SandeepKhomne.

“Applied for OpenAI Fund with a product to automate instructional design using GPT-3.”

LLM workflow automation—one year before ChatGPT.

“GPT-3 is good enough for cheat-proof tests and auto-grading.”

Anticipating the reset of education frameworks.

“‘Reward is enough’ in the context of AGI and intelligence is a mind-blowing idea.”

Early alignment with reinforcement learning as a driver of intelligent behavior.

“Text, image, video and audio will blow it to insane levels. Imagine an AI-first movie.”

Early identification of the cross-modality shift.

“GPT-3 and DALL·E 2 feel like an iPhone or Internet-scale event.”

AI as the decade’s primary technological driver.

“Building with GPT-3 APIs and Codex without formally learning AI, ML or Python. Endless possibilities.”

Validating coding through natural language before the shift became mainstream.

“The AI breakthrough is already here. GPT-3 could be a Google killer—but Google has data and DeepMind.”

An early view of the tension between innovation and distribution moats.

“ChatGPT has all the signs of becoming a huge platform across fiction, code, travel and search.”

Recognizing the horizontal nature of LLMs immediately after launch.

“OpenAI above $100B, multimodal AI, and new highs for Tesla and SpaceX.”

Aggressive valuation targets rooted in AI-first growth models.

“The nature of software—especially AI—makes it easy for big technology companies to kill startups overnight.”

The thin-wrapper problem and platform risk for venture-backed software.

“Integrating generative AI into two billion daily users will make Google grow exponentially.”

A counter-narrative to the ‘Google is dying’ thesis.

“Autonomous robots are a wild, high-impact field. Most people are missing it like they missed GenAI.”

Real-world robotics as the next compute frontier.

“AI is the tool today. Future versions will achieve what existing tools do—and replace them.”

The thesis of AI cannibalizing software vertical by vertical.

“AI is both hardware and software. Content will 1000×, driving exponential demand for compute.”

The content-explosion thesis for GPUs, servers and cloud.

“The safest AI exposure starts with hardware, then model leaders; pure software is trickier.”

A strategic asset-allocation framework for the AI decade.

“Founder-led technology companies win through technical depth, obsessive control and contrarian patience.”

A synthesis of operating principles from years of studying Elon Musk and Jensen Huang.