About Greg Brockman

Career Journey

Greg Brockman co-founded Stripe, scaling it from startup to global enterprise before departing in 2019 to pursue new ventures, citing a desire to create something novel after reaching learning limits in company building [1]. He transitioned to OpenAI as co-founder and President (previously CTO), focusing on AGI development [22][25][41]. Earlier work includes Stellar board membership for digital currency integration [47] and contributions to collaborative coding environments like Recurse Center [48].

AI Scaling and Compute

Brockman views scaling compute as the primary driver of AI progress, discovered empirically during OpenAI Five's Dota 2 project where doubling compute doubled performance without plateau [30]. OpenAI prioritizes compute scarcity management, with GPUs as a key bottleneck [23], and recent models like GPT-5.4 processing 5 trillion tokens daily [18]. Energy infrastructure now emerges as the next limiter post-algorithmic gains [30]. Inference costs have dropped ~1,000x since GPT-4 [29].

AGI Vision and Strategy

AGI is a continuous process, not a binary event, pursued via large models, iterative deployment, and 'capped-profit' structure for broad benefit [26][31][34][35]. OpenAI shifted from broad research (e.g., robotics) to scalable language models [35], embracing 'agent-first' development [22]. Current level: 'Level 3' (agents) in a 5-level framework toward innovators and organizations [30]. GPT-5 introduces hybrid reasoning architectures with reinforcement learning for reliability [29].

AI Safety and Ethics

Safety involves scalable oversight, iterative deployment for societal feedback, and alignment via positive initial conditions [43]. OpenAI emphasizes beneficial AGI, international cooperation, proactive governance to mitigate risks like job displacement [31][32][37]. Philosophical commitment: AI amplifies humanity, not replaces it [36].

Coding and Developer Tools

Pioneered Codex (GPT descendant for code gen) [40], now evolved into agentic apps with subagents, plugins (Vercel, Slack), and UI surpassing IDEs [3][6][8][12][13][16]. GPT-5.4 Mini excels in coding, multimodal tasks [17]. Use cases span bug detection, proofs, games [11][15][19]. Pricing updates and credits drive adoption [9].

Practical AI Applications

AI empowers startups (doubling revenue) [5], healthcare navigation [4], real estate [20], and non-coders via 'skills' gallery [12]. Excel AI builds games [19]; agents handle complex tasks [16]. Vibecoded companies highlight speed but ethical risks [10].

Software Engineering and ML Transition

Software engineering is AI's new frontier; ML accessible via project-driven self-study despite beginner discomfort [41][42]. Tools like tee-output, doom-py show early infra focus [33][45]. OpenAI Five demonstrated RL generalizability [44].

Early Tech Contributions

Stripe API gists [49][53], CoreOS Kubernetes hacks [46], Stellar blockchain ops [50][52]. Keybase identity proof [51].