Today at a Glance:
Quote: Play the game in front of you
Thinking: The middle game, friction worth keeping, the 2028 scenario
Podcast: Inside Anthropic, AI optimism
Tactics: Friction logs, OpenClaw
Try: The four burner audit
Tweet: UI going to zero, golden handcuffs, AI-free investing
#QUOTE
Quote I’m pondering...
“Everyone has a plan until they get punched in the mouth.” — Mike Tyson
Most people skip the middle game. They vision-cast the end state and never simulate the hundred moves it takes to get there. The opening is exciting, the endgame glamorous. But plans die in the middle.
AI accelerates execution, not strategy. The path from here to there still requires taste, sequence, and willingness to get punched.
#THOUGHTS
What I’m thinking about...
The Middle Game Problem
Will Manidis wrote something provocative: too many tech projects get stuck in the middle game. People jump to the vision without simulating the opening moves required to reach the endgame.
The essay is light on the “so what,” which I think is intentional. Simulation is the point. Play out the moves before you make them. AI makes this easier than ever. You can run scenarios, stress-test assumptions, iterate on plans in minutes. The people who simulate the middle game will outpace those who only dream about the end.
Friction Worth Keeping
Most software is bad. The only way to find the badness is to document every moment of friction while using it. Nothing is too small. The typo. The confusing spacing. The hover that doesn’t reveal what you expect.
But some friction is worth keeping. The struggle to articulate a problem is how you understand it. The tedium of research is how you develop taste. AI removes friction indiscriminately. The winners know which friction to protect.
The 2028 Scenario
Citrini Research published a thought experiment: what if AI bullishness is right, and that’s actually bearish?
The scenario assumes we do not react. AI displaces white-collar workers faster than new jobs emerge. Productivity gains flow to capital, not labor. Consumer spending collapses because machines don’t buy discretionary goods.
I think the trajectory is directionally right, but doomsday is wrong. We will react. The question is how fast and how well. Read it as a simulation, not a prophecy.
#PODCASTS
My favorite listens...
Inside Anthropic: The Creator of Claude Code Boris Cherny built Claude Code. The way Anthropic operates is genuinely different. Agent infrastructure is a people problem. Someone has to onboard, coach, and quality-check agents like junior employees. Most companies will fail at AI because they don’t know how to manage non-human workers.
AI at the Ideal Moment Marc Andreessen’s optimism is infectious. AI productivity gains can translate to prosperity, not displacement. Distribution is the harder problem.
Definitely listen to minutes 22-30, which is a concentrated dose of why AI makes the world better. Worth isolating if you’re feeling nihilistic.
#TACTICS & TOOLS
Tools and techniques I’m using...
Friction Logs The technique that reveals where your product actually breaks. Role-play a user, document every micro-friction, and rank the top 3 problems. Most teams are overwhelmed by implied work; the ranking is the insight. https://blog.sbensu.com/posts/friction-logs/
OpenClaw An open-source framework that turns AI into an always-on assistant. Connect it to your email, calendar, messaging apps, and let it work while you sleep. It’s really high leverage; I’ve basically onboarded an AI employee. I’ve been running mine on a Mac Mini (but any old computer will do)—processing email drafts, researching topics, triggering automations overnight, and creating docs for me. Start supervised, then graduate to autonomous. https://github.com/openclaw/openclaw
#THINGS TO TRY
Experiments worth running...
The Four Burner Audit Family, work, health, friends. You can’t run all four on high. Which are dimmed right now? Name them. Decide if that’s intentional or accidental.
“If You Had 1000 Support Reps” If you could run ChatGPT in your sleep, what would you build? If you had 1000 support reps working 24/7, what would you automate? Your answer reveals the bottleneck.
#TWEET
Social media that made me stop and think...
AI + Work
PMs will operate like product CEOs with an army of AI agents. Taste and judgment are becoming more valuable than knowing how to build. “I’m not technical” is no longer an excuse.
UI usage should trend toward zero. The best interface is none.
Pretty much how I work now. AI as pair programmer, strategist, editor.
Product + Shipping
Tony Fadell’s filter for everything: “Why does it matter?”
Investing
Goldman Sachs launched SPXXAI, an S&P 500 index that excludes AI companies. For people who are anti-AI, now you can put your money where your mouth is. https://www.axios.com/2026/02/20/ai-goldman-sachs-stocks-index



















































