Engineering

The world's largest sand battery has been inaugurated in Finland. Developed by Polar Night Energy, this high-temperature thermal energy storage system stores heat in sand using low-cost, clean electricity. The project is a powerful example of how thermal storage can enhance grid flexibility, decarbonise heating, and accelerate the energy transition. - It can store up to 100 MWh of thermal energy. - It has a round-trip efficiency of 90%. - It offers a cost-effective alternative to lithium-ion batteries for long-term heat storage. - By replacing an old woodchip plant, the sand battery is expected to cut the local heating network's carbon emissions by 70%.

Last week, I described four design patterns for AI agentic workflows that I believe will drive significant progress: Reflection, Tool use, Planning and Multi-agent collaboration. Instead of having an LLM generate its final output directly, an agentic workflow prompts the LLM multiple times, giving it opportunities to build step by step to higher-quality output. Here, I'd like to discuss Reflection. It's relatively quick to implement, and I've seen it lead to surprising performance gains. You may have had the experience of prompting ChatGPT/Claude/Gemini, receiving unsatisfactory output, delivering critical feedback to help the LLM improve its response, and then getting a better response. What if you automate the step of delivering critical feedback, so the model automatically criticizes its own output and improves its response? This is the crux of Reflection. Take the task of asking an LLM to write code. We can prompt it to generate the desired code directly to carry out some task X. Then, we can prompt it to reflect on its own output, perhaps as follows: Here’s code intended for task X: [previously generated code] Check the code carefully for correctness, style, and efficiency, and give constructive criticism for how to improve it. Sometimes this causes the LLM to spot problems and come up with constructive suggestions. Next, we can prompt the LLM with context including (i) the previously generated code and (ii) the constructive feedback, and ask it to use the feedback to rewrite the code. This can lead to a better response. Repeating the criticism/rewrite process might yield further improvements. This self-reflection process allows the LLM to spot gaps and improve its output on a variety of tasks including producing code, writing text, and answering questions. And we can go beyond self-reflection by giving the LLM tools that help evaluate its output; for example, running its code through a few unit tests to check whether it generates correct results on test cases or searching the web to double-check text output. Then it can reflect on any errors it found and come up with ideas for improvement. Further, we can implement Reflection using a multi-agent framework. I've found it convenient to create two agents, one prompted to generate good outputs and the other prompted to give constructive criticism of the first agent's output. The resulting discussion between the two agents leads to improved responses. Reflection is a relatively basic type of agentic workflow, but I've been delighted by how much it improved my applications’ results. If you’re interested in learning more about reflection, I recommend: - Self-Refine: Iterative Refinement with Self-Feedback, by Madaan et al. (2023) - Reflexion: Language Agents with Verbal Reinforcement Learning, by Shinn et al. (2023) - CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing, by Gou et al. (2024) [Original text: https://lnkd.in/g4bTuWtU ]

Hospitals are healing patients faster with 30-year-old Australian technology. Most healthcare facilities still operate in the dark. SolarTube skylights channel natural sunlight through reflective tubes directly into patient rooms and treatment areas. No electricity needed. Just free healing light all day. The healthcare transformation numbers: ↳ Faster patient recovery rates documented ↳ 15% staff productivity increase ↳ Reduced eye strain for medical professionals ↳ Lower patient anxiety during procedures Think about that. Tigoni Medical Center in Kenya installed SolarTubes in their COVID-19 facility. Healthcare workers reported less fatigue, increased alertness during long shifts. Patients showed dramatically improved morale and energy levels. At Rogaska Medical Center, natural daylight flooded clinics without unwanted heat. Staff comfort improved. Patient outcomes followed. Italian dental offices meeting occupational daylight standards found something unexpected: patients felt less anxious. Procedures became more comfortable. Natural light calmed nerves that fluorescent bulbs couldn't. Traditional Healthcare Lighting: ↳ Fluorescent tubes causing eye strain ↳ High electricity costs ↳ Artificial environments ↳ Staff fatigue increases SolarTube Healthcare Reality: ↳ Natural light reduces stress hormones ↳ Serotonin production increases ↳ Circadian rhythms regulate properly ↳ Recovery accelerates naturally But here's what stopped me cold: We're medicating depression while keeping people in artificial light. Jim Rillie invented this solution in the 1980s. Launched Solatube International in 1991. Now 2 million units worldwide bring natural light indoors. Healthcare facilities that adopt it see measurable improvements. Staff wellness increases. Patient satisfaction scores rise. Recovery times shorten. The Multiplication Effect: 1 hospital = hundreds healing faster 100 facilities = thousands of staff energised 1,000 installations = healthcare transformed At scale = medicine working with nature VCC in the UK experienced enhanced well-being building-wide. Staff and patients reported feeling calmer, healthier, happier. Simply from abundant daylight. We're not just installing skylights. We're installing wellness. One beam of natural light at a time. Follow me, Dr. Martha Boeckenfeld for innovations that heal environments and people. ♻️ Share if you believe healthcare should harness nature's healing power.

Exciting updates on Project GR00T! We discover a systematic way to scale up robot data, tackling the most painful pain point in robotics. The idea is simple: human collects demonstration on a real robot, and we multiply that data 1000x or more in simulation. Let’s break it down: 1. We use Apple Vision Pro (yes!!) to give the human operator first person control of the humanoid. Vision Pro parses human hand pose and retargets the motion to the robot hand, all in real time. From the human’s point of view, they are immersed in another body like the Avatar. Teleoperation is slow and time-consuming, but we can afford to collect a small amount of data.  2. We use RoboCasa, a generative simulation framework, to multiply the demonstration data by varying the visual appearance and layout of the environment. In Jensen’s keynote video below, the humanoid is now placing the cup in hundreds of kitchens with a huge diversity of textures, furniture, and object placement. We only have 1 physical kitchen at the GEAR Lab in NVIDIA HQ, but we can conjure up infinite ones in simulation. 3. Finally, we apply MimicGen, a technique to multiply the above data even more by varying the *motion* of the robot. MimicGen generates vast number of new action trajectories based on the original human data, and filters out failed ones (e.g. those that drop the cup) to form a much larger dataset. To sum up, given 1 human trajectory with Vision Pro  -> RoboCasa produces N (varying visuals)  -> MimicGen further augments to NxM (varying motions). This is the way to trade compute for expensive human data by GPU-accelerated simulation. A while ago, I mentioned that teleoperation is fundamentally not scalable, because we are always limited by 24 hrs/robot/day in the world of atoms. Our new GR00T synthetic data pipeline breaks this barrier in the world of bits. Scaling has been so much fun for LLMs, and it's finally our turn to have fun in robotics! We are creating tools to enable everyone in the ecosystem to scale up with us: - RoboCasa: our generative simulation framework (Yuke Zhu). It's fully open-source! Here you go: http://robocasa.ai - MimicGen: our generative action framework (Ajay Mandlekar). The code is open-source for robot arms, but we will have another version for humanoid and 5-finger hands: https://lnkd.in/gsRArQXy - We are building a state-of-the-art Apple Vision Pro -> humanoid robot "Avatar" stack. Xiaolong Wang group’s open-source libraries laid the foundation: https://lnkd.in/gUYye7yt - Watch Jensen's keynote yesterday. He cannot hide his excitement about Project GR00T and robot foundation models! https://lnkd.in/g3hZteCG Finally, GEAR lab is hiring! We want the best roboticists in the world to join us on this moon-landing mission to solve physical AGI: https://lnkd.in/gTancpNK

No company can give you the most exciting, challenging, and impactful project around the year, so the ups and downs in the kind of work you'd get are expected. On average, you would get 4 months of high-impact work, 4 months of moderate work, and 4 months of mundane work in a year. Instead of feeling bad about it, leverage the time and energy to advance your career and secure more impactful projects in the future. Remember, the most important and impactful project is not given to the smartest engineer, but it is given to the one holding the track record of getting things done well and on time. The trust you instill by doing grunt work or maintenance would and should put you in a position to get the most impactful project. Given that the maintenance work does not eat up a lot of your mental bandwidth, use this breathing room to identify more significant problems within your organization. Spend time digging deeper into systemic issues, find a solution, create a project plan, and present it to the leadership. This showcases your initiative and capacity for impactful contributions. When I was given some mundane work, I used the additional time to figure out the root cause of recurring outages and proposed an architecture change to solve it once and for all. Because I was thorough with my homework, it became a no-brainer for leadership to approve. You can also use this time to enhance your visibility and influence within your organization. The easiest of all is to use this time to hold tech talks and mentor early engineers. These interactions will make it more likely that you'll be top of mind when exciting projects are conceived. To be honest, this is what I did at all of my stints across all the companies I have worked at. I gave talks on some of the lessons learned from past projects, new industry trends, or introductions to new technologies. This helped me establish a good reputation within the org. Remember, every phase cannot be exciting, but it is important to know how to leverage it well. Impactful projects will not be served to you on a silver platter, you need to earn it. ⚡ I keep writing and sharing my practical experience and learnings every day, so if you resonate then follow along. I keep it no fluff. youtube.com/c/ArpitBhayani #AsliEngineering #CareerGrowth

12,6 billion euros – BP and Total are paying this record sum to develop 7 GW of offshore wind in Germany. Working in an energy intensive industry and with my background in offshore wind, I am looking at this with very mixed feelings.    Sure, from a political point of view it is a success. For the first time, companies are paying money for the right to develop wind farms in Germany. What was advanced in the form of long-term subsidies in the past is now being paid back. In the long run, however, it is a Pyrrhic victory!   It’s not good news for a low electricity price. Even if only 10% of the bid is to be paid in advance. These immense sums must be earned back. The risk is that this will be to the disadvantage of the customers. Both private and commercial customers will have to deal with higher prices. This poses challenges for energy-intensive industries in particular – and underlines the importance of an industrial electricity price for these industries.   It’s not good news for the supply chain industry. Turbine manufacturers are already under great pressure. At a time when we need a stable European supply chain to shoulder the massive expansion of wind energy, high bids intensify this pressure - especially as qualitative criteria, such as carbon footprint, were not taken into account.   And it’s also not good news for offshore wind in Germany. The cost risks for project developers and operators have become significantly higher than in other (European) countries. It is telling that none of the established offshore wind operators were willing to pay so much money upfront, but instead the projects went to large oil and gas multinationals. Companies that have softened their climate targets one after the other in recent months.

An eagerness to learn is essential for innovation. But the way we learn—and the order in which we partake in various learning activities—can make the difference between effective growth and potential missed opportunities. Jean-François Harvey, Johnathan Cromwell, Kevin J. Johnson, and I studied more than 160 innovation teams and found that the key to faster, clearer progress is: Structured learning 👷🏗️ Our research, published in the Administrative Science Quarterly Journal, highlights four distinct types of learning behaviors used by high-performing teams and examines variations in the sequence and blend of these types of team learning. Without a deliberate rhythm, teams risk becoming overwhelmed by continual information intake, leading to confusion and burnout. But by honing a team's ideal 'learning rhythm,' you can avoid overwhelm and instead focus on strategic decision-making and sustainable innovation. Read our research summary now in the Harvard Business Review: https://lnkd.in/e5nU-Kka

𝗠𝗮𝗴𝗻𝗶𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗼𝘃𝗲𝗿𝘃𝗶𝗲𝘄 𝗼𝗳 𝗮𝗻 𝗲𝗹𝗲𝗰𝘁𝗿𝗶𝗰𝗮𝗹 𝘀𝘂𝗯𝘀𝘁𝗮𝘁𝗶𝗼𝗻 Substations are used at the generation, transmission, and distribution levels. Generators (at various power plants) generally produce electricity at lower voltages. However, these lower voltages are not efficient for long-distance transmission primarily due to technical losses (such as power loss (I^2*R) or voltage drops). This is because the current is higher at a lower voltage for the same amount of power transmitted. This contributes to huge losses (I^2*R), where "I" is the load current and "R" is the line's resistance. A transmission substation is used to step up the generation voltage for long-distance delivery to reduce losses. Most power generation facilities are located far from customers (homes, businesses, and commercial or industrial electricity consumers). A transmission line length is considered: ✅ Short if it's less than or equal to 𝟱𝟬 𝗺𝗶𝗹𝗲𝘀 (𝗼𝗿 𝟴𝟬 𝗸𝗺).  ✅ Medium if it's greater than 𝟱𝟬 𝗺𝗶𝗹𝗲𝘀 (𝟴𝟬 𝗸𝗺) but less than or equal to 𝟭𝟱𝟬 𝗺𝗶𝗹𝗲𝘀 (𝟮𝟰𝟭 𝗸𝗺) ✅ Long if it's greater than 𝟭𝟱𝟬 𝗺𝗶𝗹𝗲𝘀 (𝟮𝟰𝟭 𝗸𝗺) The distribution substation takes the power from a transmission or sub-transmission substation and further steps down the voltages for distribution. For instance, a solar PV power plant is a generator. An inverter(s) is/are needed to convert the DC power from the solar panels to AC power before injecting it into a distribution or transmission network. Let's assume the expected power to be delivered is 2 MVA, and we have one central inverter at 600 V. The load current (I) at 600 V will be (𝟮 𝘅 𝟭𝟬^𝟲)/(𝟭.𝟳𝟯𝟮*𝟲𝟬𝟬) = 𝟭𝟵𝟮𝟱 𝗔. For simplicity, let's assume a conductor resistance of 0.5 ohms (keep constant) Power loss = 𝟭𝟵𝟮𝟱*𝟭𝟵𝟮𝟱*𝟬.𝟱 = 𝟭,𝟴𝟱𝟮,𝟴𝟭𝟮 𝗪 A load current of 1925 A is large, so we must buy large conductors and associated support systems to transport the 2 MVA apparent power. The technical losses and voltage drops at this current are significant and uneconomical. A transformer is used to transform the 600 V to say 34,500 V, and the current at such medium voltage will be: (𝟮 𝘅 𝟭𝟬^𝟲)/(𝟭.𝟳𝟯𝟮*𝟯𝟰,𝟱𝟬𝟬) = 𝟯𝟯 𝗔 and power loss 𝟯𝟯*𝟯𝟯*𝟬.𝟱 = 𝟱𝟰𝟱 𝗪 Same power, but now, we have a smaller load current to evacuate through a distance. For long distances and larger power, it's even more economical to step up the 34,500 V to a transmission level, say 115,000 V. At 115,000 V, the transferred current is further reduced to: (𝟮 𝘅 𝟭𝟬^𝟲)/(𝟭.𝟳𝟯𝟮*𝟭𝟭𝟱,𝟬𝟬𝟬) = 𝟭𝟬 𝗔.  and power loss is 𝟭𝟬*𝟭𝟬*𝟬.𝟱 = 𝟱𝟬 𝗪 These assumptions give a better perspective on the discussion. But remember that an increase in voltage will require you to consider factors such as increasing the cost of equipment insulation. A lot happens between these systems, so it can't be explained in this limited space. This is just an overview. 📹 Surdu Alexandru Andrei

Struggling teams don't need another framework. They need a leader. I've taken over bad teams filled with good people. I learned to embrace three themes for a successful reset: ✅ Change requires honoring the past and building the future ✅ Trust is rebuilt through actions, not just words ✅ Culture lives in daily micro-decisions Here are the 8 lessons that make it work: 1/ Honor the Past ↳ Don't play the blame game ↳ Value those who stayed through hard times 2/ Name What Stops Here ↳ Be specific about what changes ↳ Get them to help rewrite the new rules 3/ Own Your Role ↳ Acknowledge where you fell short ↳ Build trust through self-accountability 4/ Reset the Target ↳ Paint a clear 6-month vision ↳ Define what excellence looks like 5/ Define Winning Behaviors ↳ Skip empty corporate speak ↳ Make expectations crystal clear 6/ Create New Rituals ↳ Build sacred team habits ↳ Engineer connection, especially remote 7/ Embrace Iterations ↳ Progress isn't linear ↳ Celebrate small wins, learn from setbacks 8/ Rebuild Trust Daily ↳ Start from trust at zero ↳ Do what you say you'll do 9/ Catch Them Winning ↳ Be specific about what you see ↳ What gets recognized gets repeated Want more detail?  Flip through the full playbook below. Remember:  Your team likely knows the path forward. They're just waiting for you to walk it first. If this was helpful: 📌 Please follow Dave Kline for more ♻️ Share to help other leaders turn things around.