Paul Kinlan, who leads the Chrome and the Open Web Developer Relations team at Google, asks and answers the question (with a no.): Frameworks are abstractions over a platform designed for people and teams to accelerate their teams new work and maintenance while improving the consistency and quality of the projects. They also frequently force a certain type of structure and architecture to your code base. This isn't a bad thing, team productivity is an important aspect of any software.

I'm of the belief that software development is entering a radical shift that is currently driven by agents like Replit's and there is a world where a person never actually has to manipulate code directly anymore. As I was making broad and sweeping changes to the functionality of the applications by throwing the Agent a couple of prompts here and there, the software didn't seem to care that there was repetition in the code across multiple views, it didn't care about shared logic, extensibility or inheritability of components... it just implemented what it needed to do and it did it as vanilla as it could.

I was just left wondering if there will be a need for frameworks in the future? Do the architecture patterns we've learnt over the years matter? Will new patterns for software architecture appear that favour LLM management?

Google's head of research Yossi Matias maintains that learning to code remains "as important as ever" despite AI's growing role in software development. While AI tools have reduced coding time for some developers -- and Alphabet CEO Sundar Pichai noting that AI now generates a quarter of all code, Matias stressed that human engineers still review and approve AI-generated code.

The Google executive, who also serves as a company VP, acknowledged that junior professionals have faced challenges gaining experience as AI handles entry-level tasks. Google has launched initiatives to support early-career employees through this transition. Matias compared coding literacy to basic mathematics, arguing it provides crucial understanding of technology regardless of career path.

Nathan Sobo "joined GitHub in late 2011 to build the Atom text editor," according to an online biography, "and he led the Atom team until 2018." Max Brunsfeld joined the Atom team in 2013, and "While driving Atom towards its 1.0 launch during the day, Max spent nights and weekends building Tree-sitter, a blazing-fast and expressive incremental parsing framework that currently powers all code analysis at GitHub."

Last year they teamed up with Antonio Scandurra (another Atom alumnus) to launch a new startup called Zed (which in 2023 raised $10 million, according to TechCrunch). And today the open source blog It's FOSS checks in on their open-source code editor — "Zed Editor". Mainly written in Rust, it supports running in CLI, diagnosing project-wide errors, split panes, and markdown previews: By default, any added content is treated as plain text. I used the language switcher to change it to Rust so that I would get proper syntax highlighting, indentation, error detection, and other useful language-specific functions. The switch highlighted all the Rust elements correctly, and I then focused on Zed Editor's user interface. The overall feel of the editor was minimal, with all the important options being laid out nicely.

[Its status bar] had some interesting panels. The first one I checked was the Terminal Panel, which, as the name suggests, lets you run commands, scripts, and facilitates interaction with system files or processes directly from within the editor. I then moved to the Assistant Panel, which is home to various large language models that can be integrated into Zed Editor. There are options like Anthropic, GitHub Copilot Chat, Ollama, OpenAI, and Google AI... The Zed Editor team has also recently introduced Zed AI in collaboration with Anthropic for assisting with coding, allowing for code generation, advanced context-powered interactions, and more...

The real-time collaboration features on Zed Editor are quite appealing too. To check them out, I had to log in with my GitHub account. After logging in, the Collab Panel opened up, and I could see many channels from the official Zed community. I could chat with others, add collaborators to existing projects, join a call with the option to share my screen and track other collaborators' cursors, add new contacts, and carry out many other collaborative tasks.

One can also use extensions and themes to extend what Zed Editor can do. There are some nice pre-installed themes as well.

"The Rust trademark policy has been updated and a new draft is available to view," announced the Rust Foundation this week.

The last proposed trademark policy (in April of 2023) was criticized by open source advocate Bruce Perens in The Register as "far awry of fair use which is legally permitted." The Rust Foundation says this new version has "incorporated a number of suggestions from the Rust community," in a blog post that summarizes the feedback and enumerates specific ways it's been addressed: 1. We primarily plan to lean on community reports for enforcement and have no intention of spending our limited resources policing the work of small creators.

2. We have removed the non-legal language summary and instead have clarified wording throughout as best we can while keeping the policy valid.

3. The Rust trademark does not cover use of the word "Rust" in general and instead pertains to its use in relevant technical settings.

4. We have updated the logo usage policy. Color modifications are allowed.

5. The non-endorsement rule is about managing perception of official affiliation with the Foundation and Rust Project, and is thus subjective.

6. We removed restrictions on the use of "Rust" and "Cargo" in package names. The crates prefixes "rust-" and "cargo-" are no longer reserved to the Rust Project.

7. We will usually allow the community to use the marks on limited merchandise (more details in the updated draft)....

[T]he central purpose of these updates is to empower all Rustaceans to engage with the Rust language ecosystem more confidently. As a final step in this process, we invite you to review the updated policy and share any blocking concerns you might have... Thank you to everyone who weighed in with helpful suggestions on the initial trademark policy draft we shared. The level of engagement and passion within the Rust community is inspiring to all of us at the Rust Foundation.
The tech news site Heise Online writes "It is noticeable that the language is much clearer and dispenses with a lot of legal jargon," in a piece which argues the new draft "should calm the waves and create clarity." The new draft is not only formulated more simply, but is also significantly shorter. Some restrictions have been softened in the new rules or have disappeared completely...

Meanwhile, the Foundation has also adapted its logo so that it is clear which logo stands for the programming language and which for the Foundation. The use of the name Rust is explicitly permitted to identify projects that are either written in the programming language or are compatible with it...

Before the new trademark rules come into force, the Rust Foundation is collecting feedback on the current draft. The web form is open until November 20, 2024.

I had the opportunity to interview esteemed programmer Steve Yegge for the SourceForge Podcast to ask him all about AI-powered coding assistants and the future of programming. "We're moving from where you have to write the code to where the LLM will write the code and you're just having a conversation with it about the code," said Yegge. "That is much more accessible to people who are just getting into the industry."

Steve has nearly 30 years of programming experience working at Geoworks, Amazon, Google, Grab and now SourceGraph, working to build out the Cody AI assistant platform. Here's his Wikipedia page. He's not shy about sharing his opinions or predictions for the industry, no matter how difficult it may be for some to hear. "I'm going to make the claim that ... line-oriented programming, which we've done for the last 40, 50 years, ... is going away. It is dying just like assembly language did, and it will be completely dead within five years."

You can watch the episode on YouTube and stream on all major podcast platforms. A transcription of the podcast is available here.

Software engineer David Gomes writes in a blog post: I sometimes need to search for a website that will "convert a PNG to SVG", or "remove page from PDF" or "resize svg". And these apps are... okay. I don't really trust most of them with my data, and also a lot of times they just don't work or have too many ads. So, I've been noticing a trend of people just using AI agents to create full blown apps for these simple use cases.

I decided to try it myself for a "resize SVG" app since I recently had to go through a bunch of websites to do this. So, I pulled up Replit Agent and even though I've used it before, it doesn't cease to amaze me just how insanely good it is. The level of polish on this product is unlike any other AI agent out there right now. It starts off by drawing up a plan and asking you for feedback on that plan. Then, it'll just go to town and try to build the app. But what's super clever about it is that the agent asks you for feedback along the way. Effectively, the Replit Agent guides you, not the other way around (as one might have expected).

Anton Shilov reports via Tom's Hardware: FFmpeg is an open-source video decoding project developed by volunteers who contribute to its codebase, fix bugs, and add new features. The project is led by a small group of core developers and maintainers who oversee its direction and ensure that contributions meet certain standards. They coordinate the project's development and release cycles, merging contributions from other developers. This group of developers tried to implement a handwritten AVX512 assembly code path, something that has rarely been done before, at least not in the video industry.

The developers have created an optimized code path using the AVX-512 instruction set to accelerate specific functions within the FFmpeg multimedia processing library. By leveraging AVX-512, they were able to achieve significant performance improvements -- from three to 94 times faster -- compared to standard implementations. AVX-512 enables processing large chunks of data in parallel using 512-bit registers, which can handle up to 16 single-precision FLOPS or 8 double-precision FLOPS in one operation. This optimization is ideal for compute-heavy tasks in general, but in the case of video and image processing in particular.

The benchmarking results show that the new handwritten AVX-512 code path performs considerably faster than other implementations, including baseline C code and lower SIMD instruction sets like AVX2 and SSSE3. In some cases, the revamped AVX-512 codepath achieves a speedup of nearly 94 times over the baseline, highlighting the efficiency of hand-optimized assembly code for AVX-512.

GitHub released its annual "State of the Octoverse" report this week. And while "Systems programming languages, like Rust, are also on the rise... Python, JavaScript, TypeScript, and Java remain the most widely used languages on GitHub."

In fact, "In 2024, Python overtook JavaScript as the most popular language on GitHub." They also report usage of Jupyter Notebooks "skyrocketed" with a 92% jump in usage, which along with Python's rise seems to underscore "the surge in data science and machine learning on GitHub..." We're also seeing increased interest in AI agents and smaller models that require less computational power, reflecting a shift across the industry as more people focus on new use cases for AI... While the United States leads in contributions to generative AI projects on GitHub, we see more absolute activity outside the United States. In 2024, there was a 59% surge in the number of contributions to generative AI projects on GitHub and a 98% increase in the number of projects overall — and many of those contributions came from places like India, Germany, Japan, and Singapore...

Notable growth is occurring in India, which is expected to have the world's largest developer population on GitHub by 2028, as well as across Africa and Latin America... [W]e have seen greater growth outside the United States every year since 2013 — and that trend has sped up over the past few years.
Last year they'd projected India would have the most developers on GitHub #1 by 2027, but now believe it will happen a year later. This year's top 10?

1. United States
2. India
3. China
4. Brazil
5. United Kingdom
6. Russia
7. Germany
8. Indonesia
9. Japan
10. Canada

Interestingly, the UK's population ranks #21 among countries of the world, while Germany ranks #19, and Canada ranks #36.)

GitHub's announcement argues the rise of non-English, high-population regions "is notable given that it is happening at the same time as the proliferation of generative AI tools, which are increasingly enabling developers to engage with code in their natural language." And they offer one more data point: GitHub's For Good First Issue is a curated list of Digital Public Goods that need contributors, connecting those projects with people who want to address a societal challenge and promote sustainable development...

Significantly, 34% of contributors to the top 10 For Good Issue projects... made their first contribution after signing up for GitHub Copilot.
There's now 518 million projects on GitHub — with a year-over-year growth of 25%...

"AI-driven 0-day detection is here," argues a new blog post from ZeroPath, makers of a GitHub app that "detects, verifies, and issues pull requests for security vulnerabilities in your code."

They write that AI-assisted security research "has been quietly advancing" since early 2023, when researchers at the DARPA and ARPA-H's Artificial Intelligence Cyber Challenge demonstrated the first practical applications of LLM-powered vulnerability detection — with new advances continuing. "Since July 2024, ZeroPath's tool has uncovered critical zero-day vulnerabilities — including remote code execution, authentication bypasses, and insecure direct object references — in popular AI platforms and open-source projects." And they ultimately identified security flaws in projects owned by Netflix, Salesforce, and Hulu by "taking a novel approach combining deep program analysis with adversarial AI agents for validation. Our methodology has uncovered numerous critical vulnerabilities in production systems, including several that traditional Static Application Security Testing tools were ill-equipped to find..." TL;DR — most of these bugs are simple and could have been found with a code review from a security researcher or, in some cases, scanners. The historical issue, however, with automating the discovery of these bugs is that traditional SAST tools rely on pattern matching and predefined rules, and miss complex vulnerabilities that do not fit known patterns (i.e. business logic problems, broken authentication flaws, or non-traditional sinks such as from dependencies). They also generate a high rate of false positives.

The beauty of LLMs is that they can reduce ambiguity in most of the situations that caused scanners to be either unusable or produce few findings when mass-scanning open source repositories... To do this well, you need to combine deep program analysis with an adversarial agents that test the plausibility of vulnerabilties at each step. The solution ends up mirroring the traditional phases of a pentest — recon, analysis, exploitation (and remediation which is not mentioned in this post)...

AI-driven vulnerability detection is moving fast... What's intriguing is that many of these vulnerabilities are pretty straightforward — they could've been spotted with a solid code review or standard scanning tools. But conventional methods often miss them because they don't fit neatly into known patterns. That's where AI comes in, helping us catch issues that might slip through the cracks.
"Many vulnerabilities remain undisclosed due to ongoing remediation efforts or pending responsible disclosure processes," according to the blog post, which includes a pie chart showing the biggest categories of vulnerabilities found:

• 53%: Authorization flaws, including roken access control in API endpoints and unauthorized Redis access and configuration exposure. ("Impact: Unauthorized access, data leakage, and resource manipulation across tenant boundaries.")

• 26%: File operation issues, including directory traversal in configuration loading and unsafe file handling in upload features. ("Impact: Unauthorized file access, sensitive data exposure, and potential system compromise.")

• 16%: Code execution vulnerabilities, including command injection in file processing and unsanitized input in system commands. ("Impact: Remote code execution, system command execution, and potential full system compromise.")

The company's CIO/cofounder was "former Red Team at Tesla," according to the startup's profile at YCombinator, and earned over $100,000 as a bug-bounty hunter. (And another co-founded is a former Google security engineer.)

Thanks to Slashdot reader Mirnotoriety for sharing the article.

Google is advancing the release timeline for Android 16, shifting it to the second quarter of 2025 to better align with new device launches and accelerate access to its latest AI and machine learning resources. It should also "enable app creators and phone companies to prepare their products for the new software more quickly," reports CNET. From the report: [I]n a big-picture sense, the change could help facilitate a new wave of apps with more AI integration, considering developers will get access to Google's latest machine learning and AI resources even sooner. "We're in a once-in-a-generation moment to completely reimagine what our smartphones can do and how we interact with them," Google's Seang Chau, who took on the role of vice president and general manager of the Android Platform earlier this year, said in an interview with CNET. "It's a really exciting time for smartphones, and we've been putting a lot of thought into what we want to do next with them."

In addition to moving up the major release, Google will roll out a minor update in the fourth quarter of 2025 with feature updates, optimizations and bug fixes. It's a notable switch from Google's usual release timeline, but it's just one of several changes the company has made to the way it distributes Android updates in an effort to add features more frequently. [...] "Things are moving quite fast in the AI world right now," Chau said. "So we want to make sure that we get those developer [application programming interfaces], especially around machine learning and AI, available to our developers so they can build these capabilities faster and get them out to our users faster."

GitHub Copilot will switch from using exclusively OpenAI's GPT models to a multi-model approach, adding Anthropic's Claude 3.5 Sonnet and Google's Gemini 1.5 Pro. Ars Technica reports: First, Anthropic's Claude 3.5 Sonnet will roll out to Copilot Chat's web and VS Code interfaces over the next few weeks. Google's Gemini 1.5 Pro will come a bit later. Additionally, GitHub will soon add support for a wider range of OpenAI models, including GPT o1-preview and o1-mini, which are intended to be stronger at advanced reasoning than GPT-4, which Copilot has used until now. Developers will be able to switch between the models (even mid-conversation) to tailor the model to fit their needs -- and organizations will be able to choose which models will be usable by team members.

The new approach makes sense for users, as certain models are better at certain languages or types of tasks. "There is no one model to rule every scenario," wrote [GitHub CEO Thomas Dohmke]. "It is clear the next phase of AI code generation will not only be defined by multi-model functionality, but by multi-model choice." It starts with the web-based and VS Code Copilot Chat interfaces, but it won't stop there. "From Copilot Workspace to multi-file editing to code review, security autofix, and the CLI, we will bring multi-model choice across many of GitHub Copilot's surface areas and functions soon," Dohmke wrote. There are a handful of additional changes coming to GitHub Copilot, too, including extensions, the ability to manipulate multiple files at once from a chat with VS Code, and a preview of Xcode support. GitHub also introduced "Spark," a natural language-based app development tool that enables both non-coders and coders to create and refine applications using conversational prompts. It's currently in an early preview phase, with a waitlist available for those who are interested.

Google has integrated AI deeply across its operations, with over 25% of its new code generated by AI. CEO Sundar Pichai announced the milestone during the company's third quarter 2024 earnings call. The Verge reports: AI is helping Google make money as well. Alphabet reported $88.3 billion in revenue for the quarter, with Google Services (which includes Search) revenue of $76.5 billion, up 13 percent year-over-year, and Google Cloud (which includes its AI infrastructure products for other companies) revenue of $11.4 billion, up 35 percent year-over-year. Operating incomes were also strong. Google Services hit $30.9 billion, up from $23.9 billion last year, and Google Cloud hit $1.95 billion, significantly up from last year's $270 million. "In Search, our new AI features are expanding what people can search for and how they search for it," CEO Sundar Pichai says in a statement. "In Cloud, our AI solutions are helping drive deeper product adoption with existing customers, attract new customers and win larger deals. And YouTube's total ads and subscription revenues surpassed $50 billion over the past four quarters for the first time."

theodp writes: "Computer science education is a necessity for all students," argues tech-backed nonprofit Code.org in its newly-published 2024 State of Computer Science Education (Understanding Our National Imperative) report. "Students of all identities and chosen career paths need quality computer science education to become informed citizens and confident creators of content and digital tools."

In the 200-page report, Code.org pays special attention to participation in "foundational computer science courses" in high school. "Across the country, 60% of public high schools offer at least one foundational computer science course," laments Code.org (curiously promoting a metric that ignores school size which nonetheless was embraced by Education Week and others).

"A course that teaches foundational computer science includes a minimum amount of time applying learned concepts through programming (at least 20 hours of programming/coding for grades 9-12 high schools)," Code.org explains in a separate 13-page Defining Foundational Computer Science document. Interestingly, Code.org argues that Data and Informatics courses -- in which "students may use Oracle WebDB, SQL, PL/SQL, SPSS, and SAS" to learn "the K-12 CS Framework concepts about data and analytics" -- do not count, because "the course content focuses on querying using a scripting language rather than creating programs [the IEEE's Top Programming Languages 2024 begs to differ]." Code.org similarly dissed the use of the Wolfram Language for broad educational use back in 2016.

With its insistence on the importance of kids taking Code.org-defined 'programming' courses in K-12 to promote computational thinking, it's probably no surprise to see that the data behind the 2024 State of Computer Science Education report was prepared using Python (the IEEE's top programming language) and presented to the public in a Jupyter notebook. Just kidding. Ironically, the data behind the 2024 State of Computer Science Education analysis is prepared and presented by Code.org in a no-code Tableau workbook.

Speaking of streaming services, an anonymous reader shares a story that looks into Apple's entertainment offering: Ever since its launch in 2019, Apple TV+ has been carving out an identity as the new home for prestige shows from some of Hollywood's biggest names -- the kind of shows that sound natural coming out of Jimmy Kimmel's mouth in monologue jokes at the Emmys. While the company never provides spending details, Apple is estimated to have spent at least $20 billion recruiting the likes of Reese Witherspoon, M. Night Shayamalan, and Harrison Ford to help cultivate its award-worthy sheen. For all the effort Apple has expended, and for all the cultural excitement around Ted Lasso during its three-season run, the streaming service has won nearly 500 Emmys ... while attracting just 0.2% of total TV viewing in the U.S.

No wonder the company reportedly began reining in its spending spree recently. (Apple did not reply to a request for comment.) "It seems like Apple TV wants to be seen as a platform that's numbers-agnostic," says Ashley Ray, comedian, TV writer, and host of the erstwhile podcast TV I Say. "They wanna be known for being about the creativity and the love of making TV shows, even if nobody's watching them."

The experience of enjoying a new Apple TV+ series can often be a lonely one. Adventurous subscribers might see an in-network ad about something like last summer's Sunny, the timely, genre-bending Rashida Jones series about murderous AI, and give it a shot -- only to find that nobody else is talking about it in their social media feeds or around the company Keurig machine. Sure, the same could be said for hundreds of other streaming series in the post-monoculture era, but most streaming companies aren't consistently landing as much marquee talent for such a limited library. (Apple currently has 259 TV shows and films compared to Netflix's nearly 16,000.)

How is it possible for a streaming service to have as much high-pedigree programming as Apple TV+ does and so relatively few viewers, despite an estimated 25 million paid subscribers? How can shows starring Natalie Portman, Idris Elba, and Colin Farrell launch and even get renewed without ever quite grazing the zeitgeist? How does a show set in the same Monsterverse as Godzilla vs. Kong, and starring Kurt Russell and his roguishly charming son, not become a monster-size hit?

For many perplexed observers, the blame falls squarely on Apple's marketing efforts, or seeming lack thereof.

"After two years of being beaten with the memory-safety stick, the C++ community has published a proposal to help developers write less vulnerable code," reports the Register.

"The Safe C++ Extensions proposal aims to address the vulnerable programming language's Achilles' heel, the challenge of ensuring that code is free of memory safety bugs..." Acknowledging the now deafening chorus of calls to adopt memory safe programming languages, developers Sean Baxter, creator of the Circle compiler, and Christian Mazakas, from the C++ Alliance, argue that while Rust is the only popular systems level programming language without garbage collection that provides rigorous memory safety, migrating C++ code to Rust poses problems. "Rust lacks function overloading, templates, inheritance and exceptions," they explain in the proposal. "C++ lacks traits, relocation and borrow checking. These discrepancies are responsible for an impedance mismatch when interfacing the two languages. Most code generators for inter-language bindings aren't able to represent features of one language in terms of the features of another."

Though DARPA is trying to develop better automated C++ to Rust conversion tools, Baxter and Mazakas argue telling veteran C++ developers to learn Rust isn't an answer... The Safe C++ project adds new technology for ensuring memory safety, Baxter explained, and isn't just a reiteration of best practices. "Safe C++ prevents users from writing unsound code," he said. "This includes compile-time intelligence like borrow checking to prevent use-after-free bugs and initialization analysis for type safety." Baxter said that rewriting a project in a different programming language is costly, so the aim here is to make memory safety more accessible by providing the same soundness guarantees as Rust at a lower cost. "With Safe C++, existing code continues to work as always," he explained. "Stakeholders have more control for incrementally opting in to safety."

The next step, Baxter said, involves greater participation from industry to help realize the Safe C++ project. "The foundations are in: We have fantastic borrow checking and initialization analysis which underpin the soundness guarantees," he said. "The next step is to comprehensively visit all of C++'s features and specify memory-safe versions of them. It's a big effort, but given the importance of reducing C++ security vulnerabilities, it's an effort worth making."

Last month C dropped from 3rd to 4th in TIOBE's ranking of programming language popularity (which tries to calculate each language's share of search engine results). Java moved up into the #3 position in September, reports TechRepublic, which notes that by comparison October "saw relatively little change" — though percentages of search results increased slightly. "At number one, Python jumped from 20.17% in September to 21.9% in October. In second place, C++ rose from 10.75% in September to 11.6%. In third, Java ascended from 9.45% to 10.51%..."

Is there a larger trend? TIOBE CEO Paul Jansen writes that the need to harvest more data increases demand for fast data manipulation languages. But they also need to be easy to learn ("because the resource pool of skilled software engineers is drying up") and secure ("because of continuous cyber threats.") King of all, Python, is easy to learn and secure, but not fast. Hence, engineers are frantically looking for fast alternatives for Python. C++ is an obvious candidate, but it is considered "not secure" because of its explicit memory management. Rust is another candidate, although not easy to learn. Rust is, thanks to its emphasis on security and speed, making its way to the TIOBE index top 10 now. [It's #13 — up from #20 a year ago]

The cry for fast, data crunching languages is also visible elsewhere in the TIOBE index. The language Mojo [a faster superset of Python designed for accelerated hardware like GPUs]... enters the top 50 for the first time. The fact that this language is only 1 year old and already showing up, makes it a very promising language.
In the last 12 months three languages also fell from the top ten:

• PHP (dropping from #8 to #15)

• SQL (dropping from #9 to #11)

• Assembly language (dropping from #10 to #16)

This week (MIT-licensed) WebAssembly runtime Wasmer announced "a major milestone in making any software run with WebAssembly."

The announcement's headline? Running Clang in the browser using WebAssembly... Thanks to the newest release of Wasmer (4.4) and the Wasmer JS SDK (0.8.0) you can now run [compiler front-end] clang anywhere Wasmer runs! This allows compiling C programs from virtually anywhere. Including Javascript and your preferred browser! (we tested Chrome, Safari and Firefox and everything is working like a charm)...

- You can compile C code to WebAssembly easily just using the Wasmer CLI: no toolchains or complex installations needed, install Wasmer and you are ready to go...!

- You can compile C projects directly from JavaScript...!

- We expect online IDEs to start adopting the SDK to allow their users compile and run C programs in the browser....

Do you want to use clang in your Javascript project? Thanks to our newly released Wasmer JS SDK you can do it easily, in both the browser and Node.js/Bun etc... Wasmer's clang can even optimize the file for you automatically using wasm-opt under the hood (Clang automatically detects if wasm-opt is used, and it will be automatically called when optimizing the file). Imagine using Emscripten without needing its toolchain installed — or even better, imagine running Emscripten in the browser.
The announcement looks to a future of compiling native Python libraries, when "any project depending on LLVM can now be easily compiled to WebAssembly..."

"This is the beginning of an awesome journey, we can't wait to see what you create next with this."

80% of software engineers will need to upskill by 2027 to keep pace with generative AI's growing demands, according to Gartner. The consultancy predicts AI will transform the industry in three phases. Initially, AI tools will boost productivity, particularly for senior developers. Subsequently, "AI-native software engineering" will emerge, with most code generated by AI. Long-term, AI engineering will rise as enterprise adoption increases, requiring a new breed of professionals skilled in software engineering, data science, and machine learning.

Researchers at BitEnergy AI, Inc. have developed Linear-Complexity Multiplication (L-Mul), a technique that reduces AI model power consumption by up to 95% by replacing energy-intensive floating-point multiplications with simpler integer additions. This method promises significant energy savings without compromising accuracy, but it requires specialized hardware to fully realize its benefits. Decrypt reports: L-Mul tackles the AI energy problem head-on by reimagining how AI models handle calculations. Instead of complex floating-point multiplications, L-Mul approximates these operations using integer additions. So, for example, instead of multiplying 123.45 by 67.89, L-Mul breaks it down into smaller, easier steps using addition. This makes the calculations faster and uses less energy, while still maintaining accuracy. The results seem promising. "Applying the L-Mul operation in tensor processing hardware can potentially reduce 95% energy cost by element wise floating point tensor multiplications and 80% energy cost of dot products," the researchers claim. Without getting overly complicated, what that means is simply this: If a model used this technique, it would require 95% less energy to think, and 80% less energy to come up with new ideas, according to this research.

The algorithm's impact extends beyond energy savings. L-Mul outperforms current 8-bit standards in some cases, achieving higher precision while using significantly less bit-level computation. Tests across natural language processing, vision tasks, and symbolic reasoning showed an average performance drop of just 0.07% -- a negligible tradeoff for the potential energy savings. Transformer-based models, the backbone of large language models like GPT, could benefit greatly from L-Mul. The algorithm seamlessly integrates into the attention mechanism, a computationally intensive part of these models. Tests on popular models such as Llama, Mistral, and Gemma even revealed some accuracy gain on certain vision tasks.

At an operational level, L-Mul's advantages become even clearer. The research shows that multiplying two float8 numbers (the way AI models would operate today) requires 325 operations, while L-Mul uses only 157 -- less than half. "To summarize the error and complexity analysis, L-Mul is both more efficient and more accurate than fp8 multiplication," the study concludes. But nothing is perfect and this technique has a major achilles heel: It requires a special type of hardware, so the current hardware isn't optimized to take full advantage of it. Plans for specialized hardware that natively supports L-Mul calculations may be already in motion. "To unlock the full potential of our proposed method, we will implement the L-Mul and L-Matmul kernel algorithms on hardware level and develop programming APIs for high-level model design," the researchers say.

An anonymous reader quotes a report from VentureBeat: Epic Games CEO Tim Sweeney opened the Unreal Fest Seattle event today with an update on news that included a blistering criticism of monopolistic platform owners. Sweeney is a big proponent of open platforms and the open metaverse. In fact, he will talk about that subject in a virtual talk at our GamesBeat Next 2024 event on October 28-29 in San Francisco. (You can use this code for a 25% discount: gbn24dean). And so Sweeney continues to pressure the major platforms to give more favorable terms to game developers. He started out on that front by giving a price cut for users of Unreal Engine 5, Epic's tools for making games. For those who release games first or simultaneously on the Epic Games Store, Epic is cutting its royalty rate from 5% to 3.5% for Unreal developers. He noted that Epic is in better financial shape than it was a year ago, when Epic had to lay off a lot of staff. Sweeney said the company spent the last year rebuilding. "We're at a point now where game development is expensive. It's low margin, and game companies are suffering. Apple and Google make way more profit from most games than the developers make themselves, while contributing nothing," Sweeney said.

Sweeney reminisced about programming on early Apple computers, aligning with Steve Wozniak's vision for Apple where users had complete freedom without corporate restrictions. He contrasted this with today's mobile platforms, accusing Apple and Google of acting as gatekeepers that stifle innovation. "Among the fights we've taken on here, he noted the case with Apple is still an ongoing fight to open up payments so developers can process payments without Apple mediation and without Apple fees," he said, noting the "massive victory" against Google in a jury trial late last year.