What does it take to climb the highest mountains on Earth, ski across icy deserts and frozen seas, or reach other ambitious goals? A solid plan, proper preparation, a strong team, and plenty of willpower. Put together all of that and what could possibly go wrong? Well, a lot ...

Listen in to learn more about the big and small mistakes Anja has lived through on her expeditions and how she has handled them – so you won’t have to do the same.

People have been programming computers for more than 80 years, but there is little agreement on how to design software or even what a good design looks like. As a community, we talk a lot about tools and processes, but hardly at all about design. In this talk I will describe my recent work to identify and communicate a set of software design principles, including a new software design course at Stanford that is taught more like an English writing seminar than a traditional programming class, and a book on software design, which is based on the concepts from the class. I will also present a few of the design principles, such as "classes should be deep" and "general-purpose classes are deeper."

Over the last two years, the authentication workflow of BestSecret's webshop has been converted from an internal solution to Auth0, an external ID provider, to support the transition from a monolithic architecture to multiple domain services. During this time, we proceeded step by step to overcome the hurdles that arose during this initially seemingly banal change. In our talk, we want to present the architecture of our new implementation and give a report on the transition period.

Factorio is a simulation game where you build a highly automated factory, producing the technology to eventually launch a rocket to space. Along the way, every new task will be manual at first, but you can (and will need to) automate it.

I stumbled upon this game a few years ago, got hooked, and spent (way!) too many hours playing it. While I was knee-deep in refactoring my base in Factorio, making it more scalable and resilient, at some point I realized that this is eerily similar to what I’m doing in my day job. (Probably also the cause why I’m so seriously hooked on that game.)

So although this talk will look like it’s about Factorio, it’s in fact about how to bootstrap and scale teams and projects, about monoliths and microservices, and about the cost of maintenance.

Digital art is currently experiencing an upswing which increases its visibility and pushes the boundaries of traditional art using technological and conceptual innovations. Artists are using software and complex technologies for creative processes and interactive artworks to actively engage visitors.

In 2025, the UBS Digital Art Museum - Europe's first museum for digital art on an enormous scale spanning over 6,500 square meters - will open in Hamburg's HafenCity. The grand opening with the exhibition teamLab Borderless will showcase the fascinating fusion of art and technology that has already sparked a whole new enthusiasm for art among millions of people in Asia.

The UBS Digital Art Museum strives to become a leading platform for digital and immersive art in Europe and to make art accessible to a broad and new audience.

Lars Hinrichs, initiator of the UBS Digital Art Museum, gives an exclusive insight into this groundbreaking project and the inspirational effect of digital art on visitors.

We live surrounded by digital devices and use the Internet excessively through our own devices. We are often unaware of the form in which data is collected about us and by whom. This often happens in the background without us knowing the technical mechanisms at all. With TRUST-AI, Bernd Lintermann and Florian Hertweck not only want to inform about these facts, but to make them experienceable for visitors. An artificial intelligence communicates with the users, and in the course of the dialogue, the visitor plays with the capabilities of the AI, which, among other things, analyzes his emotional state and measures his pulse, and finally hijacks his face and assumes his identity. The initial promise to improve lives turns into the opposite. The lecture examines the installation from a content and technical point of view - with its numerous AI applications - and gives insights into the difficulty of technically replicating human communication.

Recent advances in artificial intelligence and computation have enabled an impossible feat dreamed about for thousands of years - the recovery of lost texts from antiquity. In this talk, we will describe how a library destroyed in the famous eruption of Mt Vesuvius is on track to being understood after 1700 years underground and 270 years of mostly failed attempts. Combining high resolution synchrotron X-ray CT scanning and a race to develop software, the effort yielded thousands of characters on Epicurean philosophy, with the promise of millions more to come.

We will explore the current state of artificial intelligence, highlighting recent advancements and the potential impact on businesses. Our focus will include both the opportunities for innovation and the practical challenges associated with integrating these technologies.

This talk introduces you to the world of generative AI with a focus on text to image and text to video for the creation of images and short videos. We will explain how neural networks use diffusion models and so-called transformer architectures to multimodally generate different output formats from short text inputs.

We focus on state-of-the-art technologies such as Sora or Midjourney. The applied tools, such as latent diffusion models, allow us to generate and process images and videos by combining text understanding through attention mechanisms and transformers through image denoising processes.

An in-depth look at the video generation process using Sora shows how visual data is compressed, split into patches and then reconstructed into the final video. In addition to Sora, we also discuss alternative methods such as Runway to cover a wide range of tools for image and video generation.

By the end of this talk, you will have gained a basic understanding of diffusion models, an overview of image and video generation tools, and a deeper understanding of how one selected tool works. The talk will be complemented by practical examples and demos.

In this talk, we share our insights and experiences on the design of language model-based agent systems and provide an overview of the current state of the art. Language models open up new horizons for a variety of practical applications. Due to their ability to process and generate natural language texts, they are often attributed with the ability to plan and reason. This is also why the integration of language models into autonomous agent systems is seen as promising. We answer the following questions: What are the pitfalls and limitations in designing such systems? How can agent systems be evaluated? Does the world need generative agent systems at all?

Aviation leasing is an intriguing, awe-inspiring and an incredibly nuanced industry. Like the rest of the world, it too is pivoting to a more digital landscape either to gain throughput efficiencies, increase customer satisfaction or create new streams of revenue. Join me as we take a look as to how TNG and MTU, have been challenging the status quo in developing our flagship product - EFM Lease. We will discuss what prompted the genesis of such a product, state of the market today and what vision we aim to achieve with the product by the end of our development lifecycle.

A combination of big data, big compute and novel architectures has given us ChatGPT, Claude and Sora. These models are capable of solving increasingly challenging tasks in domains such as language, image, video, robotics, & games. At the same time, the models still hallucinate, fail to solve simple (at least to us) reasoning problems, and lack the ability to robustly plan ahead. What will it take to get AI significantly beyond its current capabilities? We discuss this through a combination of ideas in mathematics, computer science and neuroscience, touching upon themes such as information composition, openendedness, dataset generation and model architectures.

In this talk we will discuss the challenges and recent advances around three core components of a strong AI system: 1) Explainability: can we understand the system’s decision making, 2) Efficiency: can we minimise the required computational power for learning and inference, and 3) Robustness: can we trust the system to act fully autonomously?

In this talk, we will dive into the technical aspects of Meminto, an AI-based open-source tool that automatically generates meeting minutes from audio recordings. We'll explore the use of state-of-the-art open-source AI tools such as 'pyannote.audio' for speaker diarization and 'Whisper' for speech-to-text conversion. By focusing on the engineering process, the talk aims to inspire and empower developers to build their own AI-powered tools.

In this talk, we explore generative artificial intelligence (AI) and the associated challenges regarding intellectual property rights and compliance. We start with a brief explanation of what generative AI is and how it generates content such as text, images, and music. A core topic is the issue of property rights in the context of generated content. We discuss who actually owns these rights and how the legal understanding of such content has materialised. We then take a look at the European Union's AI regulation and digital strategy as well as other local laws that are intended to regulate the use of AI. Another important aspect is the licensing of AI models. Among other things, we will explain how open source concepts can be applied to AI models and how different principles are reflected in the licensing process. This talk is for anyone who wants to develop a basic understanding of the legal and regulatory aspects of generative AI and offers hands-on insights into current developments.

In this talk, we will derive that AI provides well suited methods to empower marketing use cases at scale. Furthermore, we will holistically explore selected AI solutions including their business context as well as details about implementation and deployment.

From Deep Blue's historic victory over Garry Kasparov in chess in 1997 to Cicero's success in the natural language negotiation game Diplomacy in 2022, planning algorithms have consistently played a crucial yet often overlooked role in AI milestones. This talk will explore the profound impact of planning algorithms on AI breakthroughs in chess, Go, poker, and Diplomacy. I will then discuss the implications of this research as the AI landscape shifts from narrow, task-specific systems to broadly capable foundation models.

Prompt engineering, RAG, fine-tuning - there are many ways to individualize the behavior and knowledge of LLMs. But when does optimization make sense?
GPT-4, Llama3 and many other Large Language Models (LLMs) impress with their ever-improving language skills. But going from a raw LLM to a production-ready application is usually a long way off, especially when domain knowledge is needed. This talk provides an overview of how to bring the full power of LLMs to your business case to the road.
In this talk, we will introduce the three most common methods to specialize LLMs for specific use cases – prompting, retrieval augmented generation (RAG) and fine-tuning – and classify when each of these optimizations makes sense. We start with a look under the hood at how today’s LLMs work and to what extent they are limited by this. Using an application example, a question-answering digital doppelganger, we will approach a workable product step by step. We will show how you can dramatically improve the quality of your LLM responses using prompt engineering, how you can use RAG to get your domain knowledge into the LLM and how you can use fine tuning to put the finishing touches to your LLM. Again, we take a deeper look into technology to understand when each approach makes sense – and when it does not.

In this talk I will share my experience in the world of astrophysics, machine learning and music. I will discuss how high performance computing can help us understand the universe with cosmological simulations and how machine learning and AI can help us make sense of these simulations. Then I will talk about Earth Observations, that is, what happens when you point satellite-born telescopes in the "wrong" direction to observe our planet and how they help us monitor and prevent natural disasters. In the end comes Uranus - the title of a song I wrote with my band (Nanowar Of Steel) which has been my second full-time job while working as a researcher and a data scientist.

A cloud migration is not just a question of technical implementation, it also requires a reorganization of the entire corporate IT. In this presentation, we describe how thousands of applications are being migrated to the cloud at BMW Group IT and what experience we have gained so far.

"How I Ended Up Being So Puzzled" is a lighthearted talk that follows one person's surprising journey from a childhood love of jigsaw puzzles to the adrenaline-fueled world of speed puzzling competitions. You'll hear how an uninspiring movie night led to an unexpected discovery, opening the door to competitive puzzle races and a whole new community.
Whether you're a casual puzzler or have never touched a puzzle piece, this talk will introduce you to the fascinating world of speed puzzling, sharing the excitement, strategies, and unexpected friendships that thrive around this unique hobby. Come discover a whole new way to get puzzled!

Chess has grown more in the last three years than in the 2000 years before. Why is that and what can we learn from it? How has chess managed to move from a niche to the centre of the influencer and media world? What drives the chess elite, why is chess a relevant sports case study for MIT and how did Chess.com become one of the 50 most influential and innovative companies in the world? All this and more in this talk.

At Big Techday 24, world-class speed puzzler Chiara Dellantonio has a double duty. In addition to her talk, she will be taking on a special challenge: In the BTD24 Speed Puzzling competition, she will be competing solo against up to 20 two-player teams to solve our BTD24 puzzle. This is a unique opportunity for amateur puzzlers to compete with a pro and put their skills to the test.

Our external guests can register for the competition by sending an email to bigtechday-orga@tngtech.com. Please indicate whether you are registering as a team of two or without a puzzle partner. We look forward to welcoming many puzzle enthusiasts.

In an era of rapid technological convergence, it is crucial to think in a holistic way using scenario analyses and strategic anticipation. Studies show that cross-technology innovations are increasing faster than innovations in isolated technology areas. Developments in future technologies such as AI, blockchain, metaverse and quantum computing are fueling each other. The lecture takes up current developments and scenarios of the most important technologies, puts them into context and helps to systematically decipher the code of the future. A lot of what sounds fantastic and like science fiction at first glance is already science fact. This makes it even more important to be prepared and set the course for the future now.

WARR Space Robotics is a subsection of the WARR (Scientific Workgroup for Rocketry and Spaceflight, German: Wissenschaftliche Arbeitsgemeinschaft für Raketentechnik und Raumfahrt), a student group working on projects related to space technologies. WARR Space Robotics is specifically dedicated to creating robotic technologies that would significantly de-risk future space exploration missions. Among the projects within WARR space robotics is Project CRATER, which develops a Compact Rover for Autonomous Traversal, Experimentation and Retrieval. The mission of WARR Space Robotics is to create an autonomous rover platform that would provide an interface to science-oriented experiments. Therefore, beyond participating at international rover competitions and internal educational purposes for team members, WARR Space Robotics also aims to develop a system that can be used by external customers in the field of space exploration to further investigate payload technologies and their applications. Among external stakeholders, the Chair for Lunar and Planetary Exploration Technologies (LPE) at the Technical University of Munich (TUM) is our main focus. We will offer our rover as a platform for testing scientific instrumentation that the chair is developing for volatile detection, micro-meteorite detection, exospheric measurement, among other things.

Dynamic Soaring is a flight technique that enables radio controlled gliders to reach unexpectedly high speeds. Over the last 23 years, the top speed for Dynamic Soaring has climbed from 156mph to the current record of 564mph (908kph). Our latest design pushes the limits, using swept wings to delay the onset of transonic drag rise, in hopes of recording a speed of 610mph or greater.

Elara Aerospace is the first student team to launch a methalox rocket beyond 100km.
In this talk we’ll go through everything from quality management to technical aspects of our rocket engine. We’ll share insights into our strategy on recruiting, and how a non-profit student team can get the material means to realize this ambitious project.
Join the talk & get excited with us as we make space history. It’s rocket science!

In my talk, I will share the lessons we learned from our rocket project during launch campaigns like the European Rocketry Challenge 2023. I will discuss how we functioned under immense pressure and pushed ourselves beyond our limits to achieve what seemed impossible. A specific story from 2023 will be the core of my presentation: The dramatic search for a spring at 3:45 am. Having just gotten up at that time, I scoured the entire house for a suitable spring, which was the single point of failure for the launch later that morning. And in the end? It was actually found in a household appliance! This story is a vivid example of the power of collaboration, perseverance, and creativity in challenging situations.

Let's explore the software ecosystem within German healthcare practices and deep dive into a particular kind: the prescription software. The talk includes a detailed look at the prescription software developed by CGM and the impact of the FHIR standardized interface on data exchange and system interoperability. We will also highlight the role of the regulatory environment in the product.

Pathogen genomic sequencing data has proven crucial for understanding the dynamics of outbreaks and epidemics. Every day, many new sequences are generated worldwide, with a significant increase in effort made during recent epidemics such as the SARS-CoV-2 pandemic and the 2022 mpox outbreak. However, these sequences can only be useful when there are mechanisms for them to be stored, shared, retrieved, and analyzed. Despite efforts of several institutions, there is still a lack of general and reusable software for sequence sharing.

To address this, a collaboration of the ETH Zürich and other research institutes developed Loculus: an open-source, general-purpose software package for sharing, maintaining, and accessing microbial genetic sequencing data. Loculus targets any group managing sequencing data, from individual public health or research laboratories up to international databases facilitating global pathogen sequence sharing. It is virus-agnostic and enables real-time analysis.

In the past few years, Threat Modeling has emerged from the dark art niche where it has been for a long time into an actual discipline and a recognized part of the Secure Development Lifecycle. But there is still much to understand about what the process is and what the benefits are. Besides that, is it something we can extrapolate to other domains and find uses outside systems development? Can we use the techniques of Threat Modeling to make our lives, jobs, and communities if not easier then at least more manageable?
We will look at why Threat Modeling is considered to be "difficult", and how we can go past that and actually get value out of it. We'll look at the Threat Modeling Manifesto and at the Threat Modeling Capabilities documents, and learn how they can guide a Threat Modeling program. Finally, we'll see that Threat Modeling is not for me and you, it is for everyone, and the best part is, we are already constantly doing it!

Apache Flink is one of the most successful projects at the Apache Software Foundation. The open source stream processing framework allows teams to process data streams in real time at a very large scale. With stream processing, traditional batch-oriented data warehousing, data integration, and eventing use-cases can be solved with lower cost and much lower latency. Flink has reached close to 1m downloads in 2023 and is used in production everywhere - from startups to digital natives to more traditional enterprises. All hyperscalers offer some kind of service for it in their clouds. Ten years ago, this level of success was merely a dream of a group of students at TU Berlin. They worked on a research prototype for batch processing in the cloud, called Stratosphere. Stratosphere became Flink, they started a company called data Artisans to commercialize Flink … and the rest is history.

In this talk Robert Metzger, co-founder of data Artisans, and Konstantin Knauf, Director of Product Management for Flink and an early employee of data Artisans, will provide a behind the scenes perspective on the rise of Flink, the startups, business models and exits around it, and a glimpse into the future of Flink. The talk will not be recorded, so that the speakers can talk more openly about inside stories.

Quantum computing promises advantages for a number of structured computational problems. While the idea of quantum computing is not new, only within the last a bit more than five years protagonists have set out to actually build such devices to a reasonable scale. The quantum computers we have today are still somewhat noisy and not huge - but then, such devices seemed inconceivable not very long ago, creating an exciting state of affairs. This also comes along with lots of expectations and some hype. This talk will go on a journey deciphering what we can reasonably expect from such machines in the near future. It will present some exciting perspectives concerning achieving industrially relevant applications in machine learning and optimization. It will also debunk some of the most unreasonable of expectations and provide a reality check. Overall, this will give rise to a manic depressive ride through the landscape of one of the most exciting and promising future technologies.

Quantum computers have been gaining increased attention in recent years due to their potential for highly efficient computing in certain problem areas. Although there is currently no fully functional quantum computer, with the present hardware being relatively small and calculations prone to errors, both public and industrial research are heavily invested in their development and algorithmic improvement. This talk will present the current state of development and the potential of these relatively new computers. The discussion will cover making the functioning of quantum computers and quantum algorithms understandable for non-physicists, as well as explaining possible applications and the potential deployment areas of (potentially fully functional) quantum computers. Based on this, insights into research into the construction of quantum hardware are provided, focusing on the questions of how to improve the error-prone computations and an insight into the research aiming at implementing fault-tolerant calculations despite remaining error rates.

The promise of quantum computing – and the search for applications, use cases and even business cases – has ignited the imagination and drive of governments and industry throughout the world. There is a fundamental limitation though – based on the implicit assumption that Quantum Computing will be delivered through a supercomputer, or perhaps a hyperscale compute centre. The analogy with Thomas J Watson’s claim in 1943 that there would be a market for 5 computers, and the industry focus on Computer Facilities, dominated computing until the invention of the microprocessor in 1971. With Quantum Computing we will have the ability to skip the 28 year wait, and dive straight into a world of QPU’s. the technology behind this is based on Diamonds, and in my talk I will introduce the fascinating world of spin-systems, colour centres, QDK’s and how Quantum Computers using QPU’s no bigger than todays CPUs and GPUs will enable this new technology to fulfill its full potential.

The ScanPyramids project began in 2015 and is coordinated by the Faculty of Engineering at Cairo University in Egypt and the HIP Institute in France, with the partnership of institutions from Germany, Japan, France, and Canada, and in cooperation with and under the supervision of the Supreme Council of Antiquities in Egypt. The ScanPyramids project aims to investigate the internal structures of the Giza pyramids using non-destructive techniques, in order to better understand how they were built. The project achieved two discoveries in Khufu Pyramid of Giza: The ScanPyramids Big Void (SP-BV), announced in 2017, and the ScanPyramids North Face Corridor (SP-NFC), announced in 2023. The talk at Big Techday 24 will address mainly the role of modern non-destructive techniques in the discovery of a corridor in the Khufu Pyramid that had been hidden for 4,500 years. The presentation includes a short brief of the history of pyramids, the ScanPyramids project, an overview of the discoveries of the ScanPyramids Big Void and North Face Corridor, as well as the non-destructive techniques used, site measurements, and challenges.

Accessing data and applications in private networks is part of everyday life for many people, but at the same time it is almost always associated with workarounds when on the move. We will be showing live an innovative data service that brings a new approach to the mobile data world with Network as a Service, thus combining mobile and private networks simply and elegantly. We will also show how such an innovation was developed in a company like Telefónica from an idea to a product.

In 2023, neutrinos created in the particle collisions of the Large Hadron Collider (LHC) were detected for the first time. This talk will tell the story of the FASER experiment that led to the first discovery of these collider neutrinos. In contrast to the other large experiments at the LHC, FASER was assembled in just a few years from the remains of old detectors and actually pursues a completely different physics programme. Now, however, the detection of collider neutrinos opens up a new window for testing the standard model of particle physics and allows us to gain a deeper understanding of the strong interaction. The talk is explicitly aimed at all interested parties, especially those who have no prior knowledge of particle physics.

In her talk, Dr. Carmen Köhler will talk about her non-linear career path, which led her to become an analog astronaut, among other things. Analog astronauts conduct research for future crewed missions to the Moon or Mars in similar environments on Earth. During these Mars simulation missions, Carmen gained valuable experience of extreme environments and the challenges they present. She will give insights about what we can learn about human-robotic as well as human-human interaction in a time of artificial intelligence.

The circular economy is currently a hot topic of discussion, as more and more companies are putting the sustainability concept on their strategic agenda. Expected resource scarcity due to geopolitical dependencies, increasingly strict environmental legislation, and pressure from their own value chain are making it inevitable for many companies to think about circular solutions. But can a company overcome the challenges in this area on its own or do they need new innovation concepts that fundamentally change their own role in the value chain and the way they do business?

In this keynote, Bulent Altan, Founding Partner of Alpine Space Ventures and a pioneering figure in the NewSpace industry, explores Europe's strategic advantages in the space sector. With a background that includes various leadership roles at SpaceX, Bulent draws on his extensive experience in the space industry to discuss the untapped potential within Europe's talent pool and the critical importance of integrating space technology across various industries, from agriculture to biotechnology. Highlighting the need for tighter public-private collaboration in light of global geopolitical tensions, he underscores the vital, yet often overlooked, role of space operations in everyday life.

This talk addresses a fundamental query: How do we transport ourselves beyond the confines of a laboratory and into historical settings? Picture yourself in an ancient cathedral or immersed in the ambience of a bygone Parisian theatre. Through three compelling research endeavours focused on renowned locales such as Notre-Dame, a Parisian theatre, and the Louvre Museum, we explore the transformative potential of Virtual Reality (VR) in recreating authentic historical auditive immersive experiences. Our discussion encompasses various technical aspects including acoustic simulation, architectural visualization, archaeoacoustics, spatial sound, virtual avatars, crowd auralization, 360-degree videos, digital twins, and ecological validity.

The modern wave of AI, characterized by large language models and generative bots, is largely powered by a neural network architecture introduced in 2017, the Transformer. This talk aims to step through precisely what a transformer is, visualizing how data flows through it, with an emphasis on the attention mechanism.

Want to know more about CI/CD with GitHub Actions? Can it replace other popular CI/CD solutions? We'll tackle these questions in our session. We'll discuss how GitHub Actions work, where they run, and why it matters. We'll explain the concept of 'self-hosted runners', their uses, and how they can be run within Kubernetes.
Be part of this session to learn how GitHub Actions can improve your CI/CD system.

Join us for an exclusive look into the latest agricultural technology (AgTech) innovations that are changing the game for farmers, solution providers, and the entire food value chain. Yara Farming Solutions Europe will share multiple real-world product cases that showcase the power of technology in overcoming the industry's most pressing challenges. From crop nutrition optimisation to seamless integrations, discover how digital solutions farmers across Europe to stay ahead of the curve and shape the future of farming.

I will present a mental model that enables the different functions of a company to align, and improve value creation. Concrete examples will be provided, along with addressing common concerns, challenges, and controversial issues companies face when adopting a metrics-informed approach. A particular focus will be given to controversial topic of productivity measurement in software development, specifically SPACE and DORA frameworks.

Join us for our successful migration of Idgard, a security-first file storage and collaboration platform, to a modern stack built on S3 storage and Kubernetes. We'll cover the unique security challenges we faced, our solutions and key success factors, as well as people and process insights.

Since its stabilization in April 2023, Node.js Test Runner has been garnering attention within the community. Yet the pressing question remains: Is it truly production-ready? Is the technological shift justified?

In this presentation, we will delve into the technical and human challenges associated with deploying Test Runner in an active project. We'll tackle issues ranging from TypeScript integration and test discovery to IDE support and result reporting. But it's not all about overcoming hurdles. We'll also share the significant advantages and unexpected positives that have come to light after months of intensive use in critical projects with daily deployments and millions of users.

Our talk goes beyond release notes to offer a grounded perspective that only firsthand experience in the software industry can provide. Join us as we lay out a comprehensive evaluation of Test Runner's readiness and viability for production environments, backed by real-world insights and measurable impacts.

How do you modernize the beck-online application during operation? The application has been growing for more than 20 years and is not only the leading legal database in Germany, but also by far the company's most important and best-selling product with more than 50,000 active daily users. What does this have to do with the transformation of a traditional publishing house into a technology company? What is the impact of the fact that the world has been on permanent alert since 2020, that working from home has become socially acceptable virtually overnight due to coronavirus and that new disruptive technologies such as artificial intelligence are changing entire industries?

This talk dives into the challenges and solutions for the digitalization of C.H. Beck, a 260-year-old, owner-managed family business. Join us on a journey in which classic software becomes a modern platform and technology plays an increasingly important role in a traditional company. Experience the greatest highs and lows, successes and failures, decisions and wrong decisions of recent years with us at beck.tech.

The JavaScript ecosystem has long been known for its never-ending stream of new libraries and tools and for some time now, creating blazingly fast versions of existing tools seems to have been all the rage. As the maintainer of the bundler RollupJS, I could not stand by idly. This is the story of how Rollup 4 came to have native binaries, what problems I was facing and how you, too, can make your code blazingly fast. Or at least a little faster. You will learn about tools and techniques, but in the end, this is also a glimpse at the hard reality of open source maintenance.