Author: admin

On this week’s episode of The MacRumors Show, we discuss Apple’s apparent plan to launch a low-cost MacBook powered by an iPhone chip.

Earlier this week, Apple analyst Ming-Chi Kuo reported that Apple is set to launch an all-new “affordable” MacBook powered by an iPhone chip. The machine is expected to feature a 13-inch display, the A18 Pro chip, and color options that include silver, blue, pink, and yellow. MacRumors first spotted evidence of such a device in backend code related to Apple Intelligence last summer, and subsequently confirmed its use of the A18 Pro chip. The machine features the identifier “Mac17,1.”

This would be the first Mac powered by an ‌iPhone‌ chip. To date, all Apple silicon Macs have contained M-series chips, which offer higher core counts, support for larger amounts of memory, and better external display support. The A18 Pro chip debuted in the iPhone 16 Pro last year.

With the A18 Pro chip, the device is highly unlikely to feature Thunderbolt ports, more than 8GB of unified memory, and support for more than one external display. In line with how Apple tends to handle its other low-cost devices, the new MacBook could re-use the design and chassis of an older machine like the M1 MacBook Air to keep costs down and differentiate it from the MacBook Air. Apple could also revive the simple “MacBook” moniker, separating it from the MacBook Air and MacBook Pro, and mirroring the iPad lineup, which simply has the iPad as the entry-level model.

According to Kuo, the new MacBook is expected to enter mass production late in the fourth quarter of 2025 or early in the first quarter of 2026, which situates launch in the first half of next year.

We also discuss Apple’s broader low-cost device strategy, how it seems to be positioning its device lineups going forward, and more. The MacRumors Show has its own YouTube channel, so make sure you’re subscribed to keep up with new episodes and clips.

You can also listen to The MacRumors Show on Apple Podcasts, Spotify, Overcast, or your preferred podcasts app. You can also copy our RSS feed directly into your podcast player.

If you haven’t already listened to the previous episode of The MacRumors Show, catch up to hear is talk through through the changes introduced in the second developer beta of iOS 26.

Subscribe to The MacRumors Show for new episodes every week, where we discuss some of the topical news breaking here on MacRumors, often joined by interesting guests such as Kevin Nether, John Gruber, Mark Gurman, Jon Prosser, Luke Miani, Matthew Cassinelli, Brian Tong, Quinn Nelson, Jared Nelson, Eli Hodapp, Mike Bell, Sara Dietschy, iJustine, Jon Rettinger, Andru Edwards, Arnold Kim, Ben Sullins, Marcus Kane, Christopher Lawley, Frank McShan, David Lewis, Tyler Stalman, Sam Kohl, Federico Viticci, Thomas Frank, Jonathan Morrison, Ross Young, Ian Zelbo, and Rene Ritchie.

The MacRumors Show is on X @MacRumorsShow, so be sure to give us a follow to keep up with the podcast. You can also head over to The MacRumors Show forum thread to engage with us directly. Remember to rate and review the podcast, and let us know what subjects and guests you would like to see in the future.

Plugin developer 3d-io has released Packer-IO 1.3, the latest version of the free standalone UV packing application for games, motion graphics, VFX and visualization artists.

New features include support for real-world scene scale units, and for texel density, making it possible for games artists to maintain consistent texture resolution across models.

3dio has also released a separate free Blender integration plugin, UnwrellaConnect, which makes it possible to use Packer-IO directly inside the open-source 3D software.

An efficient standalone UV packing tool for organic and hard surface models
First released last year, Packer-IO is a free standalone UV packing tool.

It automatically arranges the UV islands of an imported 3D model to minimise wasted UV space, reducing the total file sizes of its texture maps.

The software can be used on both hard surface and organic models, and supports tiled UV layouts, used in applications like ZBrush.

It works with assets with “thousands of charts and millions of polygons”, and is resolution-independent, so packing times remain constant, regardless of the resolution of the textures.

However, it is currently purely a UV packing tool: it doesn’t unwrap the UVs of a model first. You can find more details in our original story on Packer-IO 1.0.

Now supports scale units and texel density
3d-io has now released its first feature updates to the software, in the shape of Packer-IO 1.2, which came out last month, and Packer-IO 1.3, which came out earlier this week.

New features in version 1.2 include scale-aware packing, with users now able to specify real-world scene scale units, including both metric and imperial measurements.

Games artists get support for texel density: it is now possible to specify a target resolution in pixels per centimeter, so that packed textures look visually consistent with those of other assets.

Version 1.3 provides “vastly improved” hole filling, with Packer-IO packing smaller UV islands into the space left between larger islands more intelligently.

New UnwrellaConnect plugin makes it possible to pack UVs directly inside Blender
3d-io has also released a free connector plugin for Blender, UnwrellaConnect, which integrates Packer-IO with Blender’s UI, making it possible to pack UVs of models directly inside Blender.

It supersedes UV-Packer for Blender, 3d-io’s old Blender plugin, which was based on the same core technology.

The Blender download link on the UV-Packer website has now been redirected, although the other editions, UV-Packer for 3ds Max and UV-Packer for Unreal Engine, are still available.

So how do Packer-IO and UnwrellaConnect differ from UVPacker for Blender?
If you already have UV-Packer for Blender, one reason to update to Packer-IO and UnwrellaConnect would be better integration with the other tools in your production pipeline.

Packer-IO itself supports 3D models imported from any DCC app, while UnwrellaConnect – as the name suggests – also works with Unwrella-IO, 3d-io’s commercial UV unwrapping tool.

Packer-IO also features a more streamlined UI and workflow than previous UV packing tools.

Licensing and system requirements
Packer-IO 1.3 is compatible with Windows 10+ and macOS 10.15+. The software is free, and is licensed for use in commercial projects.

UnwrellaConnect is compatible with Blender 4.2+. Source code is available under a MIT license.

Download free standalone UV packing software Packer-IO from 3d-io’s product website

Download free Blender integration plugin UnwrellaConnect from the plugin’s GitHub repository

Have your say on this story by following CG Channel on Facebook, Instagram and X (formerly Twitter). As well as being able to comment on stories, followers of our social media accounts can see videos we don’t post on the site itself, including making-ofs for the latest VFX movies, animations, games cinematics and motion graphics projects.

Meet AUTOSEL, a Linux maintainer that helps keep the kernel stable.

AUTOSEL is a script, one that uses a Large Language Model (LLM) to complete its task.

It does a thankless job, one loathed by all kernel maintainers, that of backporting patches.

Backporting patches is a “very tedious and frustrating process” that “doesn’t scale,” said NVIDIA Distinguished Software Engineer Sasha Levin, in a talk at the Open Source Summit. The presentation focused on how AI is starting to be used in the Linux kernel community to help keep up with the Herculean tasks of maintaining the Linux kernel.

It may not be writing exciting new features for the open source operating system kernel yet, but AI has excelled in tasks that are mind-numbingly repetitive yet still necessary. In other words, AI is already making the lives of Linux kernel developers easier, said Levin, who helps maintain the Linux Kernel Stable and LTS trees.

Patch Inspection

As the world’s single largest open source project, the Linux kernel gets updated and upgraded … a lot.

The pace has remained constant: As many as 10,000 new patches have landed in the mainline kernel over a 10-week period.

Stable and Long-Term Support (LTS) kernel maintainers usually review around 100 patches per day, every day, including weekends and holidays.

Only a few, about 5-10, turn out to be suitable for backporting.

Levin’s AUTOSEL, written in Rust, takes a first pass at incoming commits, looking for similarities across submitted commits and past backporting decisions, and suggesting only those to human committers to seem to be worthy of closer review.

AUTOSEL is built from multiple LLMs, with each LLM used for a particular strength, as well as for cross-validation to reduce errors and hallucinations.

For each commit, the tool creates mathematical representations (or “embeddings”) of the text that preserve semantic meaning for every commit in the kernel’s history, making them easily comparable.

For human maintainers, the tools cuts down the number of commits that humans have to review. It even explains its reasoning in an email.

Know Your Tools

Like any other tool, the value of an LLM corresponds to how well it is understood by it user.

You can think of Large Language Models as the next generation of compilers, providing developers with a jump in productivity, Levin said. They act like “massive state machines,” though what is unusual about them is that they perform state transitions in a probabilistic, rather than deterministic, manner.

They are good at matching patterns given a huge number of parameters and an input provided by a user. A “temperature” parameter controls how probabilistic the LLM is, or how liberally it interprets its material.

Other Uses

And like any other technology, LLMs are first being tested in minor tasks.

LLMs excel at “small, well-defined tasks.” Levin said.

One such use is code generation and refactoring. Tightly-defined bug fixes, or converting code to other forms, such as standard APIs, are good tasks.

For the 6.15 kernel release, Levin had an LLM write a patch to convert the open-coded hash table implementation to a standard API.

Linux kernel 6.16 included git-resolve. This tool resolves incomplete or incorrect commit IDs, a nagging issue for core developers, though not one that occurs often enough to spend a lot of time manually writing a tool to figure out which commit an incomplete SHA-1 is actually connected to.

It took Levin all of 20 minutes to work with the LLM to create the tool.

It would take an engineer about half a day to create such a tool, making it not worth the effort given the relative rarity of the issue it addressed. Plus the LLM did a lot of extra credit: It created a set of self-tests and even documentation, which a human engineer would have done begrudgingly if at all.

There’s no end of clean-up tasks that could be done with the kernel, Levin said. An LLM could help non-native English speakers write descriptive commit messages.

CVE Classification

Another tedious chore is classifying security vulnerabilities (CVEs), a task that the Linux kernel community took on in 2024.

The work involves inspecting commits to see if they address security issues.

Originally, a set of “hacky Bash scripts” was written to help.

LLMs were used to replace the scripts with a set of far more refined tools written in Rust, which included a full set of testing tools and documentation.

Using AUTOSEL as a launching point, a CVE classifier uses LLMs to identify commits that address security issues, and then goes on to check the vulnerability hasn’t already addressed in a previous patch. This is an overwhelming task for humans, given the 40 million lines of code that make up the Linux kernel.

LLMs can understand the semantic meaning of the commits, which provides for a far more comprehensive matching capability. A Retrieval Augmented Generation (RAG) cycle pulls in the kernel’s development history and documentation (e.g., Git repositories) to minimize hallucinations.

The LLMs effectively act as AI agents, Levin noted. They can run git commands, such as git blame, directly against the code repository to learn from the history of the kernel development itself.

In summary, AI has thus far helped Linux scale its maintenance efforts, while enhancing consistency and reducing manual labor for tedious tasks.