David Brooks (@nytdavidbrooks) in The Atlantic: what will separate people in the AI age is not how smart they are, but their relationship to mental effort. Use AI to skip the thinking and it hollows you out. Make it attack your thinking instead. https://t.co/nkyyULcuQ2
Prototyper, Builder, Sweeper, Grower, Maintainer. @bcherny's five archetypes as functional roles melt. The part everyone will skip: the Sweeper, who unships and simplifies, is in every stage mix. No career ladder rewards that. https://t.co/kLQd6tasId
Bloomberg: founder Matt Van Horn keeps 6+ Claude Code agents running, one babysitting the others while he sleeps. 100x the output and "never worked harder in my life." The agents did not remove the constraint. They moved it to attention. https://t.co/SjCb3Ob5Rl
GraphRAG is the default upgrade right now. I ran a decision gate on my own 200k-chunk notes stack: verdict no. RAG + a reranker beats LLM-extracted graphs on fact retrieval (83% vs 61%). If you write wikilinks, the graph already exists. https://t.co/b4PHRcbpA1
~90 uncoordinated teams inside OpenAI building the same feature, then curating the best. That's @ajambrosino's estimate on @lennysan's podcast. The old process assumed you could only afford to build once. That assumption is out the window. https://t.co/84bNydpT8v
BCG studied 1,488 workers and put numbers on what a lot of us have been feeling. High AI oversight correlated with 14% more mental effort and 19% more information overload. They're calling it 'brain fry.'
The real ROI from AI won't come from a model that's 10x faster or 10x smarter. It will come from humans who learn to use it in ways that change how they think, decide, and execute.
That's an ROI few are considering.
The tool sharpened the toolmaker.
Through teaming with AI, I've picked up thinking habits that show up whether AI is involved or not. I look for lateral reframes. I anticipate blind spots in my own reasoning.
That's not augmentation. That's development.
The US Air Force is exploring "teaming" as an approach to hybrid human-AI workflows. One pilot coordinating multiple AI aircraft that scout ahead and cover blind spots.
It is an interesting model for knowledge work too. Not replacement. Expanded reach and peripheral vision.
Microsoft's own researchers found that AI shifts thinking more than it saves it. Workers reported spending MORE mental effort verifying AI outputs on high-stakes tasks. The labor moved, it didn't shrink.
Used AI to stress-test a go-to-market decision. Ran the problem through multiple analytical frameworks.
The most valuable output wasn't a better answer. It was realizing I was asking the wrong question.
That reframe changed my execution entirely.
Two books that pair well for strategic thinking per @robdwillis: Thinking in Systems (Meadows) teaches you what's happening. Good Strategy Bad Strategy (Rumelt) teaches where to intervene. Together, leverage instead of detail. https://t.co/XqZphOxsZJ
.@Upwork found 39%% of employees spend more time reviewing AI output than they save. Neuroscience explains why: your brain treats agent supervision like air traffic control. Same cognitive drain, same performance drop after 15 minutes. https://t.co/ZuuaLda5tx
YouTube passed Disney as the world's largest media company. $62B revenue, valued at $500B+ standalone. AI touches 95% of user interactions. The real shift: subscription growth is outpacing ad revenue. https://t.co/sLnNJXcCQD
Bainbridge, 1983: "By taking away the easy parts of a task, automation makes the difficult parts more difficult." Everyone saying AI makes work easier should sit with this for a minute. https://t.co/4QFIVQn208
Jeff Kaplan on @lexfridman: WoW's team innovated because they didn't know what they couldn't do. Blizzard later assembled their best for Project Titan and it failed. Expertise became a liability. https://t.co/A2aei5vFNA
Yasar Ahmad's framework for leadership updates: outcome, reasoning, next step. Three sentences. Always end with what you need from them. Leaders don't remember data. They remember clarity. https://t.co/Ao3irwZA19
The AI productivity narrative has a blind spot. A 2008 vigilance study found your brain's monitoring performance drops within 15 minutes. We know this about pilots and air traffic controllers. Nobody's applying it to people supervising AI agents. https://t.co/lZJCFwrFVy