Google’s Antigravity 2.0 apparently used 96 agents running in parallel to write an operating system from scratch in 12 hours for under a thousand US dollars in token costs. And it runs Doom.

That’s the claim, anyway.

The Doom thing has become a genuine benchmark meme at this point. Someone ran Doom on a pregnancy test display a few years back. Doom runs on ESP32 microcontrollers. Doom runs on graphing calculators. If your new piece of technology can’t run Doom, that’s probably the more interesting story. So let’s hold that particular detail lightly.

So I’ve been down a rabbit hole this Easter weekend, and it has nothing to do with chocolate eggs. Google DeepMind dropped Gemma 4, and the local AI community has basically lost its collective mind — in the best possible way.

For those not deep in the weeds on this stuff, Gemma is Google’s family of open-weights AI models. The new Gemma 4 lineup ranges from tiny models designed to run on phones all the way up to a 31 billion parameter beast that’ll give your home server a decent workout. And the specs are genuinely impressive: multimodal input handling text, images, video and audio, context windows up to 256K tokens, native tool calling, built-in reasoning modes, and support for over 140 languages. That last point is actually more significant than most people give it credit for — more on that in a moment.

Google’s latest AI announcement has the tech world buzzing again. Apparently, they’ve built an AI that “learns from its own mistakes in real time.” Cue the usual chorus of “holy shit” reactions and breathless headlines about revolutionary breakthroughs. But hang on a minute – let’s take a step back and actually think about what this means.

Reading through the various reactions online, it’s fascinating to see the divide between those who understand the technical details and those who just see the marketing speak. The more technically-minded folks are pointing out that this sounds a lot like glorified RAG (Retrieval-Augmented Generation) – essentially fancy context management where the AI stores its reasoning process and refers back to it when similar problems arise. It’s not actually changing its core weights or truly “learning” in the way we might imagine.

The tech world’s been buzzing lately about Google’s latest move to charge developers for app verification outside the Play Store ecosystem. What started as a promise of openness and choice in the Android world is rapidly becoming another corporate cash grab, and frankly, it’s getting under my skin.

Google’s decision to implement a tiered verification system for Android developers feels like a betrayal of everything the platform once stood for. Sure, they’re keeping a “free” tier for hobbyists and students, but with undefined installation limits and heavy encouragement to upgrade to the paid tier. The paid verification will cost developers $25 - the same as Play Store registration - just for the privilege of distributing apps outside Google’s walled garden.

There’s something almost comically ironic about my current predicament. Here I am, a DevOps engineer who spends his days wrestling with code, infrastructure, and the endless march of technological progress, and I’ve stumbled across a question that’s been gnawing at me for weeks now.

It started with a post on Reddit that made me pause mid-scroll. Someone was asking whether the Go programming language itself could be a privacy concern, simply because Google created it. At first glance, it sounds almost absurd – worrying about the privacy implications of a programming language is like being suspicious of the pencil because you don’t trust the company that made the graphite. But the more I thought about it, the more I realised this question touches on something much deeper about our relationship with technology in 2024.

Recently, I’ve been diving deep into the rabbit hole of digital privacy, specifically focusing on Google’s data retention policies. The topic hits close to home, especially since I’ve spent countless hours tinkering with development tools and cloud services, always with that nagging feeling about the digital footprints I’m leaving behind.

Google’s “My Activity” deletion feature presents itself as a simple solution to wipe your digital slate clean. But let’s be real - it’s about as effective as using a garden hose to clean up after a flood. Their own policy states that while deleted activity is “immediately removed from view,” they still retain certain information for the “life of your Google Account.” That’s corporate speak for “we’re keeping whatever we want.”

The tech world is buzzing about Google’s latest AI demonstration, Project Astra, and honestly, it’s bringing back memories of countless “revolutionary” product launches I’ve witnessed over my decades in IT. While watching the polished demo of someone using AI to fix their bike, I found myself caught between excitement and skepticism.

Let’s be real - the demo looks impressive. The seamless interaction between human and AI, the contextual understanding, the ability to make phone calls and find specific information… it’s the stuff we’ve been promised since the early days of Siri. But having lived through numerous Google demos that never quite materialized (remember Duplex?), I’m keeping my expectations in check.

The digital landscape keeps throwing curveballs at us, and the latest one’s particularly fascinating. Recently, there’s been quite a stir about Google’s AI pulling “citations” directly from Reddit comments. The example making rounds involves a Smashing Pumpkins performance at Lollapalooza, where Google’s AI confidently declared it was “well-received” based on a single Reddit comment using the phrase “one-two punch” - despite historical accounts suggesting they were actually booed off stage after three songs.

The tech world never ceases to amaze me with its rapid advancements. Google just dropped something fascinating - new quantization-aware trained (QAT) checkpoints for their Gemma models that promise better performance while using significantly less memory. This isn’t just another incremental improvement; it’s a glimpse into the future of AI model optimization.

Running large language models locally has always been a delicate balance between performance and resource usage. Until now, quantizing these models (essentially compressing them to use less memory) usually meant accepting a noticeable drop in quality. It’s like trying to compress a high-resolution photo - you save space, but lose some detail in the process.

Something rather amusing is happening in the world of AI right now. Google’s latest Gemini model (specifically Exp 1114) has climbed to the top of the Chatbot Arena rankings, matching or surpassing its competitors across multiple categories. But there’s a catch - it seems to be having an identity crisis.

When asked about its identity, this Google-created AI sometimes claims to be Claude, an AI assistant created by Anthropic. It’s a bit like walking into a McDonald’s and having the person behind the counter insist they work at Hungry Jack’s. The tech community is having a field day with this peculiar behaviour, with some suggesting Google might have trained their model on Claude’s data.