About the Author
Kent McCrea has over 15 years of experience leading one of North America’s top staffing firms, delivering consulting and workforce solutions to multiple Fortune 500 organizations. As AI capabilities become more advanced and widely adopted, he brings a unique perspective on the evolving staffing landscape, with deep insights into emerging market trends and industry shifts.
I previously wrote about developers negotiating token budgets in their compensation and how access to AI is becoming inseparable from a creator’s identity. When the tools are the same but the token access is rationed, everything we thought we knew about evaluating talent, scoping roles, and retaining people gets more complicated.
But the compensation question might be the easy part. I’d been sitting on a piece by Will Manidis called “Tool Shaped Objects” (all links in comments) and shared it with a colleague. It sparked a conversation that made me rethink the assumptions underneath everything in that first post.
Manidis opens with a story about master Japanese woodworkers who spend days setting up hand-forged blades to produce shavings so thin they’re translucent. The shavings are beautiful. They are also, in any practical sense, worthless. A power planer does the same work in a fraction of the time. The blade exists so that the ritual can exist.
He calls this a “tool-shaped object”, something that feels like a tool, produces the unmistakable sensation of work being done, but doesn’t produce work. The activity is the output. And AI, he argues, is the most sophisticated version of this ever created — because it can produce the sensation of anything.
Now layer that onto what’s happening in every company I talk to. Senior leaders are pushing their teams to adopt AI. The urgency is real. I share it. But urgency without direction produces one thing reliably — very expensive piles of wood shavings.
And we’re reinforcing it. The whole token budget conversation carries an assumption underneath it — more is better. But AI researcher Andrej Karpathy (who coined the term vibe coding) just distilled an entire GPT algorithm into 243 lines of pure Python. Less can be more. Measuring AI productivity by token volume is like measuring code quality by line count — the developers reading this are already wincing.
Manidis is careful to say that AI genuinely does real work. The line between tool and tool-shaped object isn’t a line — it’s a gradient. You can only fail to notice when you’ve crossed from one side to the other.
In the AI creator era, the ability to connect output with impact is no longer a best practice. It’s an essential skill. The people who thrive won’t be the ones who produce the most with AI. They’ll be the ones who can tell you what it built, what outcome it drove, and why it mattered.
That’s moving fast from differentiator to baseline expectation.
I strongly recommend reading the Manidis piece. He tapped into something I was feeling but couldn’t quite articulate – it’s worth the read.
And I’m curious: have you had a wood shavings moment with AI? A time you realized, in your own work or a colleague’s, that the output looked impressive, but the impact wasn’t there?
