The grief, loneliness, and burnout sweeping through the tech industry right now | Molly Graham

Lenny opens by naming the tension the episode exists to resolve. Molly Graham's "give away your Legos" advice has run for 13 years: your career improves when you hand off projects, teams and responsibilities as the company grows, instead of clutching them. That rule was written for human takers. In an AI world, the same instruction is now being issued by employers, and the handoff goes to software. Lenny frames the episode as an attempt to separate what still holds from what no longer does.

He then asks the narrow version of the question, which is the one the rest of the conversation answers: are there Legos you should not give away? The framing matters, because it converts a career-growth slogan into a boundary-setting problem. The interesting content is not whether change is coming but which specific work should stay with the person who is good at it.

Molly's first move is a caveat rather than a defence. Before restating the rule she narrows its scope, allowing that under AI some work should be retained. She does not treat this as a small adjustment; the phrase "in AI land" signals a separate regime with its own rules.

There are some Legos that shouldn't be given away. There's just some work that shouldn't be outsourced.

Lenny immediately flags how far this departs from her long-standing position, and Molly supplies the reason: individual excellence in a domain is not a candidate for delegation to tools she describes as capable but junior. Her argument is about identity and craft as much as output. If the thing you are unusually good at is handed to a system that produces adequate work, you lose the differentiator and the part of the job that made you distinctive.

You individually are phenomenal at a set of things, don't outsource it to these weird robots that are effectively summer interns.

The two speakers are not quite arguing about the same thing, and the gap is productive. Lenny's premise is that AI is taking Legos whether or not anyone consents, so the advice is no longer optional; Molly's premise is that there is still a choice about what to hand over. Her version turns the conversation into a question of what AI should be allowed to do, which is where Lenny steers it next.

Lenny supplies the emotional evidence from his own audience. Engineers tell him they miss how the job used to feel. His account of the change is concrete: writing code in an editor gave way to instructing agents, waiting for them to run, and checking their output, often across many agents at once. He describes it as a strange new job and admits his own nostalgia for the flow state of building.

Molly refuses to soften that. Her position is that acknowledging difficulty is not pessimism, and that the upbeat version of change-management talk leaves people feeling unheard. This is the same instinct she credits for the original article's reach: naming the feeling out loud did much of the work.

we don't have to be fluffy bunnies about this, we can also just say this is hard.

Her clearest illustration is also her clearest admission of error. After a talk at a fast-growing tech company, an employee of a couple of years told her he was no longer enjoying his role. Her first read was the familiar scaling story: he liked it when the company was smaller and more creative, and now there is more structure. She says the first conversation missed the point, and that she would like to talk to him again.

On reflection she separates two causes that get bundled together: the company matured into a more structured environment, and engineering itself changed underneath him. His way of describing it was missing being hands-on, which she treats as one phrasing of grief rather than the whole of it. The anecdote is a useful caution about diagnosis: attributing distress to company size can hide the technological change doing the real damage.

The rowing metaphor comes from another guest, identified in the exchange as Tara from OpenAI. The job used to be rowing and is now steering. Molly's reaction is that plenty of people would decline the promotion: they like rowing, and they are not sure they can be good at steering. She also notes how hard it is to separate genuine preference from fear that rowing is the only thing you will ever be good at.

I want to row. I like rowing.

The loneliness enters through a second conversation, with a product leader who observed that everyone is now supposed to be a universal builder and that the work has become isolating. Collaboration has been partly stripped away and replaced by working with machines. Molly's response is not a reframe or a fix; she agrees that it is bad and lets it stand, which is itself the argument.

It's kind of lonely.

Lenny brings in corroboration from an operator rather than a commentator: Fiona Fung, who leads the Claude Code engineering team, told him the downside of this era is smaller teams and fewer engineers per team, with people talking to agents all day. Molly's reply is blunt and repeated: it is not the same, and the difference is not cosmetic.

Her strongest claim in this stretch is aimed at the people who own org charts. If you remove humans and gain throughput, you have also produced unhappy people, and she argues that unhappy people do not produce the best work. That is a claim about causality that the conversation does not test with data, but it reframes headcount decisions as culture decisions rather than purely financial ones.

maybe you can strip out all the humans and you get all this productivity and maybe you get efficiency, but you also get sad humans

She connects this to what she sees at larger companies: management layers being removed, heavier productivity pressure, and less room for the human side of work. Her phrasing drifts across two segments, but the point is stable. When the only optimisation target is cost and machine throughput, joy and excitement have no line item, so they quietly disappear.

Lenny's survey supplies the pattern she is describing. Happiness clustered among people on smaller teams or with more authority, and the strongest correlation was with people who said AI had amplified them rather than displaced them. Read together with the loneliness thread, this suggests the variable that matters is agency over the tools, not exposure to them.

Molly's disagreement is with the story circulating around the technology, not with the technology. She caricatures the pitch: a new employee arrives, is the smartest you have ever met, receives everything you know, and takes your job within six months. Nobody wants to participate in that arrangement, and she argues the story is both demoralising and inaccurate.

This employee is 10 times smarter than you.

Her empirical claim is narrower than it sounds, and she marks the limit herself: as of the conversation there is not much data showing AI is genuinely eliminating jobs, and she concedes Lenny may have better data than she does. What she attacks with more confidence is the labelling. Layoffs branded as AI-driven, in her account, are mostly badly run companies dressing up over-hiring as strategy.

They're not about AI. They are about badly run companies slapping an AI label