My First Tech

There's an incredible buzz around AI agents, with the prevailing wisdom suggesting that bigger is always better. The industry has poured billions into monolithic, Large Language Models (LLMs) to power these new autonomous systems. But what if this dominant approach is fundamentally misaligned with what agents truly need?This episode dives deep into compelling new research from Nvidia that makes a powerful case for a paradigm shift: the future of agentic AI isn't bigger, it's smaller. We unpack the core arguments for why Small Language Models (SLMs) are poised to become the new standard, offering superior efficiency, dramatic cost savings, and unprecedented operational flexibility.Join us as we explore:Surprising, real-world examples where compact SLMs are already outperforming massive LLM giants on critical tasks like tool use and code generation.The key economic and operational benefits of adopting a modular, "Lego-like" approach with specialized SLMs.A clear-eyed look at the practical barriers holding back adoption and the counter-arguments from the "LLM-first" world.A concrete, 6-step roadmap for organizations to begin transitioning and harnessing the power of a more agile, cost-effective SLM architecture.This isn't just an incremental improvement; it's a potential reshaping of the AI landscape. Tune in to understand why the biggest revolution in AI might just be the smallest.The research paper discussed in this episode, "Small Language Models Are the Future of Agentic AI," can be found on arXiv:https://arxiv.org/pdf/2506.02153

It looks incredibly impressive when a large language model explains its step-by-step thought process, giving us a window into its "mind." But what if that visible reasoning is a sophisticated illusion? This episode dives deep into a groundbreaking study on the new generation of "Large Reasoning Models" (LRMs)—AIs specifically designed to show their work.We explore the surprising and counterintuitive findings that challenge our assumptions about machine intelligence. Discover the three distinct performance regimes where these models can "overthink" simple problems, shine on moderately complex tasks, and then experience a complete "performance collapse" when things get too hard. We'll discuss the most shocking discoveries: why models paradoxically reduce their effort when problems get harder, and why their performance doesn't improve even when they're given the exact algorithm to solve a puzzle. Is AI's reasoning ability just advanced pattern matching, or are we on the path to true artificial thought?Reference:This discussion is based on the findings from the Apple Machine Learning Research paper, "The Illusion of Thinking: Understanding the Strengths and Limitations of Large Language Models with Pyramids of Thought."https://machinelearning.apple.com/research/illusion-of-thinking

What happens when AI becomes vastly smarter than humans? It sounds like science fiction, but researchers are grappling with the very real challenge of ensuring Artificial General Intelligence (AGI) is safe for humanity. Join us for a deep dive into the cutting edge of AI safety research, unpacking the technical hurdles and potential solutions. We explore the core risks – from intentional misalignment and misuse to unintentional mistakes – and the crucial assumptions guiding current research, like the pace of AI progress and the "approximate continuity" of its development. Learn about the key strategies being developed, including safer design patterns, robust control measures, and the concept of "informed oversight," as we navigate the complex balance between harnessing AGI's immense potential benefits and mitigating its profound risks.An Approach to Technical AGI Safety andSecurity: https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/evaluating-potential-cybersecurity-threats-of-advanced-ai/An_Approach_to_Technical_AGI_Safety_Apr_2025.pdfGoogle Deepmind AGI Safety Course: https://youtube.com/playlist?list=PLw9kjlF6lD5UqaZvMTbhJB8sV-yuXu5eW

Ever tried to understand how AI actually learns, only to get lost in a sea of equations and jargon? This episode is your fast track through the fundamentals of machine learning, breaking down complex concepts into understandable nuggets.Drawing inspiration from Stanford course materials, we ditch the dense textbook approach and offer a clear, conversational deep dive into the core mechanics of AI learning. Join us as we explore:Linear Predictors: The versatile workhorses of early ML, from classifying spam to predicting prices.Feature Extraction: The art of turning raw data (like an email) into numbers the algorithm can understand.Weights & Scores: How AI weighs different information (like ingredients in a recipe) to make a prediction using the dot product.Loss Minimization & Margin: How do we measure when AI gets it wrong, and how does it use that feedback (like the concept of 'margin') to improve?Optimization Powerhouses: Unpacking Gradient Descent and its faster cousin, Stochastic Gradient Descent (SGD) – the engines that drive the learning process.Whether you're curious about AI or need a refresher on the basics, this episode provides a solid foundation, explaining how machines learn without needing an advanced degree. Get ready to understand the building blocks of artificial intelligence!Stanford's Algorithms for Artificial Intelligence: https://web.stanford.edu/~mossr/pdf/alg4ai.pdf

 It's over 50 years old, yet the C programming language remains a fundamental building block beneath countless technologies we use daily, from operating systems to embedded devices. Join us as we dive deep into the core principles of C, guided by the classic Kernighan & Ritchie text. We explore its history, key syntax, control flow, functions, and the low-level power that explains why C still matters profoundly today.The C Programming Language (2nd Edition): http://cslabcms.nju.edu.cn/problem_solving/images/c/cc/The_C_Programming_Language_%282nd_Edition_Ritchie_Kernighan%29.pdf

Open source software powers our lives, from our phones to our fridges. We uncover its enormous economic impact, revealing how much businesses save (trillions of dollars!), the importance of different programming languages, and how to ensure its sustainable future.The Value of Open Source Software: https://www.hbs.edu/faculty/Pages/item.aspx?num=65230

Ever wondered how AI is transforming our lives, from smart homes to the stock market? In this episode, we explore the fascinating world of AI agents – systems that can perceive, reason, and act to achieve goals. We'll unpack the core concepts from "AI Foundations of Computational Agents," delve into real-world examples like trading and tutoring agents, and discuss the blend of neural networks and logical reasoning that powers them. Get ready to understand the building blocks of AI's future!Artificial Intelligence: Foundations of Computational Agents (3rd Edition): https://mrce.in/ebooks/AI%20Foundations%20of%20Computational%20Agents%203rd%20Ed.pdf

Unlock the secrets behind lightning-fast data retrieval! We explore the inner workings of database storage engines, from simple bash scripts to complex B-trees. Discover how indexes work, the trade-offs between speed and storage space, and how data warehouses organize vast amounts of historical data.Designing Data-Intensive Applications (free chapters): https://www.scylladb.com/wp-content/uploads/ScyllaDB-Designing-Data-Intensive-Applications.pdfDDIA book: https://dataintensive.net

Large Language Models are everywhere, but how do they actually work? Join us as we unravel the science and engineering behind these powerful AIs that can write, translate, and even create. From pre-training to fine-tuning, we explore how they learn and what it means for the future of tech. Foundations of Large Language Models: https://arxiv.org/pdf/2501.09223

Dive into the core of software engineering and system design with us! This episode, we're breaking down key concepts, real-world case studies, and emerging trends using the fantastic insights from ByteByteGo's System Design 101. Whether you’re prepping for an interview, just curious about how systems work, or simply love a good tech deep dive, we've got you covered. We demystify everything from programming languages and network protocols to API security and cloud-native architectures. Get ready to unlock the secrets behind building, deploying, and managing modern applications. Key Themes Covered: - Programming Languages (Compiled vs Interpreted) - Network Protocols (HTTP, TCP, UDP) - API Security and Authentication - Software Architecture Patterns (MVC, MVP, MVVM) - Design Patterns - Solid Principles and "KISS" - Microservices - Database Types and Management - Cloud Computing (IaaS, PaaS) - Cloud Native Approach - CI/CD - Technical Interviews - Object Oriented Programming - The Human Side of Software Development ByteByteGo's System Design 101: repository and document.

Are you working on a machine learning project and feeling overwhelmed? This episode dives into the practical strategies outlined in Andrew Ng's "Machine Learning Yearning", offering a roadmap for building and improving your AI systems. Learn how to avoid common pitfalls and make rapid progress by understanding key concepts like: Setting up effective development and test sets: Discover how to choose data that reflects your future needs, and why your dev and test sets should come from the same distribution. Avoid the pitfall of using training data that does not match what you want to perform well on. The importance of a single-number evaluation metric: Understand how to establish a clear, measurable goal for your team, and how to combine multiple metrics into one. Learn the difference between optimizing and satisficing metrics. The power of iterative development: Learn to quickly build a basic system and then use error analysis to identify the most promising directions for improvement. Error analysis techniques: Explore how to manually examine misclassified examples to identify error categories and prioritize your work. You will also learn how to evaluate multiple ideas in parallel. Discover the importance of cleaning up mislabeled data. Understanding bias and variance: Learn to diagnose and address the two main sources of error in machine learning models. Explore how to interpret learning curves and compare your algorithm to human-level performance. The critical role of human-level performance: Discover how comparing your system to human-level performance can help you estimate optimal error rates and set achievable goals. Tackling data mismatch: Learn strategies for addressing situations where your training and dev/test sets come from different distributions, including how to create artificial data. Debugging inference algorithms: Use the Optimization Verification test to pinpoint problems in your scoring function or search algorithm. End-to-end learning: Understand the pros and cons of end-to-end learning, and how to decide when to use it. Pipeline design: Learn how to choose appropriate components for your pipeline, balancing data availability with task complexity. Understand how to analyze your pipeline by parts to focus on specific improvements. This episode will equip you with actionable strategies to make your machine learning projects more efficient and effective, helping you become a "superhero" in your field.

Join us to discuss about the book "Operating Systems: Three Easy Pieces" which teaches operating system concepts through a dialogue format between a professor and student. The excerpts cover various topics, including CPU and memory virtualization, concurrency, persistence (file systems and devices), scheduling algorithms (like MLFQ, lottery, and CFS), process management, memory management (paging, segmentation, allocation), and distributed systems. The book uses a combination of explanation, code examples, and figures to illustrate concepts, and includes homework assignments and references to seminal operating systems papers. It emphasizes the importance of understanding the underlying mechanisms of operating systems and the trade-offs involved in designing them.

This episode explores how to approach problem-solving through computer programming using Python. We'll learn about the core concepts in "Think Python," including variables, expressions, functions, and control structures, and how these elements work together to create effective programs. We'll focus on developing a "computer science" mindset, and using concepts like incremental development and debugging to tackle complex problems. We will also discuss data structures like lists and dictionaries, object-oriented programming, and file handling to write practical programs, with examples drawn from the textbook. Think Python 2nd edition: book

Join us as we delve into the classic textbook, "Structure and Interpretation of Computer Programs" (SICP), exploring its key concepts and timeless principles. From abstraction and recursion to metacircular evaluators and logic programming, we unpack the ideas that have shaped generations of programmers. We’ll discuss how these ideas are still relevant today, drawing on concrete examples and explanations from the text and our own exploration. We will be referencing different concepts in the text, including Lisp as a language for describing processes, data abstraction, generic operations, message passing, stream processing and the implementation of evaluators, to fully understand this foundational text.

This episode explores the mathematical concepts underlying computer science. We'll begin with logic and proofs, examining predicates, propositional formulas, and truth tables. We'll then discuss induction, recursive definitions, and structural induction for analyzing data types. The podcast will cover state machines, the stable marriage problem, and how derived variables help in proving termination. We'll also touch on binary relations, generating functions, and basic probability. Finally, we'll delve into the connections between these concepts, random walks, and a simplified version of Turing's code, showing the link between math and cryptography. This episode is ideal for anyone wanting to grasp the mathematical side of computer science. Mathematics for Computer Science: https://ocw.mit.edu/courses/6-042j-mathematics-for-computer-science-spring-2015/mit6_042js15_textbook.pdf

This episode delves into the fascinating world of competitive programming, exploring the crucial role of algorithms and their real-world applications. Uncover the importance of time complexity in crafting efficient solutions, going beyond simple code to strategies that optimize performance. Explore a variety of fundamental algorithms, from classic sorting techniques like binary search to powerful data structures like vectors, sets, and maps. Gain insights into the art of complete search, a powerful approach to problem-solving that guarantees correct answers by systematically exploring all possibilities. Journey into the realm of greedy algorithms, understanding how making the best choice at each step can lead to elegant and efficient solutions. Discover dynamic programming, a technique that blends the thoroughness of complete search with the speed of greedy algorithms. Join us as we unravel the secrets of competitive programming and learn how to master the algorithms that power the digital world. Competitive Programmer’s Handbook: book

Mobile apps are complex systems with unique reliability challenges. Unlike servers, where engineers have complete control, mobile app reliability depends on billions of devices and user behaviors. This episode explores how Site Reliability Engineering (SRE) principles can be applied to build and manage reliable native mobile applications. Join us as we discuss: Key concepts in engineering reliable mobile apps. Unique challenges of mobile app reliability such as scale, control, monitoring, and change management. Real-world case studies of mobile app outages at Google and the lessons learned. Practical strategies for measuring and improving app availability, performance, and efficiency. The impact of mobile app changes on server-side infrastructure. Whether you are a mobile app developer, an SRE, or just interested in learning about mobile app reliability, this episode has something for you. Book Engineering Reliable Mobile Applications: https://sre.google/resources/practices-and-processes/engineering-reliable-mobile-applications/

Join us as we explore Google's in-depth course, "Comprehensive Rust," a free resource designed to take you from Rust novice to proficient developer.​We'll discuss the course's structure, covering Rust fundamentals, specialized deep dives into Android, Chromium, and bare-metal development, and a full day dedicated to concurrency.​We'll highlight the course's hands-on approach, emphasizing practical examples and exercises that solidify your understanding of key concepts like memory management, ownership, borrowing, and error handling.​We'll also touch upon the course's focus on real-world applications, showcasing how Rust integrates with Android, Chromium, and embedded systems.​Whether you're a seasoned programmer or just starting your Rust journey, this episode will offer valuable insights and practical tips to accelerate your learning and empower you to build robust and efficient software with Rust.Comprehensive Rust course: https://google.github.io/comprehensive-rustRust programming language: https://www.rust-lang.org

In this debut episode, we dive into the world of Kotlin, the modern programming language that’s taking the tech world by storm. Whether you’re a beginner or an experienced developer, Kotlin offers a fresh, efficient way to code—whether for Android apps, server-side development, or even multiplatform projects. We’ll explore what makes Kotlin unique, why it’s loved by developers worldwide, and how you can get started with it today. Tune in to learn about its concise syntax, powerful features, and why it’s the future of programming!Learn more about Kotlin: https://kotlinlang.org