Daily Tech Brief · September 16th
From storyflo. This is your daily audio brief. It's Theo. September 16th, tech roundup — five stories, here's number one. Let's get into it. First, from DevClass. Apple iPhone Duo makes developers think in folds.
From storyflo. This is your daily audio brief. It's Theo. September 16th, tech roundup — five stories, here's number one. Let's get into it. First, from DevClass. Apple iPhone Duo makes developers think in folds.
The October 23 ship date of Apple’s iPhone Duo grows nigh, so developers wanting to optimize their iPhone apps for the foldable will need to get to grips with its new display behaviors and form factor. “A fold is a UI event with backend consequences,” Atharva Deosthale, Appwrite developer advocate, wrote in an introductory blog post.
TypeSafe AI just launched a new model called Jev, designed specifically for machines, not humans. Unlike traditional AI that responds with natural language, Jev provides structured, probabilistic decisions that software can easily process. This is a big deal because it sidesteps the usual parsing issues that come with text responses, making it faster and more efficient for tasks like customer service or even playing Doom. Jev operates on a unique architecture that allows it to return decisions in 70 to 500 milliseconds, significantly quicker than existing models. Plus, it’s much cheaper to use, costing a fraction of what other models do. The idea behind Jev is to create a tool that can reliably handle complex workflows without the risk of generating misleading information, which is crucial for applications that require quick, accurate decision-making. It’s fascinating to see how this could reshape automation and AI interactions in various industries.
There’s now a sitcom pilot produced entirely with AI. Of course there is. Oscar-nominated Sci-fi filmmaker Neill Blomkamp recently tried something similar, and it was awful. Now, according to the Guardian, some former startup investor named Andrew Dickinson has given it a shot, and it’s awful. Dickinson reportedly wrote the script in the hopes of getting it filmed—like with human actors. With that in mind, watching this show, called No Big Deal, will feel familiar if you’ve ever read someone’s sitcom pilot and been asked to give notes.
I’ve been thinking about how the engine under Whiteout Survival quietly shifted. GPT‑6’s Astra model can actually read a screenshot, map the 3‑D layout, and suggest concrete level structures—something that used to need a designer’s eye.
At the same time Unity rolled out a Claude Code plugin that lives inside the editor. It can spin up objects, tweak scenes, write C# scripts, and chase down compile errors without you opening a separate IDE.
Put those together and you get a loop: Astra sketches the world, Claude fills in the code, Unity packages it for iOS. The workflow skips the months‑long hiring and iteration that a traditional studio would need.
The upshot isn’t “type a game and get a million dollars,” but you can now prototype a full‑featured mobile title without ever having been a game dev. That’s why the $41 million month feels suddenly reachable.
AI is evolving to be more flexible, allowing agents to adapt and create tools on the fly. This shift includes test-time training, where models can update themselves while in use, enhancing their memory, strategy, and even the criteria for what constitutes improvement. However, a critical question emerges: what boundaries should remain firm? Meta’s Muse exemplifies this by maintaining control over permissions and data access, even as it improvises and builds new capabilities.
The conversation around recursive self-improvement is heating up, but it’s essential to clarify what makes an improvement truly beneficial. While no AI can fully self-improve yet, researchers are exploring various aspects of this concept, working on pieces like strategy and memory enhancements. Test-time training is a significant step forward, enabling models to adapt in real-time, which is particularly valuable for agents facing unpredictable situations. This week’s discussions highlight the delicate balance between empowering AI and ensuring it operates within safe limits.
Insiders are claiming that OpenAI may soon receive a funding round valuing it at $1.2 trillion, according to the Financial Times. The intriguingly timed round could give OpenAI a higher valuation than its rival Anthropic once again, though it could also come around the same time as Anthropic’s IPO. The decision to give OpenAI this capital injection “will hinge on exactly when OpenAI decides to go public,” the Financial Times writes. In an interview published in Fortune this past weekend, OpenAI CEO Sam Altman said the company would not go public in 2026.
A new post on Apple’s Security Research blog details the fascinating tech, architecture, and thinking behind the iPhone 18 Pro’s new Reference Image camera mode. Here are the details Apple explains why it developed the new Reference Image camera mode In a blog post titled “Apple Reference Image: A New Approach for Verified Photography,” Apple goes in-depth into how its new photographic authenticity system works. The post confirms that the company’s motivation to build Apple Reference Image was the ever-lowering barrier to creating or adjusting images synthetically.
All three devices come with an upgraded octa-core processor for better performance and now run on Android 16. As always, the new Boox devices have access to the Google Play Store, plus they'll have Bluetooth 6.0, and Boox's BSR refresh technology for a smooth display with decreased ghosting. Here's what we know about the three new devices. The Boox Note Air6 C is available now, while the Mini C and Palma 3 are coming soon. All devices will be available for purchase from Boox and Amazon.
[Clough42] has started a new CNC control cabinet build, and uses it as an opportunity to demonstrate why he models the whole thing out in CAD before ordering parts or physically building anything. One may wonder why to bother, and the simple answer is to ensure there are no surprises or waste. [Clough42] has built plenty of these cabinets and there’s always something that isn’t as expected. Modeling out an entire cabinet sounds like it should be easy, because today one can obtain 3D models for components from industrial suppliers with ease.
Seattle-based Interlune has raised $5 million in fresh funding to support a business plan that involves building space infrastructure as well as mining the moon for helium-3 and other valuable resources. The funds were raised through a Simple Agreement for Future Equity, or SAFE, in which investors provide upfront cash to a startup in exchange for shares to be issued during a future priced round. Interlune used the same approach to raise $5 million earlier this year.
Send this story to anyone — or drop the embed into a blog post, Substack, Notion page. Every play sends rev-share back to storyflo · tech.
We’ve simplified responses to 👍 / 👎. Past comments are archived but no longer visible.