‘A.I.-Washing’ Layoffs? + Why L.L.M.s Can’t Write Well + Tokenmaxxing - Hard Fork Recap

Podcast: Hard Fork

Published: 2026-03-20

Guests: Jasmine Sun

What Happened

Atlassian and Block have implemented significant layoffs as part of their strategic shift towards AI investments and restructuring. Atlassian's 10% staff reduction is aimed at funding further investments in AI and enterprise sales, while Block's decision to cut 40% of its workforce seeks to create smaller, flatter teams. Meta is also reportedly considering laying off 20% or more of its workforce, adding to the 20,000 employees already let go in the past year, as it plans to spend $135 billion on capital expenditures, including AI infrastructure.

The UK government's proposal to allow AI companies to train on copyrighted works was withdrawn after backlash from artists like Dua Lipa. This decision highlights the ongoing tension between protecting intellectual property rights and advancing AI technologies. Meanwhile, some AI startups are spending more on developing AI tools than on payroll, reflecting the high priority placed on AI development within the industry.

AI models continue to struggle with creative writing, as highlighted by Jasmine Sun's article in The Atlantic. Despite advancements, AI-generated content often lacks the personal voice and life experiences that make human writing compelling. Post-training and reinforcement learning from human feedback have made AI models more predictable but less creative, with companies hiring creative writing experts to evaluate and improve AI-generated content.

Token usage in AI models is being closely monitored, with leaderboards showing which employees use the most tokens. This metric is increasingly being used in performance reviews, raising concerns about potential incentives to maximize token usage unnecessarily. The high cost of tokens, exemplified by an OpenAI employee using 210 billion tokens in a week, underscores the financial implications of this trend.

The concept of 'centaur models' suggests that human-AI collaboration could enhance creative tasks more effectively than AI alone. Jasmine Sun exemplifies this by using AI tools to edit her work, providing personalized feedback based on her writing archive and notes. This approach demonstrates the potential for AI to support, rather than replace, human creativity.

There is speculation about the potential for unionization among tech workers in response to widespread layoffs. With companies like Meta, Atlassian, and Block making significant workforce reductions, there is growing concern about job security and the changing landscape of skills required in the AI-driven economy.

Key Insights