This post contains an interview with Andrew Atkinson, a PostgreSQL expert at StepChange, which discusses his experience scaling Rails applications for millions of users. Atkinson shares valuable lessons learned from his work at Microsoft, where he optimized the infrastructure for Flip, a video discussion app. He highlights the challenges of scaling a Rails application with Postgres, particularly when dealing with heavy read and write operations, and provides practical advice for developers on optimizing queries, identifying bottlenecks, and utilizing tools like PG stats and explain plans. The post encourages developers to leverage ActiveRecord's features like counter caches and sharding to improve performance and scalability.
Microsoft has donated the Mono Project, an open source implementation of the .NET Framework, to the Wine team, and has acknowledged its significance in enabling cross-platform .NET development.
Whatever you want to build in the cloud, Vultr [4] is the easiest way to get started. Whether it's Wordpress websites, multi-tenant web services, or Elasticsearch clusters — Vultr has you covered:
OpenAI is reportedly planning to launch a new AI as part of a chatbot this fall. Codenamed Strawberry, the AI has advanced mathematical reasoning, programming, and other skills that allow it to answer questions on more subjective topics, like marketing strategies. It can be used to generate high-quality synthetic training data for training large language models. The model could help OpenAI obtain the data it needs to train the GPT-4's successor.
SpaceX's Starlink system for mobile phones will be free for users connecting to emergency services. The company is currently waiting for approval from the Federal Communications Commission to let it commercially operate its 'direct to cell' Starlink technology. Providing free emergency services access would benefit public safety communications and first responders. There aren't many details about how Starlink's system will work, but it will likely be similar to Apple's Emergency SOS feature for iPhones, which is free for the iPhone 14 and up.
SpaceX's Polaris Dawn mission has been delayed until later this week at the earliest due to a helium leak. The mission will send four private citizens (billionaire entrepreneur Jared Isaacman, retired Air Force Lt. Col. Scott “Kidd” Poteet, and SpaceX engineers Sarah Gillis and Anna Meno) to space on a mission that will include the first spacewalk carried out by a civilian crew. It was scheduled to lift off early Tuesday. Polaris Dawn is the first of three planned space flights funded and organized by Isaacman and SpaceX.
Digital computing is reaching relative maturity, with traditional electronics well into the final phase of their S-curve and GPUs in the middle of theirs. The demand for compute is showing no sign of slowing. Next-gen approaches will need to harness entirely different materials systems with more amenable physics. This article looks at some of the possible technologies that may bring the industry to the next level.
Vanilla Prime is a spiritual successor to VANILLA TODO with no frameworks, no bundlers, and no required dependencies. It aims to provide a simpler, more sustainable way of web development. While it is less ergonomic than a fully integrated framework, it produces websites that are performant, lightweight, independent, hackable (in a good way), and respects users' and the world's resources. Vanilla Prime can be difficult to get into and hard to master, but the outcomes can be incredibly rewarding.
There are significant changes happening with distributed systems. These changes will influence how systems are operated and how they are programmed. This article shares insights into the changes in transactional and analytical systems, especially around object storage and programming models. There are many possible disruptive technologies, so it is challenging to pick the winners and losers.
Of the 208 startups currently listed in Y Combinator's S24 startup directory, 156 are working on AI-related products. Nearly half of all US venture capital investment went to AI companies last quarter. It is unclear whether Y Combinator is deliberately selecting startups that are working in the space, if more founders are choosing to work in AI, or if startups are just saying that their product is AI-related because they think that's what investors or customers want to hear. Only 33 Y Combinator startups were related to crypto and Web3 in 2022, when the technology was the hottest thing to work on.
While Lyft has been working on innovative features for riders, Uber's focus in the US has been on supply. With self-driving cars, the platform power shift will be in the direction of who controls the demand - most people who own self-driving cars will rent out their cars to the platform that has the most demand for it. Uber will have the most types of vehicles available and more flexibility in terms of how much it can charge. It owns the most demand worldwide today, and self-driving will make it even more dominant.
You don't need to put your career on pause to get a high-quality education. Browse 200+ affordable, online degree programs at SNHU — an accredited, nonprofit institution. Start here (US learners only)
Starliner's failure is only the tip of the iceberg - Boeing's contributions to NASA's Moonbound Space Launch System are many years behind and way over budget.
Stack Overflow has been a cornerstone of developer communities for years, but it seems to be losing its prominence due to several issues, including over-moderation, a toxic community, and how difficult it is for new users to start out on the platform.
OpenAI is planning to release a new AI product called "Strawberry" in the fall. It will feature advanced reasoning capabilities, such as the ability to solve previously unseen math problems, and can perform high-level tasks like developing market strategies.
The 3rd part of a series, this paper explores scaling laws and how many bits are required to store knowledge in a model. The answer seems to be around 2 bits of knowledge per parameter.
Boson AI has introduced Higgs-Llama-3-70B-v2, a new model that excels in dialog and comprehension benchmarks like Arena-Hard and AlpacaEval 2.0. The model reduces response regeneration rates by 21.6% and boosts day 1 retention by 5.3% compared to Claude 3.5 Sonnet. Enhanced by an in-house reward model, Higgs Judger, it ties with Google's Gemini 1.5 Pro on performance.
Customers don't always use the words you expect them to, which tends to break keyword-based solutions for knowledge base and in-product search. Luckily, genAI does much better. Make your help center actually helpful with Ask AI's Generative Search [13]. Minutes to
Pre-training hybrid (Mamba style) models is different from pre-training normal Transformers. This post explores how to scale different hyperparameters, data acquisition, and others to get the performance you want.
This is a framework that fine-tunes small models to serve as a fallback when closed API models go down. It shows how to migrate from a large to a small model smoothly.
LitServe is an easy-to-use, flexible serving engine for AI models built on FastAPI. Features like batching, streaming, and GPU autoscaling eliminate the need to rebuild a FastAPI server per model.
Llava BitNet is the first ternary (-1, 0, 1) weight model trained on VLM tasks. The model, weights, and scripts are in the process of being fully open sourced. The technical report will be released soon and suggests the model has promising performance.
While large language models (LLMs) may not significantly improve in reasoning capabilities, their decreasing costs and increasing speeds will make them more useful for repetitive tasks. Although these models may lack true understanding, they can still efficiently handle straightforward tasks.
Using a number of models in sequence, this group was able to create an amazing model that generates fully playable 3D game scenes based on a single input sketch.
Meridian is the Stellar Development Foundation's annual conference — and this year's London event [4] is a must-attend for anyone in the blockchain space.
Tether is investing its massive profits into AI and neural implant companies with assistance from controversial tech investor Christian Angermayer. It has already spent $1.5 billion on majority stakes in Northern Data, a German data center operator, and brain-computer interface maker Blackrock Neurotech. These deals mark an expansion of Tether's business from stablecoins, where it earns $4 billion annually, into new technologies, though Angermayer has a questionable track record on past deals.
Toncoin and Ethereum have led a downturn in the crypto market, with Toncoin seeing a 20% decline following the arrest of CEO Pavel Durov. The crypto market overall saw over $144M in long liquidations within 24 hours, heavily impacting these two cryptocurrencies amid broader market corrections.
MakerDAO has rebranded to "Sky" and is set to launch a new stablecoin, USDS, and a governance token, SKY, on September 18, marking a large overhaul in its operations. The rebranding includes a new UI that aims to make the platform more approachable and expandable with a metaDAO ecosystem called "Sky Stars" for decentralized project governance. While the existing DAI and MKR tokens will remain, they can be voluntarily upgraded to the new tokens.