Latest Interviews

Building for the Enterprise with Aaron Levie (How to Start a Startup 2014: Lecture 12)

Aaron Levie

Box CEO and co-founder Aaron Levy advocates for a strategic pivot toward enterprise software, highlighting the $3.7 trillion global market and the sector's shift from on-premise systems to cloud-based platforms. Levy details specific tactical approaches for building scalable ventures, including targeting niche "wedge" markets, exploiting economic asymmetries, and maintaining consumer-grade user experiences to bypass traditional sales friction. He concludes by predicting that deep technology partnerships will become essential for all businesses over the next decade, urging startups to balance self-service adoption with consultative sales models to navigate complex enterprise environments.

Hiring and Culture with Patrick and John Collison and Ben Silbermann (HtSaS 2014: 11)

John Collison, Ben Silbermann, Patrick Collison, Sam

Stripe and Pinterest founders articulate a culture-building strategy rooted in the strategic hiring of the first ten employees, who serve as multiplicative agents for future growth while embodying specific traits of intellectual honesty and obsessive detail. To sustain this integrity as organizations scale, leaders transition from rigid architectural planning to adaptive "gardening" practices that prioritize radical transparency, autonomous units, and structured onboarding to manage complexity. This approach relies on selling the hard challenges of the mission rather than guaranteed success, ensuring that only those with genuine conviction join teams that operate with high self-awareness and rapid feedback loops.

How to Get Started, Doing Things that Don't Scale, and Press (How to Start a Startup 2014: 8)

Stanley Tang, Walker Williams, Justin Kan

DoorDash founder Stanley Shao, Teespring CEO Walker Williams, and Twitch founder Justin Hunt share specific strategies for early-stage growth, emphasizing the necessity of manual, non-scalable operations like founder-led customer service and direct sales to validate market demand. The discussion highlights how successful startups often prioritize rapid iteration and targeted press outreach over perfecting infrastructure or relying on expensive agencies during their initial phases. Key takeaways include avoiding false validation metrics, leveraging mobile technology to minimize capital expenditure, and treating media coverage as a calculated tool for acquiring the first thousand users rather than chasing broad recognition.

Growth with Alex Schultz (How to Start a Startup 2014: Lecture 6)

Alex Schultz

This presentation outlines a growth philosophy where the entire organization, led directly by the CEO, must function as a unified team to optimize a single North Star metric rather than relying on isolated departments. Drawing on case studies from Facebook, eBay, and Airbnb, the speaker details how companies can accelerate retention and achieve viral expansion by identifying their unique "magic moment," applying dimensional reasoning to market saturation, and executing high-volume experiments. Ultimately, the discussion emphasizes that sustainable scaling requires prioritizing long-term user retention over acquisition volume and utilizing precise data modeling to predict product-market fit within the first few months of operation.

The deep ocean is the final frontier on planet Earth

Bobby Moore, Greg stone, Jim Delgado

Utilizing the US government vessel *Okeanos Explorer* and its advanced *Deep Discover* ROV, an international team is investigating the World War I-era submarine wreck of the S-19 while contending with extreme pressures and critical technical failures. Although scientists hope to uncover clues about the origins of life in deep-sea ecosystems and assess the economic potential of rare earth mineral mining, they simultaneously document pervasive marine debris to underscore the urgency of stewarding these fragile environments. This mission highlights the accelerating race to explore the uncharted depths, balancing historical discovery and resource extraction against the need to protect the ocean's closed-system integrity.

Before the Startup with Paul Graham (How to Start a Startup 2014: Lecture 3)

Paul Graham

Y Combinator partners deliver counterintuitive advice to aspiring entrepreneurs, emphasizing that startup success requires suppressing corporate instincts, prioritizing deep user expertise over formal business training, and waiting until after college to launch ventures. The event warns founders against mimicking established company structures or relying on academic performance to predict resilience, instead urging them to pursue personal intellectual curiosity and immerse themselves at the technological edge. By clarifying that organic growth and trust in interpersonal judgment outweigh efficiency systems or business school credentials, the discussion outlines a framework where total life commitment and genuine user value become the primary determinants of long-term viability.

Team and Execution with Sam Altman (How to Start a Startup 2014: Lecture 2)

Sam Altman

This comprehensive guide outlines critical operational frameworks for early-stage startups, emphasizing that co-founder selection based on personal history and shared traits is the primary determinant of survival. It further details rigorous hiring protocols that prioritize aptitude and mission belief over experience, alongside a strict management philosophy that links rapid execution to a bias toward action and extreme focus. The advice concludes by stressing the necessity of maintaining momentum through tangible wins, swift termination of toxic elements, and the absolute rejection of remote co-founder teams to ensure exponential growth.

Arjun Guha: On Verification for System Configurations Languages

Arjun Guha

Researchers developed "Rehearsal," a verification tool designed to detect determinism and idempotency violations in Puppet configuration manifests, which are frequent causes of major system outages. The system employs a transformation pipeline that converts complex Puppet syntax into Datalog and a low-level modeling language, leveraging the Z3 SMT solver alongside optimizations like partial order reduction and state pruning to analyze real-world code efficiently. Benchmarks on GitHub-scraped manifests demonstrate that this approach successfully identifies previously unknown non-deterministic bugs and generates fixes, although current limitations exclude shell script execution and specific file permission modeling.

How to Build an Exchange

Claudio, Ron, Pete

A US equities matching engine utilizes a deterministic, single-instance architecture to process three million messages per second while ensuring fair, latency-critical order execution across thousands of symbols. The system employs C++ and OCaml on a dedicated machine to replace complex distributed consensus with high-speed, single-threaded logic, allowing state machines to recover from failures within 30 to 60 seconds via message log replay. This design guarantees regulatory reproducibility and financial reliability by eliminating consensus overhead, enabling clients to achieve precise atomicity and global risk enforcement through simplified, high-throughput message sequencing.

Why Functional Programming Doesn't Matter

John Hughes, Tony Hoare

Jane Street, a high-frequency proprietary trading firm executing millions of daily trades, attributes its engineering success to OCaml's expressive static types rather than traditional functional features like laziness or strict purity. This type system enforces correctness by eliminating null pointer exceptions, forcing logic updates at compile-time to prevent bugs, and encoding business invariants to exclude impossible states. Consequently, the firm prioritizes predictable performance and verifiable clarity to protect its capital against errors, viewing advanced type systems as a more critical asset than higher-order functions or declarative purity.

MIT 6.S094: Deep Learning for Human-Centered Semi-Autonomous Vehicles

Researchers are collecting billions of high-speed video frames from semi-autonomous Teslas to train deep learning models that detect critical driver metrics such as body pose, gaze direction, and cognitive load. By shifting from fully supervised to semi-supervised annotation strategies, the team achieves an 84-fold reduction in human effort while accurately identifying micro-saccades and emotional cues to overcome current privacy and trust barriers. This data-driven approach aims to replace static crash test assumptions with dynamic occupant monitoring, ultimately enabling vehicles to adapt passive safety systems based on real-time human behavior.

MIT 6.S094: Recurrent Neural Networks for Steering Through Time

The lecture provided a comprehensive technical overview of recurrent neural networks, contrasting vanilla architectures with Long Short-Term Memory (LSTM) units to address vanishing gradient challenges in processing sequential data. It detailed core optimization mechanics, including backpropagation and gradient stabilization, while showcasing diverse applications ranging from machine translation and medical diagnosis to autonomous driving systems that utilize image sequences to predict steering and speed. The session concluded by emphasizing the heavy reliance on manual hyperparameter tuning and massive datasets, while setting the stage for future discussions on driver state analysis and an upcoming White House AI policy speaker.

MIT 6.S094: Convolutional Neural Networks for End-to-End Learning of the Driving Task

This lecture explores the application of Convolutional Neural Networks to computer vision challenges, specifically demonstrating how deep learning models surpass human performance on benchmarks like the CIFAR-10 dataset to enable autonomous driving systems. It details the architectural differences between convolutional, pooling, and fully connected layers while contrasting browser-based training tools like ConvNet.js with robust offline implementations in TensorFlow. The session concludes by addressing critical industry hurdles such as data scarcity for rare edge cases and the necessity for near-perfect accuracy to ensure safety in real-world deployment scenarios.

MIT 6.S094: Deep Reinforcement Learning for Motion Planning

Lex

Participants in the Deep Traffic competition develop deep reinforcement learning agents to navigate a seven-lane highway simulation, aiming to achieve an average speed of 65 mph or higher through autonomous decision-making. Utilizing client-side JavaScript and Andrej Karpathy's ConvNet.js library, competitors train neural networks via experience replay and Q-learning algorithms without relying on explicit ground truth data for vehicle actions. The initiative serves as a practical framework for exploring the complexities of autonomous driving, highlighting both the potential of simulation-based learning and the critical challenges of aligning reward functions with real-world safety constraints.

MIT 6.S094: Introduction to Deep Learning and Self-Driving Cars

Lex Friedman, Dan Brown, William Angio, Spencer Dodd, Benedict Jenick, Andrej Karpathy, Hans Moraveck

MIT Course 6S094, led by Lex Friedman, utilizes self-driving cars as a case study to teach deep learning through two simulation projects: the reinforcement learning game Deep Traffic and the image-based control system Deep Tesla. The curriculum contrasts standard supervised learning with complex real-world challenges such as adversarial attacks and data inefficiency, requiring students to train neural networks to drive virtual vehicles at speeds exceeding 65 mph for credit. By analyzing the architectural modules of autonomy and historical milestones like the DARPA Grand Challenge, the course bridges theoretical computer science with the practical safety constraints of deploying artificial intelligence in unstructured environments.