The analyses
6 analysesDeepMind's AlphaFold Breakthrough
How DeepMind solved protein folding and demonstrated AI's potential to transform fundamental scientific research
Using a grand scientific challenge as a proving ground for AI's transformative potential
The strategic move — DeepMind, Google's AI research laboratory, applied deep learning to the protein folding problem through a system called AlphaFold. After a strong showing at the CASP13 competition in 2018, DeepMind developed AlphaFold2 — a transformer-based neural network architecture that achieved accuracy comparable to experimental methods at the CASP14 competition in November 2020. DeepMind then made a pivotal strategic decision: rather than commercializing the technology exclusively, they partnered with the European Bioinformatics Institute (EMBL-EBI) to release predicted structures for nearly every known protein — over 200 million structures — as a free, open-access database. They also open-sourced the AlphaFold2 code.
Read analysis →GitHub Copilot's AI-First Developer Tool
How GitHub leveraged OpenAI to create the first mainstream AI coding assistant and redefined the developer experience
The first AI product to prove that language models could transform knowledge work at scale
The strategic move — GitHub partnered with OpenAI to develop Copilot — an AI pair programmer that integrates directly into the code editor, suggesting code completions in real-time as the developer types. Powered by OpenAI Codex (a GPT model fine-tuned on billions of lines of public code), Copilot could understand context from the current file, adjacent files, and natural language comments to generate relevant code suggestions. GitHub launched Copilot as a technical preview in June 2021 and as a paid product ($10/month for individuals) in June 2022, making it the first commercially successful AI-powered coding tool.
Read analysis →NVIDIA's AI Pivot Strategy
How NVIDIA transformed from a gaming GPU company into the backbone of AI infrastructure and became the world's most valuable company
The two-decade bet on parallel computing that made one company indispensable to the AI revolution
The strategic move — Starting in 2006, NVIDIA launched CUDA (Compute Unified Device Architecture), a software platform that allowed developers to program GPUs for general-purpose computing. Rather than waiting for demand, NVIDIA invested heavily in building the developer ecosystem, providing free tools, university partnerships, and dedicated support for researchers exploring parallel computing applications. When deep learning emerged in 2012, NVIDIA GPUs were already the de facto hardware for neural network training — not by accident, but because a decade of ecosystem investment had made CUDA the standard platform for GPU computing. NVIDIA then systematically expanded from gaming into data centers, AI training, inference, autonomous vehicles, and scientific computing.
Read analysis →OpenAI's ChatGPT Launch Strategy
How OpenAI's consumer product launch changed the AI industry overnight and ignited the modern generative AI race
The research-lab-to-consumer-phenomenon pivot that redefined an industry
The strategic move — On November 30, 2022, OpenAI released ChatGPT as a free research preview — a conversational interface built on top of GPT-3.5, fine-tuned with Reinforcement Learning from Human Feedback (RLHF). Rather than licensing the technology exclusively to enterprises or launching behind an API paywall, OpenAI made a counterintuitive choice: give the product away for free as a simple chat interface anyone could use. This decision prioritized mass adoption and public mindshare over immediate revenue, turning millions of ordinary users into evangelists who demonstrated AI capabilities to their own networks.
Read analysis →Tesla's Autopilot Data Flywheel
How Tesla built the world's largest real-world driving dataset to lead autonomous driving through a vision-only, fleet-learning approach
Turning every customer vehicle into a data collector for the world's largest autonomous driving dataset
The strategic move — Tesla took a radically different approach by equipping every production vehicle with cameras, sensors, and onboard computing capable of collecting driving data. Starting with Autopilot in 2014 and evolving through Full Self-Driving (FSD), Tesla turned its entire customer fleet — millions of vehicles worldwide — into a distributed data collection network. Through "shadow mode," Tesla cars silently ran their neural networks alongside human drivers, comparing the AI's decisions with actual human behavior to identify failure cases. When the AI would have made a different decision than the human, that data was flagged and uploaded for training. This approach generated billions of miles of real-world driving data from diverse conditions no test fleet could replicate.
Read analysis →TikTok's Algorithm-First Strategy
How TikTok's recommendation engine disrupted social media by prioritizing content over connections and redefined attention economics
Replacing the social graph with a content graph as the organizing principle of a media platform
The strategic move — TikTok, launched internationally in 2018 as the global version of ByteDance's Chinese app Douyin, introduced a fundamentally different architecture: the content graph. Instead of showing users content from people they follow, TikTok's For You Page (FYP) used a recommendation algorithm to surface content predicted to be interesting to each individual user, regardless of who created it. A video from an unknown creator with zero followers could reach millions if the algorithm determined it was engaging. This approach inverted the traditional social media model: instead of building a social network and then surfacing content through it, TikTok built a content recommendation engine and let social connections form organically around shared interests.
Read analysis →Explore other themes
Platform & Ecosystem
Strategies for building and leveraging platform-based business models
Explore →Business Model
Innovative approaches to value creation and capture
Explore →Market Entry
Strategies for entering new markets and segments
Explore →Growth & Acquisition
Strategies for scaling through organic and inorganic growth
Explore →Culture & Talent
How organizational culture drives strategic outcomes
Explore →Product Innovation
Strategies behind breakthrough products and services
Explore →Cautionary Tales
Strategic failures and the lessons they teach
Explore →From analysis to the next read.
Every Vault analysis is one read in full — the decision, the rationale beneath it, and how it played out. Follow the thread across the lenses and companies it connects to.