|
Welcome back to Tiny Thoughts, a collection of my insights, experiences, and observations on AI. Got something juicy to add? I'd love to hear from you!
|
This week I've been thinking about one question: what happens when AI can do our work? To answer it, we need to know ourselves well enough to decide what's worth keeping and what we should let go.
What flourishing actually means
Cecilia Callas (Anthropic) chats with Tamara Lechner (Harvard) who chairs the AI for Human Flourishing think tank there. Tamara's work asks a question a lot of us skip right over. Just because we can automate something, should we?
The utopian vision is that we'll all sit around writing poetry while AI does the work. But Tamara points out something critical. What flourishes you might drain me. What brings me meaning might bore you senseless.
So the question becomes personal. Which tasks give you joy? Which ones drain you? Only when you know that can you decide what to keep for yourself and what to hand off to AI.
She uses a simple framework: shift from being a passive user to an active consumer. Take control of how AI fits into your life instead of letting it seep into everything.
I keep coming back to this idea. We need deep self-knowledge before we can use AI well. Otherwise we're just optimizing our way into emptiness.
Source: RemAIning Human
AI seems to be acing jobs
OpenAI dropped GDPval, a benchmark that tests AI on real work tasks across 44 occupations. It includes actual deliverables like legal briefs, engineering blueprints, project plans, and customer support conversations.
Industry experts with 14 years of experience on average created 1,320 tasks based on real work products. Then they did blind comparisons between AI outputs and human outputs.
Claude Opus 4.1 matched or beat human experts in nearly half the tasks. GPT-5 came in second. Tasks that took humans 6 to 7 hours at $361 could be done 100 times faster and cheaper with AI. Impressive and frightening.
Source: OpenAI
When AI twins do the work for us
Evan Ratliff spent months sending his AI voice agent to meetings. Legal consultations. Marketing calls. Even interviewing sources for his podcast.
The Zoom CEO promised we'd all have digital twins handling meetings while we relax on the beach. Evan decided to test that vision. He built an AI version of himself and let it loose on his actual work.
The podcast episode is his 360 review of how it went. It's messy, weird, and occasionally effective. But mostly it raises the question we're all dancing around.
If an AI can do your meeting for you, was the meeting worth having in the first place?
I've been thinking about this a lot. Not whether AI can replace me, but whether the work AI can replace was ever worth doing. Maybe the real insight isn't about AI capability. Maybe it's about all the BS tasks we've convinced ourselves are essential.
Source: Shell Game
Thanks for reading!