"Finding people who are high agency and work with urgency... if I was hiring five people today, those are some of the top two characteristics. You can take on the world if you have people who have high agency and not needing to get 50 people's different consensus."
High agency: seeing a problem and solving it without permission
Urgency: shipping when it matters, not when it's perfect
These two traits compound at scale — imagine a team of 200+ people like this
Eliminates bureaucratic consensus loops that kill innovation
Real Examples
How Companies Use GPTs to Win
Marketing GPT: Generates ad copy ideas for Facebook & Google, offloads brainstorm work
Data Science GPT: Interprets experiment results, answers follow-up strategy questions without analyst involvement
OKR Planner: Forces consistency in goal-setting and metric alignment across quarters
VC Diligence GPT: Gets multiple perspectives on deal flow analysis in minutes
The patternBest internal GPTs solve horizontal problems (planning, analysis, writing) where you can inject company voice & context, not general assistants.
Where it failsBuilding a general-purpose assistant to compete with ChatGPT requires something radically different. Otherwise you're just adding engineering effort for marginal gains.
The Opportunity Map
Where to Build on OpenAI (And Where NOT To)
Don't: Build another general assistant. ChatGPT owns that. You lose.
New interaction paradigm (sketch → code) beats ChatGPT's text box interface.
Playbook
How OpenAI Ships Fast
No OKRs (counterintuitively) — principles-based decision making instead
Core tenets: Does it help us reach AGI? Is it reliable? Is it what users actually need?
Trust high-agency people to assemble solutions without approval chains
Assistants API example: team saw customer feedback, designed & shipped in weeks
Measure outcomes (adoption, developers, revenue as proxy for compute) not activities
The secret sauceNo institutional legacy = no slowdown. Young companies with high agency win because they don't ask permission first.
Contrarian
AI Product Development Myths
✗We need to outbuild OpenAIINSTEAD →✓ You can't. Build something so different (vertical, interface, integration) that you're not competing directly.
✗OKRs and detailed plans = executionINSTEAD →✓ Principles, trust, and high-agency people move faster than bureaucracy ever will.
✗Revenue is the goal at OpenAIINSTEAD →✓ Revenue funds compute. Compute trains better models. Models reach AGI. That's the chain.
✗Internal tools don't matter muchINSTEAD →✓ Custom GPTs with company context beat generic AI by 10x for knowledge work (30-50% engineering boost documented).