OpenAI plans to release its next big AI model by December (2 minute read)
OpenAI plans to launch its next frontier model, Orion, by December. The model has been teased by an OpenAI executive as potentially up to 100 times more powerful than GPT-4. Orion will initially be released to companies that OpenAI works closely with so they can build their own products and features. Engineers inside Microsoft are preparing to host the model on Azure as early as November. It is unclear if OpenAI will call the model GPT-5, despite it being seen internally as GPT-4's successor.
TSMC's Arizona Chip Production Yields Surpass Taiwan's in Win for US Push (3 minute read)
Taiwan Semiconductor Manufacturing Co. (TSMC) has achieved early production yields at its first plant in Arizona that surpasses similar factories back home. The share of chips manufactured at the facility that are usable is about four percentage points higher than comparable facilities in Taiwan. The yield is critical as it determines whether companies will be able to cover the enormous costs of a chip plant. The accomplishment is a sign of progress for US efforts to revitalize local semiconductor manufacturing.
How Intel Got Left Behind in the A.I. Chip Boom (10 minute read)
The story of how Intel got left behind in AI is representative of the broader challenges the company now faces. The company missed opportunities, made wayward decisions, and executed poorly. The trail of missteps was a byproduct of a corporate culture born of decades of success and high profits that was hard-changing and focused on its franchise in personal computers and later in data centers. Its corporate ethos worked against it as the company tried and failed repeatedly to become a leader in chips for artificial intelligence.
New Architecture is here (22 minute read)
React Native 0.76 with the New Architecture by default is now available on npm.
Google plans to announce its next Gemini model soon (10 minute read)
Google is aiming to release its next major Gemini model in December. Gemini 2.0 will be widely released at the outset as opposed to being rolled out in phases. While the model isn't showing the performance gains experts had hoped for, it will likely still have some interesting new capabilities. It appears that the top AI developers will continue to race to release ever-bigger and more expensive models even as performance improvements start leveling off.
A proposal to split JavaScript into two languages has been presented to Emca TC39, the official standardization committee. The proposal argues that the foundational technology of JavaScript should be simple because security flaws and the complexity cost of the runtimes affect billions of users. New language features only benefit developers and applications that actually use those features to their advantage - adding them almost always worsens security, performance, and stability. The proposal suggests changing JavaScript's approach to one where most new features are implemented in tooling rather than in the JavaScript engines.
A mysterious new image generation model has appeared (1 minute read)
red_panda is a mysterious new image generation model that is beating other top models. The model is over 100 times faster than OpenAI's DALL-E3. It is unknown which company made the model and when it will be released. AI companies are increasingly using community benchmarks to drum up anticipation ahead of announcements, so it is likely a reveal will be coming soon.
The Battery Revolution Is Finally Here (12 minute read)
Solid-state cells, an emerging, ground-breaking technology, are already hitting the market. These batteries solidify the traditionally liquid electrolyte, enabling advantages such as higher energy density and lower cost and weight. There are many different types of solid-state cells. This article looks at several of the competitors in this industry and their approaches to developing solid-state batteries.
GitHub Spark lets you build web apps in plain English (3 minute read)
GitHub Spark allows developers to quickly build small web apps using nothing but natural language.
Dropbox cuts its workforce by 20 percent in latest round of layoffs (1 minute read)
Dropbox is laying off 528 people amid slowing growth for its core cloud storage business. The company is making significant cuts in areas where it is over-invested or underperforming while designing a flatter, more efficient team structure overall. Affected employees will receive sixteen weeks of severance pay, equity, bonus plan lump sums, payouts of approved leave, and immigration consultation for those on work visas. Dropbox reported an increase of 63,000 paid users quarter over quarter for its second-quarter earnings this year, light compared to its total 18 million-plus user base.
Automattic has offered two rounds of severance deals - 159 employees out of roughly 1,900 employees took the first deal.