Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Nvidia will be leaning more heavily into robotics in 2025. It will launch a new generation of compact computers for humanoid robots called Jetson Thor in the first half of next year. Nvidia aims to provide robot makers with an underlying OEM as its largest customers work to lessen their dependence on the company's AI chips by developing their own.
Blue Origin successfully performed a hotfire of its New Glenn rocket on December 27, the same day it was issued a commercial space launch license from the US Federal Aviation Administration. The company remains tightlipped about when it will launch its rocket, but it had claimed earlier in the year to be targeting a launch before the end of the year. New Glenn was designed to be partly reusable. Its first stage, which is designed to be good for a minimum of 25 missions, will land on a downrange sea-based platform.
Commonwealth Fusion Systems has announced plans to build the world's first nuclear fusion power plant. The plant will be located in Chesterfield County, Virginia. If successful, the plant will infuse 400 megawatts power into the state's electrical grid starting in the early 2030s. Fusion technology still hasn't been proven to be viable - previous attempts by well-financed tech companies have failed.
China plans to start building a demonstration thorium-based molten-salt reactor in the Gobi Desert next year. The 10-megawatt reactor project is scheduled to be operational by 2030. Thorium-232 is more abundant on Earth than uranium-235, but it isn't directly fissile, so it must first be transformed into fissile uranium-233, which is technically feasible, but it is unknown whether that is economical and practical. If feasible, thorium, which is waste product of China's huge rare earth mining industry, could reduce dependence on uranium.
SQLite is the most deployed and used database. It is likely used more than all other databases combined. SQLite is only maintained by three people who don't allow outside contributions. This blog post lists 24 facts about the database.
A dependency consists of a checksum, a location, a name, and a version. The checksum is a cryptographic hash of the contents of the dependency. The location tells systems how to acquire the dependency. The name tells when two different dependencies correspond to different versions of the same thing. The version is a specific rule about dependency substitutability.
Gene drives are a clever genetic trick that forces a particular gene to be inherited by nearly all offspring rather than the usual 50-50 chance in normal inheritance. They can propagate a desired mutation through an entire population theoretically with only a few mating events. Some iterations of gene drives have been demonstrated to work in fruit flies, but mosquitoes are much more complicated and we don't fully understand them yet. This technology is also being developed in countries where malaria is not endemic - the countries that have problems with mosquitoes may not want to deal with the potential ecological issues that may result from the use of gene drives.
The modern web is incredibly complex. WebAssembly made web technology the core of the future of computing. We are already using the browser as a replacement for operating systems' graphics toolkits. Simplifying the web won't fix many of the widespread problems. There are a lot of smaller networks currently available that provide a look into what an alternative to the web might look like.
Run:ai offers GPU orchestration software that organizes groups of GPUs more efficiently.
Founders are always selling - sales is a skill that anyone can get good at.
Researchers from Johns Hopkins University and Stanford University have trained robots to manipulate needles, tie knots, and suture wounds using video recordings of human-controlled robotic arms performing surgical tasks.
Passkeys were supposed to fix the problems with passwords, but their inconsistent implementation has made them less practical than password managers.
A look over a year of writing by Anu Atluru, a founder and writer from NYC.
People with high salaries and net worth tend to live longer lives due to access to better healthcare, nutrition, and safer environments.
Chinese AI startup DeepSeek has released DeepSeek-V3, a 671B parameter model employing a mixture-of-experts architecture, available on Hugging Face. DeepSeek-V3 has outperformed leading models like Meta's Llama 3.1 and rivals closed models such as OpenAI's GPT-4o. The model emphasizes efficient performance with innovations like multi-token prediction, offering substantial training cost savings.
Microsoft and OpenAI define AGI as AI systems generating $100 billion in profits, which OpenAI is far from achieving. OpenAI is reportedly losing billions and doesn't expect profitability until 2029, affecting how long Microsoft will have access to its technology. Speculation on OpenAI declaring AGI sooner is countered by these financial metrics.
OpenAI is considering building its own humanoid robot, leveraging previous investments in robotics firms like Figure and 1X. The company disbanded its robotics division in 2021 and would face challenges re-entering this competitive market.
This paper introduces a new approach to Super-Resolution (SR) that challenges the conventional method of training separate models for each scale.
SceneCraft introduces a method for creating detailed 3D indoor scenes based on user-provided text descriptions and layout preferences.
Label Critic is an innovative tool designed to streamline medical dataset annotation by using AI-generated labels instead of starting from scratch.
CHOTA metric (Cell-specific Higher Order Tracking Accuracy) improves the evaluation of cell tracking methods in biomedical research. Unlike current metrics, which focus on local accuracy, CHOTA unifies the assessment of cell detections, global coherence, and lineage tracking, making it more effective for high-level biological analysis.
This repository, along with the companion paper, contains a list of services, models, datasets, and systems used to generate music.
Meta's COCONUT is a novel approach that allows LLMs to reason in continuous latent space rather than discrete language tokens by encoding reasoning steps in continuous vectors. The method improves reasoning abilities but sacrifices interpretability. It could be a valuable addition to future LLMs despite the trade-off.
Tenstorrent's open-source AI hardware approach offers a promising alternative to Nvidia's dominance by using a unique CPU and AI core integration strategy. The company is leveraging Samsung Foundry's cost-effective SF4X process and aims to tackle latency issues for scaling out AI workload efficiently. Tenstorrent's recent $2B valuation highlights its potential, especially as an option for high-performance RISC-V IP amidst ARM's pricing pitfalls.
Waymo's self-driving ride-hailing service has expanded to Los Angeles, joining San Francisco and Phoenix. Riders appreciate the smoother and more private experience compared to traditional rideshares. However, while ridership is increasing, Waymo's profitability remains uncertain.
ChatGPT Search, a new AI-powered search engine, can be tricked into generating misleading summaries by hiding text in websites.
Meta's Ray-Ban Smart Glasses now feature live AI assistance, language translation, and Shazam music identification.
SAPNet is a neural network developed by Italian scientists to enhance spectral data analysis from handheld Raman spectroscopy devices.
Google plans to introduce a number of AI features in the first half of next year. The company recently held a strategy meeting for next year where executives set the stage for a year of increased competition, regulatory hurdles, and advancements in AI. CEO Sundar Pichai has warned employees that 2025 will be a challenging year as the company is facing scrutiny across the world. While Google's search business still has dominant market share, generative AI has brought with it a host of new competitors.
OpenAI will create a public benefit corporation to oversee commercial operations when it moves towards a for-profit structure in 2025. The public benefit corporation will remove some of OpenAI's nonprofit restrictions and allow it to function more like a high-growth startup. The scale of capital that the company needs requires conventional equity and a less complicated structure. OpenAI's effort to restructure faces major hurdles, including a lawsuit from Elon Musk and an outflow of high-level talent.
Humans think at a rate of 10 bits per second. Our bodies' sensory systems gather environmental data at a rate of a billion bits per second, 100 million times faster than our thought processes. It is unknown why humans are only able to think a single thought at a time rather than many at a time in parallel the way our sensory systems work. One theory is that the brain was first evolved for navigation - it would make sense that we can only follow one 'path' of thought at a time if our brain evolved from a simple system to follow paths.
Space exploration dazzled the world in 2024 - NASA's Europa Clipper began its journey to study Jupiter's moon Europa, SpaceX's Starship achieved its first successful landing, China's Chang'e 6 mission successfully returned samples from the far side of the Moon, and the International Space Station continued to host international crews and private missions. Space agencies are gearing up for even more ambitious goals in 2025. This article looks at the most exciting missions planned for the coming year. These include CLPS, an initiative from NASA to deliver science and technology payloads to the Moon using commercial landers; SPHEREx, a mission that will survey the sky in near-infrared light to create a comprehensive map of the universe; Space Rider, and uncrewed spaceplane from the European Space Agency; and Japan's M2/Resilience mission, which will launch a lander and micro-rover to the lunar surface.
Software designers usually have a simple concept in mind when writing programs - understanding this concept allows them to find sections of programs that must be altered when an update or correction is needed. Changes made by people who don't understand the original design concept almost always cause the structure of the program to degrade. The changes will be inconsistent or invalidate the original concept. After many such changes, nobody will understand the modified product. Software that has been repeatedly modified in this way becomes very expensive to update.
It's probably easier to get to staff engineer by getting hired as doing so via promotions requires you to be known and valued by your organization. Companies value engineers who ship, so ship projects that the company cares about and don't spend time on pet projects. Don't worry too much about mentoring and leading, as long as you're leading projects successfully, or if doing so delivers value for the company. Getting to staff manager requires a lot of work from your manager - don't invest time in the promo packet until they're fully on board.
All small and medium-sized portable electronic devices (under 100 watts) sold in the EU must now be compatible with USB-C ports. The rule will also apply to laptops starting April 28, 2026. Apple removed the iPhone 14, iPhone 14 Pro, and third-generation iPhone SE from its EU online stores to comply with the regulation earlier this month. Other major markets, such as California and India, are following the EU's lead and will implement similar rules in the coming years.
This article talks about the implications of labor-replacing AI and the possible future that we might be facing. Labor is the human mental and physical effort that produces something of value. Capital goods are things that are used in the production of goods and services. Capital can refer to both the stock of capital goods and the money that can pay for them. AI will make capital a more general substitute for labor. There will be less need to pay humans for work as you can replace that with capital - data centers running software.
This person tracked their default apps throughout the year to see how their app use changed.
This Hacker News thread contains a discussion on the state of employment for software engineers and the effect that offshore/immigrant workers are having on the industry.
Running large language models locally gives you total control over your data.
Listing the things that you found joy in throughout the year can be a cathartic exercise.
A list of people who are going to change the world, along with links to their work and X accounts.
A study analyzing 62 queries compared ChatGPT search and Google, highlighting strengths and weaknesses. Google led in informational, local, and commercial queries, while ChatGPT search showed promise in content gap analysis and disambiguation tasks. Both platforms had issues with errors, omissions, and incomplete responses, but Google generally provided more reliable search results.
The EU-U.S. 6G-XCEL project and joint efforts like ACCoRD and COSMOS are advancing 6G research through collaborations on AI-integrated network architectures. Workshops at Rutgers highlighted innovations in 6G technology and emphasized open-source initiatives and industry partnerships. These initiatives aim to accelerate progress and create interoperability frameworks for the next generation of wireless networks.
Google acquired Character AI primarily for its cost-effective inference technology, which enabled scalable AI interactions. This efficiency allows Google to offer AI models for free via AI Studio without harming unit economics. The acquisition complements the trend toward inference-time optimizations as pre-training returns diminish.
MovieChat introduces an approach to understand long videos by merging video foundation models with large language models using a zero-shot learning method.
The Chordonomicon dataset offers over 666,000 songs with chord progressions annotated by genre, structure, and release date, filling a gap in deep learning resources for music.
ClassiFIM is a new approach for estimating the Fisher Information Metric in unsupervised learning of phase transitions.
An AI-powered hedge fund that uses multiple agents to make trading decisions.