1.
Jensen's purchase of a new toy could reshape the entire AI industry
2.
Thailand pauses all datacenter builds and approvals
ASIA IN BRIEF Thailand’s National Economic and Social Development Council last week decided to pause all datacenter builds in the country, while it works out appropriate regulations. The Council announced the pause last Friday in a communiqué issued after a meeting of its Datacenter Business Policy Committee, at which Thailand’s Prime Minister and Interior Minister Anutin Charnvirakul said datacenters are an important part of the nation’s economy – but admitted his government doesn’t have good data about how many of the facilities currently operate, or the pipeline for new builds.
3.
Might as well blame AI for this giant Linux release candidate, says Linus Torvalds
4.
This AI-generated iPhone 18 Pro mock ad makes me actually love dark cherry color
Can you believe it’s new iPhone week already? Apple’s “Rise and Surprise” event is just a few days away on Wednesday, September 9, at 10 am PT. The company is set to unveil the foldable iPhone Ultra and new iPhone 18 Pro and iPhone 18 Pro Max, among other new hardware products. The interesting new color for the Pro phones this year is rumored to be a dark cherry red finish.
5.
Roku's 24/7 AI Slop Channel Is Even Worse Than Expected
A new streaming TV channel shows films made with genAI, reports Engadget. "Fairground AI Creator TV" is free — and supported with ads — describing its material as "AI Cinema": "AI-generated" can make it sound as though someone typed a sentence into a machine and came back five minutes later to find a finished movie. Fairground's catalog shows why that description can be too simple. Take Lost Garden: The Awakening of the Lantern Knight. According to its Fairground page, creator Frank Houbre wrote the world, characters, mythology, emotional arc and screenplay himself.
6.
Inside the Book: Graph Engineering for Agentic AI Systems, Chapter by Chapter
As you move through the chapters, it becomes clear that understanding the anatomy of these graphs is crucial for effective design. There’s a strong focus on practical applications, like how to handle shared state and the implications of adding more agents to a system. The author even lays out a pattern language for common topologies, which could be a game changer for anyone shipping AI systems.
What I found particularly interesting is the way the book stresses the importance of not just following frameworks blindly, but rather understanding how to map your unique topology onto them. It’s all about making informed choices that can prevent costly mistakes down the line. If you’re into AI or systems design, this could be a valuable resource to explore.
7.
GPT-6 Astra just completed Portal, but it burned through $571.18 in tokens to do it
About a year and a half ago, I watched Claude try to beat a Pokémon game on the Twitch channel ClaudePlaysPokemon. Back then, Claude kept getting stuck on things, and it was mostly frustrating to watch him bumble around and undo progress. Now, GPT-6 Astra has shown how AI has evolved by beating Portal by itself, although it had to churn through $571.18 in API calls to do it.
8.
Stop Uploading PDFs to LLMs to Save Thousands of Tokens: Do This Instead
This week, I opened a one-page setup guide from 100xEngineers about a Microsoft project called MarkItDown. And it sent me down a rabbit hole. Most of us have the same habit. We drag a PDF into an LLM, type a prompt, then hope it can make sense of a file that was designed for a human eye, not a language model. I used to treat a PDF like clean text. It isn’t always. A PDF can carry page layout, repeated headers, columns, tables that fall apart when extracted, scanned images, and a lot of stuff you never meant to put in the conversation. For a short, text-heavy document, that might be fine.
9.
Java News Roundup: TornadoVM 6, JReleaser, LangChain4j, Java Operator SDK, JHipster, Yupiik Fusion
TornadoVM 6.0 just hit general availability, and it’s pretty interesting how it’s enhancing Java’s ability to leverage GPUs for parallel processing. This means developers can tap into some serious performance boosts for compute-heavy applications without diving into complex native code.
JReleaser and LangChain4j both got point releases, which usually means some nice tweaks and bug fixes to improve stability and usability. It’s always good to see these tools getting regular updates, keeping them fresh and functional.
The Java Operator SDK and JHipster also made strides with their own updates, which could streamline workflows for building cloud-native applications. Plus, Kotlin Toolchain and Yupiik Fusion are stepping up with maintenance releases, ensuring everything runs smoothly.
Lastly, Micronaut and GraalVM Development Kit are in the mix with maintenance updates too, which is crucial for keeping those frameworks robust and reliable. It’s a busy time in the Java ecosystem, and these updates reflect a commitment to continuous improvement.