2024 has been a landmark year for tech: the AI hype train is at full steam, with new LLM models emerging and Apple finally entering the race. This year has brought a whirlwind of updates: from fresh versions of frameworks and runtimes to monopoly accusations against Google over Chrome. Is 2025 the year we see quantum computers become mainstream? This is the 2024 tech news roundup.
Making AI profitable
The release of ChatGPT in 2022 sparked a chain reaction that continues unabated. With billions of dollars invested in training models, companies are racing to discover sustainable business models.
OpenAI new models
With a valuation of $157 billion but less than $2 billion in revenue, OpenAI is under pressure to generate returns on its massive investments. This year, OpenAI launched GPT-4o and the more advanced o1 model.
Part of this strategy includes introducing new plans, such as the ChatGPT Pro plan, which grants access to GPT-o1’s professional version. Whether the model justifies its $200-per-month price tag remains to be seen (and if that seems steep, don’t even think about a subscription to Devin).
In addition, OpenAI has partially released Sora, the text-to-video AI generator that was teased earlier this year. Sora can create 20-second videos in 1080p without sound, which suggests how compute-heavy AI video generation is.
Apple Intelligence
Apple has been bidding on its time to adopt AI. This year, the company introduced Apple Intelligence, which has full on-device capabilities.
While Apple’s AI offerings remain modest, including improved proofreading, summarization, enhanced Siri suggestions, and image generation, one standout feature has captured attention: the iPad now boasts an official calculator app, and it’s surprisingly impressive.
Meta’s new Llama
Meta released its latest Llama 3.2 large language model in September. The largest models are multimodal, supporting image reasoning, while the smallest are efficient enough to run on modest machines and phones.
The Llama 3.2 family can be downloaded from HuggingFace or llama.com. Llama remains the closest alternative to an open-source LLM, offering a foundation for AI applications beyond mere OpenAI wrappers.
Anthropic Claude can control your machine
Anthropic continues to refine its large language models. Its latest innovation, Computer Use, allows Claude to control your device: viewing your screen, moving the mouse, and typing. This capability is both awe-inspiring and alarming, underscoring AI’s potential for task automation.
This is as amazing as it’s terrifying, but it shows that AI has great potential to automate tasks.
AI blunders
Wherever you look there’s AI now. It’s on your Google searches, your phones, and your computers, and it’s even made its way into your messaging apps. Under this kind of pressure, not all AI projects have succeeded.
Two notable failures this year were the Rabbit R1 and the Humane Pin. These tech wearables, aimed at replacing smartphones, received harsh reviews and failed to deliver on their promises.
Microsoft also faced backlash with its Copilot Recall feature, which takes periodic screenshots of your computer. While its purpose — to enable an in-device memory you can query — has merit, the feature’s default activation and poor announcement left many users uneasy, evoking Orwellian undertones.
State of DevOps 2024
Google released the annual State of DevOps and this year the focus of the survey has fallen predictably on AI.
The report is well worth a read, but this is the TL;DR:
- Increased productivity but not delivery performance: While AI significantly boosts productivity, flow, and job satisfaction, it hasn’t improved software delivery performance as expected.
- Over-reliance concerns: AI adoption correlates with less time spent on meaningful work, suggesting developers may rely too heavily on AI for complex tasks.
- Low trust in AI-generated code: Despite widespread usage, 39.2% of respondents expressed little to no trust in AI-generated code quality. Developers feel compelled to validate and tweak AI outputs, comparing this reliance to the early days of StackOverflow.
- Environmental concerns: Many developers are worried about the environmental impact of AI, citing increased power consumption and resource costs. Microsoft’s decision to reactivate a nuclear facility to power its AI systems has only amplified these fears.
- AI code generation rises: By October, AI-generated code accounted for over 25% of new code at Google.
Developer highlights
2024 brought us new runtimes and new versions of popular frameworks.
Deno 2.0
Deno hit the 2.0 milestone this year, marking it with an impressive ad. Fully compatible with Node.js, Deno now supports NPM packages, making it an excellent time to give it a spin.
Deno 2.0 also includes:
- A stable standard library
- Support for private NPM registries
- JSR, an alternative package registry for JS and TS modules
- Monorepo support
- A new LTS support channel
Bun on Windows
Bun, the latest JavaScript runtime, has finally released a native version for Windows. Previously, Windows users needed WSL to run Bun, but that’s no longer necessary.
Key updates in the latest version of Bun:
- Windows compatibility
- A CSS bundler
- Enhanced NPM compatibility
- Bytecode compilation
React 19 gets a compiler
Following the steps of Svelte, React 19 now includes a new compiler that converts React code into efficient JavaScript, eliminating the need for manual memoization with hooks like useMemo and useCallback.
We’ve covered the new compiler in another post, so here are the highlights:
- Automatic Memoization: The compiler automatically applies optimizations equivalent to
useMemoanduseCallback, reducing unnecessary re-renders and improving application performance without manual code adjustments. - Enhanced Performance: Transforming React code into optimized JavaScript can potentially double application performance, leading to faster load times and a more responsive user experience.
- Automatic Code Splitting: The compiler analyzes your application and automatically splits code into optimized bundles, improving load times and ensuring efficient resource utilization.
- Enhanced Tree Shaking: By identifying and eliminating unused code, the compiler reduces the final bundle size, leading to faster application performance.
- Improved Server-Side Rendering (SSR): The compiler optimizes components for SSR, resulting in faster initial page loads and better SEO performance.
- Support for New Syntax and Features: It enables the use of upcoming JavaScript and React features, ensuring your application stays up-to-date with the latest advancements.
Framework releases in 2024
This year a lot of popular frameworks got updated:
- Angular 19: introduced experimental zoneless change detection support and server-side rendering improvements
- Laravel 11: brought new features and improvements, including a first-party web socket server called Laravel Reverb, enhancing real-time capabilities for applications.
- Remix 2.10: continued to enhance its full-stack web framework capabilities, focusing on server-side rendering and client-side routing.
- Svelte 5: added Runes, a new signal-powered reactivity API that serves as the core of Svelte 5. This version includes
sv, a brand-new CLI to manage Svelte projects. - Vue.js 4: introduced a new Composition API, improved TypeScript support, and enhanced performance optimizations.
- Django 4.2: continued to build upon its high-level Python framework, promoting rapid development and clean design, with updates that enhance scalability and security for web applications.
- Ruby on Rails 8: introduces several significant features and enhancements aimed at empowering developers to deploy and manage applications without relying on the cloud.
- Next.js 15: introduces React 19 support, new APIs, turbo pack improvements, and improvements on caching.
Google continues to make the headlines
Google remains at the forefront of AI innovation while simultaneously grappling with regulatory challenges. You may recall that Google was the original creator of transformers, the foundation for ChatGPT and all LLMs.
AlphaProof gets a medal
Google continues its advancements in deep learning. In 2016, Google’s AlphaGo defeated Lee Sedol in Go, the then-world champion in the sport. This year, the Alpha team focused on maths. AlphaProof and AlphaGeometry AI systems competed at the International Mathematical Olympiad, solving 4 out of 6 problems and winning the equivalent of a silver medal.
New quantum computer chip
A few weeks ago, Google announced another breakthrough in quantum computing, with a new quantum chip named ‘Willow’. The new chip benchmarks are impressive: it can complete Random Circuit Sampling (RCS) tasks in under five minutes, something that would take the fastest classical supercomputers billions of years.
Google browser monopoly
On a more controversial note, U.S. regulators found Google guilty of engaging in monopoly practices related to its search and browser products. The U.S. government is pushing Google to sell Chrome, which currently holds a 67% global market share. If that happens, which company has deep enough pockets to buy a browser valued at $20 billion?
Conclusion
As the relentless march of AI continues, we are witnessing the tech industry reach new levels of maturity. Developers are making clear decisions about where they want AI — and where they don’t. Let’s hope 2025 brings a more balanced approach to AI integration.
As we look forward to another year, one thing is certain: change. It’s an exciting time to be part of the developer community. Here’s to another year of building, learning, and innovating — happy coding!
Originally published at https://semaphoreci.com on December 23, 2024.
