building things directly disappear, and they grow anxious about whether they'll be good at the new role. Loving "rowing" may be mixed with the fear that "it might be the only thing I'm good at."

Molly recalls a conversation with an employee at a tech company. The employee said that as the company grew and became more structured, the work wasn't as enjoyable as it used to be. At first she thought the problem was the loss of the freedom and creativity unique to small companies, but looking deeper, she found the employee was grieving over not knowing what the job itself was turning into.

"Change is hard. It can turn out to be wonderful, but we don't have to force ourselves to be positive about it. It's okay to just say it's hard."

AI changes not only individual work but also team structure. Where teams of five to ten engineers working together used to be typical, more teams now combine fewer people with more AI tools and agents. People end up talking to AI all day instead of collaborating with colleagues, and that can deepen a sense of isolation.

Molly warns that it's dangerous for companies to remove humans from their org structures based on productivity alone. Cost efficiency and automation may go up, but if people's satisfaction and the joy of collaboration disappear, it's hard to sustain strong performance in the long run.

"You might be able to achieve higher efficiency by taking humans out entirely. But if people end up unhappy as a result, that won't produce the best outcome for anyone."

4. AI Fear Narratives and Burnout: It's Not That Everyone Is Doomed, Nor That Everyone Is Happy

Both of them think the fear narrative that AI is wiping out jobs is overblown in some ways. Molly criticizes the fact that many of the mass layoffs blamed on AI are actually closer to over-hiring, management failures, or cost-cutting decisions, yet companies package them as AI's fault.

"A lot of what's being called 'layoffs because of AI' is honestly bullshit. Often, instead of admitting they hired too many people, companies slap the AI label on it."

She says that if people keep hearing that AI is like "a new hire ten times smarter than you," and that after teaching it everything they'll soon be replaced, it's hard for them to willingly take part in the change. If you ask people to "make yourself unnecessary" and then tell them there won't be a job afterward, naturally no one will want to give away their Legos.

Molly shares a perspective she gained from a conversation with journalist Manoush Zomorodi. The common question is "What should I do if my job disappears?" but a more useful question is this:

"What would I do if I believed my job will keep existing, but will turn into something completely different every six years?"

This perspective is neither optimistic that jobs will stay exactly the same nor pessimistic that they'll all disappear. Instead, it starts from the premise that the shape of jobs will keep being reinvented and proposes actively taking part in that redesign.

In a survey of tech workers Lenny ran, burnout responses jumped from 44% in 2025 to 55% in 2026. Behind this is the pressure to do more work faster because AI tools now exist, the anxiety that the company could be at risk if you fall behind other teams, and the reality that expectations rise while compensation stays the same.

"Now there's this vibe of 'You have AI tools, so why can't you do more?' People are being asked to do more work, and they're not necessarily getting paid more."

Molly says the pace of change itself wears people out. Just six months ago, organizations were asking, "How do we get everyone to use AI more?" Now the question has become "Are we using AI well?" and "Are we using it badly?" When the standards flip like this every few months, people burn out from constantly adapting.

But the survey also showed an important countertrend. About half of respondents said they were so excited and happy that they felt this is the best time of their careers. This is especially true for people who work on small teams, have a lot of autonomy and authority, and feel that AI amplifies their abilities.

"About half of people feel this is the happiest time of their careers. They're working harder, but they're also having the most fun they've ever had."

By contrast, designers reported relatively high dissatisfaction. As everyone became able to create design mockups with AI, design experts' judgment was easily encroached upon, and while product launches sped up, the time to think things through, build consensus, and refine shrank.

5. AI Slop and the Productivity Illusion: AI Is Not a Genius but an Intern to Be Managed

Lenny says, "More and more, other people are cleaning up the AI output that people created to do their own jobs." In other words, so-called AI slop—unreviewed AI-generated content and low-quality output—is piling up inside organizations.

Molly argues that we need to start by fixing the attitude of seeing AI as a "superintelligent, perfect employee." Today's AI tends to produce wrong or superficial results without enough context, and it needs human review and repeated revision. So she describes AI more accurately like this:

"AI is an intern. Often a terrible intern. I call it the lazy intern."

No one would send a presentation an intern made straight to an executive without reviewing it. You check the content, fill in missing context, revise it several times, and only then take responsibility for passing it on. AI output should be treated exactly the same way.

"You wouldn't copy what an intern made and send it straight to your boss. AI is the same."

The problem is that the moment you copy and send AI output as is, the person who should be responsible for its quality dodges that responsibility. The burden then shifts to whoever receives it. Someone has to reread, verify, and fix the sloppy document, strategy, code, or design. This process breeds fatigue and distrust across the whole organization.

Molly is especially concerned about CEOs sending AI-written strategy memos to their organizations with barely any edits. This isn't just about the quality of one document. It signals to the organization that "it's fine to outsource even thinking and strategy" and "you don't have to take responsibility for the quality of what you send."

She also says many companies are over-tracking metrics like token usage, lines of code, and number of AI uses. This is no different from the old days of counting what time people arrived in the parking lot, how many Slack messages they sent, or how many lines of code they wrote.

"We can end up running at top speed on a treadmill and still not getting anywhere."

The key distinction here is between productivity and effectiveness. Productivity is the ability to produce more output; effectiveness is the ability to actually produce better results. AI can explosively increase output, but if that output increases rework, security incidents, and review costs, you can't call it effective.

In fact, in an engineering example mentioned in the conversation, development productivity rose with AI, but the amount of code that had to be rewritten increased eightfold and security issues multiplied. In other words, whether AI adoption succeeds depends not on "how much you produced" but on how good the results were and how responsibly you produced them.

6. Giving Legos to Humans Is Different from Handing Work to AI

The biggest reason Molly revised her Lego advice for the AI era is that handing work to a person and handing work to AI are fundamentally different. When you fully hand work to a person, you can hand over ownership and responsibility for the outcome along with it. That actually frees up space in your head so you can take on new opportunities.

She describes the best way to give your Legos to a human with an exaggerated metaphor:

"The best way to give away your Legos is to throw them in the person's face and run in the other direction."

Of course, she doesn't literally mean handing them off irresponsibly. The point is not to try to control the other person into building the same tower as you, but to entrust the future of that work to them. If you keep interfering, approving, and revising, it's not real delegation.

But when you hand work to AI, you can't walk away like that. Even if AI drafts, writes code, or does analysis, a human still has to review and revise the output and bear final responsibility. In other words, AI can do the work for you, but the burden of supervision and responsibility doesn't go away.

"Handing work to a robot is not the same as giving your Legos to a person and walking away completely. You still supervise, own the final result, and take responsibility."

Because of this, AI can reduce workload while simultaneously creating a new kind of mental burden. When multiple AI agents work at the same time, the user has to keep receiving notifications, answering questions, and unblocking them. Lenny also says running multiple tools and agents is exciting, but it means handling requests all day like "How's this?", "I'm stuck here," and "What should I do next?"

Molly likens this to a manager's role. The skills needed to use AI well are providing context, setting clear expectations, correcting output, giving feedback, and establishing standards. In other words, even people with no management experience become managers the moment they use AI.

"Now everyone in the world has effectively become a manager."

Right now, AI is especially like an employee even more junior than a junior employee. With an excellent senior engineer you can give just a rough direction and trust the result, but AI requires much tighter instructions and repeated review. So using AI may not be an easy or enjoyable change for everyone.

7. From Fear to Opportunity: Designing the Job of the Future Rather Than Protecting Your Job

The original Lego advice rested on the premise that "when you hand something off, a new opportunity appears." The reason people are anxious in the AI era is precisely that this second premise is shaky. They're not sure there will really be an opportunity left for them after handing their work to AI.

While Molly understands this anxiety, she says we shouldn't get trapped only in today's fear narrative. She thinks AI is likely to bring about a massive shift like the mobile revolution and create many new industries and jobs.

"If AI automates the work I hate and lets me spend more time on the work I love, that's the future we want."

She thinks that the more expertise-heavy a profession is—law, journalism, design, product management, engineering—the stronger the mindset can become that "I have to hide and protect my expertise so I won't be replaced." But rather than clinging to past definitions of a role, a better strategy is to design what that profession will look like in the future.

"Lawyers will keep existing in the future. They just might look completely different every three or six years. Do you want to take part in designing that future, or stay focused on protecting what you know now?"

This demands agency and an entrepreneurial attitude from individuals. Instead of simply being dragged along by change, you should ask, "What's the next version of my job?" and "Can I be one of the people who builds that version?"

At the organizational level, lowering the barriers between roles matters. Molly cites a case where a head of design wanted to build products directly and even ship code, but the engineering organization blocked it, saying "Designers shouldn't deploy." Of course, in fields like finance where safety and regulation matter, controls are needed. But we should re-examine whether past role boundaries are truly still necessary, or whether they remain simply because they're familiar.

"Some of the walls between designers, engineers, and product managers need to come down. It's already happening."

This shift doesn't mean abandoning your profession and becoming a completely different person. Rather, it means applying your expertise in broader ways. A designer can become not just someone who designs but someone who builds and validates products more deeply; a marketer can even ship code; and a journalist can become not just someone who works inside a big media company but someone who builds their own distribution and business model.

8. Legos You Should Never Give Away: Vision, Judgment, Trust, and Responsibility

Lenny finally raises the most important question: "Are there Legos we shouldn't give to AI?" This is where things differ most from Molly's past advice. At fast-growing companies in the past, "give it away" was generally the right answer. But in the AI era, there is clearly work you shouldn't give away.

Molly hopes AI will reduce the work humans weren't good at to begin with or don't need to do. For example, driving is a domain where rules, systems, and repetitive judgment matter, so self-driving technology may be safer than humans. On the other hand, the domains humans are uniquely good at should be protected.

The main things she identifies as belonging to humans are:

• Judgment and strategy: deciding what's right, what to prioritize, and which direction to go
• Vision and taste: defining what you want the world to look like and what a good outcome is
• Trust and relationships: work that requires accountability, commitment, empathy, and relational context between people
• Quality standards: knowing and evaluating what an excellent result is
• Final responsibility: owning and fixing things when results fail or cause harm

"If you don't know what good looks like, you can't hand it to an intern."

She especially stresses that CEOs should not have AI generate strategy and then copy and send it straight to the organization. Strategy isn't just a document-writing task; it's a human responsibility that carries the organization's direction and judgment.

Lenny sums this up with the frame of a "human sandwich." At the top is a human. The human decides the goal, the vision, the problem definition, and what the desired world looks like. In the middle is AI. AI produces multiple drafts, analyzes data, and performs repetitive tasks. At the end there's a human again. The human reviews the results, revises them, and guarantees quality and accountability.

"At the top, a human says, 'This is where we want to go'; in the middle, AI does the work; and at the end, a human reviews it. It's a human sandwich."

Molly agrees with the metaphor and says that no matter how good AI gets, for now the right attitude is to see it as "a pack of summer interns." You shouldn't outsource all of your strengths, sensibility, insight, and professional judgment to robots.

"You're really great at certain things. Don't hand those strengths over to these weird robots. In the end, they're summer interns."

Lenny uses the difference between Airbnb and Booking.com to illustrate the importance of human vision. Booking.com is designed to maximize booking conversion, inducing pressure and urgency. Airbnb, by contrast, chose the direction of creating a more delightful and comfortable experience. AI can optimize for whichever is more efficient, but choosing what kind of world and experience we want is up to humans.

"Humans are the ones who steer toward the world we want."

9. A Funeral for What's Lost, and the Hope of a Slow Takeoff

The two say that adapting to the changes AI brings and grieving what's disappearing are not contradictory. If the joy engineers felt in coding directly, the way designers had their expertise recognized, and the experience of solving problems together on big teams are fading, it's natural to feel you've lost something.

Molly says that sometimes you may even need to actually hold a "funeral" for that loss.

"Sometimes you need a funeral for certain things. Miss them for a while, and then ask what might be possible next."

"You can grieve deeply for what you've lost, and at the same time there may be opportunities in the future. We need to be able to hold both of those things together."

If people in the AI era are too quick to say "Let's just stay positive," those experiencing loss end up hiding their feelings. Molly believes that being able to say "I miss it" about the beautiful ways we used to work is what allows us to move on to the next stage in a healthy way.

Meanwhile, Lenny says AI's progress looks closer to a slow takeoff, where people have time to observe, learn from, and respond to change, than to a fast takeoff that explodes into uncontrollable superintelligence right away. Even when AI agents behave unexpectedly, people are spotting the problems, building countermeasures, and adapting.

Molly particularly rejects the "it's already too late" mood. Talk that you'll fall behind unless you automate everything and run dozens of AI agents only fuels anxiety, but she thinks we're actually just at the start of the first game, the first inning.

"It's the first inning. The distance between beginners and experts is very short right now."

AI tools can actually give newcomers and beginners a big opportunity, because the infrastructure knowledge and tool experience that used to take years to accumulate to become skilled are shrinking. New people who make good use of today's tools can quickly produce work and build their skills.

Lenny suggests the most important habit right now is asking this question every time:

"Could AI help me do this?"

This doesn't mean handing everything to AI unconditionally. It's a new thinking habit of pausing before starting a task to consider whether AI could reduce repetitive work, expand your thinking, or suggest a better approach.

At the same time, Lenny says we should ask a bigger question as well.

After "Could AI help with this?", the next question to ask is "How can I be more ambitious with this idea?"

Molly adds, however, that ambition shouldn't be mistaken for simply shipping more things faster. In the AI era, countless products and companies can be built quickly and then vanish within six or twelve months. So what matters isn't "can it be built" but is it good, is it worth lasting, and is it worth other people's time.

"Being able to build something and that thing being good are two different matters."

10. For Managers and Leaders: Managing Humans Becomes More Important

Molly says this is an especially hard time for managers and leaders. Individual contributors can use AI to build more, but leaders have to decide how to structure their teams, how far to use AI, how to handle team members' anxiety, and what humans must be responsible for.

Her first message to leaders is don't forget that you are the role model. How a leader uses AI, how much responsibility they take for AI output, and what language they use to talk about change become the team's behavioral standard.

"What you do and how you act show your team 'what good behavior looks like.'"

Second, leaders need to acknowledge the impact change has on team members' minds and identities. Just telling people "You're not the only one feeling this" and "It's hard because change is happening fast" reduces their sense of isolation.

"Other people have felt this way before. It doesn't mean the world is ending; it means too much change is happening too quickly."

Third, leaders need to clearly set standards for responsibility and quality in the AI era. Even if AI took part in producing the output, humans must be responsible for the quality of the documents, code, strategies, and decisions sent to the organization. Having used AI doesn't exempt anyone from responsibility.

"Putting AI out into the world and being responsible for the quality of the result are not separate things. However you made it, you have to own the output."

In Lenny's survey, the factor that could most affect employees' happiness right now turned out to be their direct manager. Individuals can hardly control the pace of AI progress or overall company strategy, but a good manager can give team members a sense of being respected and supported.

That's why Molly worries about the trend of large companies cutting management layers and eliminating managers. She understands the pressure of cost and efficiency, but she believes management has become not less important in the AI era but even more important.

"Good managers make people feel valued, supported, and able to solve problems together."

She says truly good managers are so rare that you might meet only one or two in your life, and advises that if you find one, don't let them go.

"If you find a truly great manager, hold on to them with everything you've got. You might only meet one or two in your whole life."

11. Closing: Adapting Humanely in the Middle of History

At the end of the conversation, both agree that this is a confusing but special historical moment. Molly compares being inside a fast-growing company to being "in the middle of a hurricane." Everything is flying around, it's easy to get swept up in small problems, and it's easy to forget why you're there.

So sometimes you need to step back and look at the bigger story. What you're doing now isn't just about your current title or a particular project; it could be part of a story where someday you can say, "I was there for that change, and I took part in building something."

"You are living through history right now. Five or ten years from now, what story will we tell about this period?"

Finally, Molly urges people not to stay trapped alone in the anxiety of the AI era. Whether leaders or practitioners, they should find colleagues and communities, compare the realities each is seeing, and talk together about what is hype and what is real opportunity.

"Don't isolate yourself in grief. Get out there and talk about what's real and what isn't."

Lenny calls this a "Trojan horse episode": people come looking for practical answers like how to use AI, and end up discovering emotional and human needs they didn't know they had. In the end, the most important message of this conversation is simple. We have to learn AI, but in the process we must not give up human judgment and responsibility, relationships and joy, or even the right to grieve what we've lost.