text stringlengths 0 2.68k |
|---|
**Salvatore Sanfilippo:** Thanks. |
**Adam Stacoviak:** Thank you, Salvatore. |
**Salvatore Sanfilippo:** Thank you. It was very great to talk with you, and thanks for the hospitality. |
**Adam Stacoviak:** Oh, yes. Glad to have you back. |
**Jerod Santo:** Yeah |
**Outro**: \[01:28:17.29\] |
**Jerod Santo:** So you recently wrote a post called, "We are destroying software", which - you know that, but I'm telling you that as a setup... Which was almost a poetic piece of maybe a little bit of a rant, with like 19 things that we're doing, roughly... I didn't count one by one. But ways that we are destroying s... |
**Salvatore Sanfilippo:** In general, everybody I look, I see that the complexity is increasing without a proportional utility, usefulness of the systems. For example, one of the most important things that we have right now, which are LLMs, are like 3,000 lines of code for the inference, and even if you look at DeepSee... |
• Zipline's start in delivering blood to hospitals in Rwanda, chosen for its health-focused problem and potential for impact |
• Initial lack of technology, with early plans made using pencil sketches and a slide deck |
• Gradual expansion into other use cases, including health supplies, animal health, and eventually e-commerce and food delivery |
• Emphasis on environmental sustainability, with a 20x improvement in supply chain environmental footprint |
• Design of the drone, featuring a larger drone that delivers a smaller "zip" to the ground for delivery |
• Reduction in bird injuries and fatalities through drone delivery, estimated to be 10x to 100x lower than traditional delivery methods |
• Designing a drone that can deliver packages in tight spaces without being too loud |
• Creating a two-part architecture with a main drone and a mini drone that comes to the ground for delivery |
• Using a mini drone with thrusters to stay precise and oriented, similar to spacecraft reaction control thrusters |
• Developing software for the two drones, including divergent operating systems and a distributed update platform |
• Conducting extensive testing, including simulation and ground testing, to develop control systems for the drones |
• Deploying software updates over the air, with a process that takes around 6 weeks |
• Managing software updates, including flight software and cloud software, and ensuring safety and reliability |
• Drones used for defense were too expensive and had limitations |
• Company had to develop their own solutions and datasets due to unique flying conditions |
• They operate at a large scale with a fleet of over 400 drones |
• Software includes cloud autonomy, fleet level autonomy, and aircraft level autonomy |
• Weather forecasting and airspace design are critical components |
• Regulation plays a significant role, with different rules in various areas |
• Company has to work closely with regulators to create precedent and navigate complex airspaces between countries |
• The speaker describes a unique regulatory environment for their company, where they have a high-stakes scenario with potential crashes or accidents, but also a willingness to collaborate with regulators. |
• The company's approach to regulation is to convince themselves of the safety and efficacy of their products before presenting it to regulators, through rigorous analysis, testing, and data review. |
• The speaker introduces the "Swiss Cheese Model" of testing, which involves multiple layers of testing (unit testing, static testing, software-in-the-loop testing, hardware-in-the-loop testing, flight tests) to ensure comprehensive coverage and confidence in the product's safety. |
• The company's release planning involves multiple layers, including testing, delivery, observability, and potential re-testing, to ensure that new capabilities are thoroughly vetted before being deployed in production. |
• The company has a large-scale testing infrastructure, with over 10,000 flights a week, and uses automated software systems to simulate extreme conditions and test the system's resilience. |
• Development of aircraft's ability to transition from hover to fixed wing and back |
• Release process, including testing and stabilization |
• Data collection and logging for safety and problem analysis |
• Use of AI for forecasting extreme weather events |
• Protocols for handling severe weather, including parachute deployment |
• Wind limits for safe flight |
• Analysis of data from thousands of flights to understand statistics and trends |
• Challenges of flying through mountain passes with high winds |
• Development of control systems to navigate through gusty winds |
• Benefits of using wings for drone flight, including increased range and efficiency |
• Company's approach to prioritizing safety and urgency in development and deployment |
• Range anxiety and limitations of current drone technology |
• Plans for future development, including the ability to dock at different locations and increase service radius |
• Cost and pricing of drone technology |
• Considerations for reliable and consistent delivery services, such as weather and terrain obstacles |
• Range limitations of 10 miles and 50 miles for different drone models |
• Importance of customer feedback and partnerships in developing a useful platform |
• User experience of ordering food through the Zipline app or partner apps |
• Logistics of ordering, prep, and delivery through the Zipline system |
• Technical aspects of API integration with partner companies (e.g. Chipotle) |
• Safety features of the system, including drone avoidance and delivery point precision |
• Future plans for expansion and human-centered delivery |
• Avoidance and navigation of the delivery drone to prevent it from becoming a game for children |
• Potential for future features such as allowing orderers to set a mode for fun or efficiency |
• Current ability to deliver 24/7, with delivery times limited only by partner hours |
• Benefits of 24/7 delivery, including increased commerce and addressing of night shift worker needs |
• Possibility of integrating with telehealth services to deliver medication and other health-related items |
• Long-term vision for Zipline's impact on the world, including widespread adoption and ubiquitous delivery |
• Importance of quiet, low-noise operation in drone delivery systems |
• Benefits of timely deliveries, including fresh food and essential items |
• Potential for drone delivery to improve customer experience and reduce wait times for services like HVAC repairs |
• Examples of creative uses of drone delivery, such as delivering Easter eggs and gifts |
• Current limitations of drone delivery (e.g. 8-pound weight limit) and potential future expansions |
• Discussion of competition and innovation in the field of drone delivery |
• Zipline's competitors and admirers in the drone delivery space, including Amazon and Google's Wing project |
• Zipline's plans for scaling up operations in 2026, including expanding the supply chain and building charging sites |
• The selection process for new cities to launch in, which focuses on testing and validation of the system in similar weather and terrain conditions |
• Technical challenges ahead, including improving weather forecasting and predicting icing and lightning conditions |
• Managing the airspace and fleet at scale, including issues with migration and swarming |
• Future use cases and capabilities, including healthcare and new systems for humans and patients |
• Discussion of autonomous flight systems and their capabilities |
• Comparison of human vs. AI performance in tasks such as soccer and drone operation |
• Mention of a soccer goalie AI that outperformed a human |
• Fault tolerance and robustness of autonomous systems |
• Example of using autonomous drones for delivery in challenging conditions |
• Safety and scalability concerns in autonomous flight |
• Tech stack and programming languages used by Zipline, including Rust and ROS |
• Future of autonomous flight and the need for open-source collaboration |
• Idea of a shared communication channel for drone systems to avoid collisions |
• Drones share information live with each other over a point-to-point radio link and IP network |
• Partnership with a commercial drone operator in Dallas will use the same protocol to share airspace |
• Studying the DNS system to learn how to create a democratized and open airspace |
• Distributed algorithms allow drones to adjust their course to avoid collisions |
• Layering and vertical separation may be used to create efficiency and safety in the airspace |
• Formalizing lanes or highways for drone traffic may help prevent congestion and improve safety |
• Soft and hard keepouts may be used to direct traffic into specific areas |
• Technical challenges in a project |
• Customer challenges in a project |
• General positivity and appreciation for the conversation |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.