The grief, loneliness, and burnout sweeping through the tech industry right now | from Lenny's Podcast

What It Covers

Molly Graham, host of WorkLife and creator of the 13-year-old "Give Away Your Legos" career framework, revisits her advice in the context of AI transformation. She examines how delegating to AI differs fundamentally from delegating to humans, why 55% of tech workers report burnout in 2025, and which professional skills should never be outsourced to machines.

Key Questions Answered

• •AI Delegation vs. Human Delegation: Handing work to AI is not the same as giving a Lego to a human colleague. With AI, the mental ownership never transfers — you retain full oversight, quality control, and accountability for the output. This means the cognitive load stays with you even as the task moves to a machine, which helps explain why burnout has risen 10 percentage points in a single year despite productivity gains from AI tools.
• •The Intern Mental Model: Treating AI as a superintelligent replacement leads to "AI slop" — low-quality output copy-pasted without review. Reframing AI as a junior intern changes behavior: you would never forward an intern's unreviewed presentation to your CEO. Apply the same standard to AI output. Provide context, iterate on drafts, and maintain editorial accountability. The quality gap between raw AI output and genuinely good work remains large.
• •Legos to Keep, Not Give Away: Three categories of work should not be delegated to AI right now: tasks requiring judgment and strategic vision (e.g., company strategy memos), work that builds trust within relationships, and anything where the person doing the delegating cannot yet define what "good" looks like. If you cannot evaluate quality, you cannot effectively supervise an AI producing it — the oversight becomes meaningless.
• •The Manouche Framework for Job Reinvention: Instead of asking "will my job disappear," ask "what would I do if my job will always exist but look completely different every six years?" Journalist Manu Somerodi built a 30-year career across BBC, crypto audio, and Ted Radio Hour by treating each disruption as a design opportunity rather than a threat. Applied to tech: engineering, design, and product management will persist in transformed forms — the question is who shapes those forms.
• •Burnout Driven by Narrative Thrash, Not Just Workload: The 55% burnout rate in 2025 tech (up from 44% in 2024) stems partly from rapid narrative reversals. One example: OpenAI shifted from "maximize AI token usage everywhere" to "are we using AI cost-effectively" within six months. Workers absorbing these contradictory directives every few months experience exhaustion independent of actual task volume. Leaders who acknowledge this thrash explicitly reduce team stress without changing any workflow.
• •Happiness Concentrates in Specific Conditions: Survey data from Lenny Rachitsky's 2025 tech poll shows the strongest correlation with workplace happiness is feeling "amplified by AI" — using it to do more of what the person values. Secondary factors include small team size and high individual autonomy. Designers report the lowest happiness, partly because AI made "acceptable design" widely accessible, eroding perceived expertise, while design's feedback-dependent process cannot compress the way coding workflows can.
• •Management Matters More, Not Less: The single most actionable lever for increasing team happiness, per the 2025 survey, is manager quality — stronger than any AI tooling or compensation variable. Simultaneously, many large companies are eliminating management layers for efficiency. Graham argues this is a strategic error: effective managers make people feel seen during periods of disorienting change, and that function becomes more valuable, not less, as transformation accelerates. Finding and retaining one or two genuinely skilled managers across a career is rare and worth prioritizing.