Amid a push toward AI agents, with both Anthropic and OpenAI shipping multi-agent tools this week, Anthropic is more than ready to show off some of its more daring AI coding experiments. But as usual with claims of AI-related achievement, you’ll find some key caveats ahead.

On Thursday, Anthropic researcher Nicholas Carlini published a blog post describing how he set 16 instances of the company’s Claude Opus 4.6 AI model loose on a shared codebase with minimal supervision, tasking them with building a C compiler from scratch.

Over two weeks and nearly 2,000 Claude Code sessions costing about $20,000 in API fees, the AI model agents reportedly produced a 100,000-line Rust-based compiler capable of building a bootable Linux 6.9 kernel on x86, ARM, and RISC-V architectures.

Carlini, a research scientist on Anthropic’s Safeguards team who previously spent seven years at Google Brain and DeepMind, used a new feature launched with Claude Opus 4.6 called “agent teams.” In practice, each Claude instance ran inside its own Docker container, cloning a shared Git repository, claiming tasks by writing lock files, then pushing completed code back upstream. No orchestration agent directed traffic. Each instance independently identified whatever problem seemed most obvious to work on next and started solving it. When merge conflicts arose, the AI model instances resolved them on their own.

The resulting compiler, which Anthropic has released on GitHub, can compile a range of major open source projects, including PostgreSQL, SQLite, Redis, FFmpeg, and QEMU. It achieved a 99 percent pass rate on the GCC torture test suite and, in what Carlini called “the developer’s ultimate litmus test,” compiled and ran Doom.

It’s worth noting that a C compiler is a near-ideal task for semi-autonomous AI model coding: The specification is decades old and well-defined, comprehensive test suites already exist, and there’s a known-good reference compiler to check against. Most real-world software projects have none of these advantages. The hard part of most development isn’t writing code that passes tests; it’s figuring out what the tests should be in the first place.

Alpine skiing live streams at the 2026 Winter Olympics will see Austria attempt to continue their historic dominance of this event, with challenges expected from Switzerland and France.

Austria have claimed 40 golds and 128 medals in total since alpine skiing was introduced to the winter games back in 1936. Switzerland came out on top in 2022, though, and they are expected to be strong again when the action gets under way on Saturday, February 7 (full schedule below).

Ski jumping at the Winter Olympics should deliver plenty of drama and entertainment at Milano Cortina 2026.

Ski jumping has been part of the Winter Olympics since the inaugural edition of the event in 1924. Norway lead the way for golds (12) and total medals (36), but the likes of Finland, Austria, Germany and Japan have enjoyed success over the years. Meanwhile, Slovenia topped the medal charts in 2022 and will hope to be contenders again this year.

Big Tech capital expenditure for this year is predicted to rise 60pc from $410bn in 2025.

Meta, Google, Amazon and Microsoft are signalling a collective 2026 capital expenditure package of around $650bn, with AI, cloud and data centres as unsurprising high-ticket items. Wary investors, however, seem unhappy, and the Financial Times reported that Amazon, Google and Microsoft are set to lose $900bn in market cap altogether.

Big Tech capital expenditure predictions would mark a rise of 60pc from the $410bn spent in 2025 and 165pc from the $245bn spent the year before.

The four competitors see the race to provide AI compute as “the next winner-take-all or winner-takes-most market”, Gil Luria, an analyst at DA Davidson told Bloomberg. None of the companies are “willing to lose”, he added.

Amazon shares fell by 11pc after the company’s earnings call yesterday (5 February) in which company president and CEO Andy Jassy announced a $200bn capital expenditure (capex) plan for the year, growing more than 50pc since last year.

He reasoned that a 24pc revenue growth in Amazon’s cloud offerings and a 22pc growth in advertising is evidence that the heavy spending is paying off. This year’s spending will be focused on AI, chips, robotics and low-Earth orbit satellites, Jassy said.

Meanwhile, Microsoft announced a $37.5bn quarterly capital expenditure bill on 28 January, just slightly more than analyst estimates. But the company was the worst hit among the four for a while, dropping 18pc since the announcement.

The company had also, for the first time, disclosed the true nature of its close economic relationship with OpenAI. It reported that roughly 45pc of its $625bn expected in future cloud contracts was from the start-up, leading to investor wariness on its over-reliance on one customer.

Google parent Alphabet initially dropped 4pc in share price after it reported its earnings on Wednesday (4 February), but climbed back up to being just below 0.5pc since yesterday. Sales and earnings per share grew by 18pc and 31pc respectively during Q4, beating analyst expectations, while Alphabet’s cloud backlog grew by 55pc quarter-over-quarter to $240bn.

The company announced capex for the year between $175bn and $185bn, doubling expenses to meet customer demand and capitalise on the growth of the company’s AI offerings. Though despite fears of heavy spending, Gemini Enterprise is selling 8m seats and the Gemini App now has more than 750m monthly active users, which, Motley Fool reported, is keeping investors relatively content.

Lastly, Meta has announced its total expenses for 2026 to be in the range $115bn to $135bn. The growth, it said, is driven by an increased investment to support its Meta Superintelligence Labs efforts as well as its core business.

While the stocks rose 10pc after the earnings announcement, the Financial Times reports that it lost those gains after overall investor fear has pushed the tech-heavy Nasdaq down 4pc over the week.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

I’ve been teaching long enough to recognize when something fundamental is shifting in the classroom. Lately, that shift sounds like a single word echoing through my courses: why.

Why are we doing this? Why does it matter? Why should I care?

At first, it can sound like pushback, the kind of challenge that might once have been mistaken for defiance. But I don’t see it that way. When Gen Z students ask “why,” they’re not questioning authority; they’re questioning meaning. They’re trying to understand whether what they’re being asked to learn aligns with a world that already feels crowded with information, competition and contradiction.

And they’re right to ask.

Gen Z has grown up surrounded by constant messaging — some genuine, some hollow. They’ve seen companies preach purpose while chasing profit, influencers claim authenticity while filtering reality, and institutions talk about mental health while rewarding burnout. So when they step into a classroom, they’re not looking for performance. They’re looking for proof.

In many ways, “why” has replaced the old-fashioned raise of the hand. It’s the new signal for engagement, not disengagement. These students aren’t rebellious for sport; they’re searching for relevance. When they ask “why,” they’re asking us to show them the thread between knowledge and purpose.

For educators, that’s both thrilling and challenging. The old classroom contract may no longer be enough. Gen Z expects transparency in exchange for trust. They want to know not only what they’re learning but how it connects to who they’re becoming. That expectation is reshaping how many of us teach.

I’ve noticed that when I take the time to explain why we’re doing something — even briefly — engagement rises. It doesn’t need to be a speech or a slide titled “Why It Matters.” It can be a few sentences woven into the moment: “You’ll use this when you’re leading a team someday,” or “This will help you understand how strategy actually plays out in a business setting.” Framing purpose in passing often lands more effectively than any formal statement could. It tells students that there’s intention behind what they’re being asked to do.

And when the connection isn’t obvious, I try to make the learning process itself transparent. I’ll tell them why I’ve designed a particular project or changed an assignment from last semester. I explain my reasoning the way I’d want a mentor to explain theirs — not to justify, but to include. Once they see the care that goes into the design, their tone shifts from skeptical to curious.

New Perspective

That shift has changed my own mindset as an instructor. I’ve started to see my role less as delivering content and more as modeling thoughtfulness — the same kind I’m asking of them. I don’t have to declare that an assignment matters; I can show that it does by connecting it to a broader purpose, by caring about it visibly.

When things don’t go perfectly, I’ve learned to acknowledge that too. I used to think admitting uncertainty would weaken credibility. It turns out it does the opposite. When I tell students, “I’m still experimenting with how to teach this,” they don’t lose confidence — they lean in. They respect honesty because it mirrors their own experience of figuring things out.

That’s the real undercurrent here: authenticity has replaced authority as the key driver of credibility. Gen Z doesn’t automatically trust titles or experience; they trust consistency between what we say and what we do. They’ve been burned too many times by institutions that preached one set of values and practiced another. In the classroom, they want something simpler — teachers who mean what they say.

This doesn’t mean lowering standards or catering to comfort. If anything, it’s raised expectations. When students believe something has meaning, they work harder. I’ve seen it when my students analyze real company challenges instead of hypothetical ones, or when they present their findings to local business leaders rather than just to me. They’re sharper, more invested, and more willing to push themselves when the stakes feel real.

Even small acts of transparency build trust. Explaining why feedback is framed a certain way, or why participation matters, helps students see that the structure exists for a reason. They might not always agree, but they rarely tune out.

Overcoming Defensiveness

Of course, this approach can be draining. There are days when the “whys” feel relentless — when every question seems to demand another explanation, and you wonder if they’ll ever just take your word for it. But over time, I’ve come to see their skepticism not as defiance but as discernment. They’re not trying to tear down the system; they’re trying to make it make sense.

When a student asks, “Why are we doing this?” they’re really saying, “Help me to see the point.” That’s not cynicism. You might call it curiosity with higher benchmarks. And if we can meet that question with openness instead of defensiveness, the classroom becomes a space of shared inquiry rather than guarded authority.

There’s an irony in all this. The very generation accused of being distracted is, in many ways, the most focused — just not on what older models of education assumed mattered. They’re focused on meaning. They want clarity, fairness and consistency, but they also want a sense of humanity behind it all. They crave professors who teach like people, not policies.

Maybe that’s the lesson for us, too. If Gen Z is asking “why,” perhaps we should start asking it of ourselves — not as a challenge, but as reflection. Why do we teach the way we do? Why do we grade like this? Why do we define learning in these terms?

Teaching a generation that questions everything isn’t easy. But it’s not resistance, it’s renewal. Their “why” invites us to rediscover our own.

Other Apple Intelligence features get all the attention, but in iOS 26 one of my favorite tools does its thing quietly in the background. On iPhone models that are capable of running Apple’s AI tech, the Adaptive Power setting is at work behind the scenes to extend battery power, even on many older iPhones.

Currently, the iPhone uses as much power as it needs to perform its tasks. You can extend battery life by taking a few simple steps, such as reducing screen brightness and disabling the always-on display. Or, if your battery is running low, you can turn on Low Power Mode, which limits background activity, like fetching mail and downloading data, and dims the screen to help extend battery life. Low Power Mode also kicks in automatically when the battery level reaches 20%.

If Low Power Mode is the hammer that knocks down power consumption, Adaptive Power is the scalpel that intelligently trims energy savings here and there as needed. Based on Apple’s description that accompanies the control, the savings will be felt mostly in power-hungry situations such as recording videos, editing photos or even playing games.

Apple says Adaptive Power takes about a week to analyze your usage behavior before it begins actively working. And it works in the background without needing any management on your part. 

Here’s how Apple describes it in the iPhone user guide: “It uses on-device intelligence to predict when you’ll need extra battery power based on your recent usage patterns, then makes performance adjustments to help your battery last longer.”

Watch this: The iPhone 17 Pro Max Has Incredible Battery Life

12:10

Which iPhone models can use Adaptive Power?

The feature uses AI to monitor and choose when its power-saving measures should be activated, which means only phones compatible with Apple Intelligence get the feature. These are the models that have the option:

• iPhone 17
• iPhone 17 Pro and iPhone 17 Pro Max
• iPhone Air
• iPhone 16 and iPhone 16 Plus
• iPhone 16 Pro and iPhone 16 Pro Max
• iPhone 16e
• iPhone 15 Pro and iPhone 15 Pro Max

Although some iPad and Mac models support Apple Intelligence, the feature is only available on iPhones.

How to turn Adaptive Power on

Adaptive Power is on by default on the iPhone 17, iPhone 17 Pro, iPhone 17 Pro Max and iPhone Air. For other models, you must opt in to use it. In iOS 26, you’ll find the Adaptive Power toggle in Settings > Battery > Power Mode. To be alerted when the feature is active, turn on the Adaptive Power Notifications option.

Advertisement

Enlarge Image

Adaptive Power sounds like an outgrowth of Gaming Mode, introduced in iOS 18, which routes all available processing and graphics power to the frontmost app and pauses other processes in order to deliver the best experience possible — at the notable expense of battery life.

Advertisement

What does this mean for your charging habits?

Although we all want as much battery life as possible all the time, judging by the description, it sounds as if Adaptive Power’s optimizations will not always be active, even if you leave the feature on. “When your battery usage is higher than usual” could include a limited number of situations. Still, considering that according to a CNET survey, 61% of people upgrade their phones because of battery life, a feature such as Adaptive Power could extend the longevity of their phones just by updating to iOS 26.

I also wonder whether slightly adjusting display brightness could be disruptive, but in my experience so far, it hasn’t been noticeable. Because the feature also selectively de-prioritizes processing tasks, the outward effects seem minimal. When it’s activated on my iPhone 16 Pro, the only indication was the Adaptive Power alert that appeared.

We’ll get a better idea about how well Adaptive Power works as more people adopt iOS 26 and start buying new iPhone models. Also, remember that shortly after installing a major software update, it’s common to experience worse battery life as the system optimizes data in the background; Apple went so far as to remind customers that it’s a temporary side effect.

Although often tossed together into a singular ‘retro game’ aesthetic, the first game consoles that focused on 3D graphics like the Nintendo 64 and Sony PlayStation featured very distinct visuals that make these different systems easy to distinguish. Yet whereas the N64 mostly suffered from a small texture buffer, the PS’s weak graphics hardware necessitated compromises that led to the highly defining jittery and wobbly PlayStation graphics.

These weaknesses of the PlayStation and their results are explored by [LorD of Nerds] in a recent video. Make sure to toggle on subtitles if you do not speak German.

It could be argued that the PlayStation didn’t have a 3D graphics chip at all, just a video chip that could blit primitives and sprites to the framebuffer. This forced PS developers to draw 3D graphics without such niceties like a Z-buffer, putting a lot of extra work on the CPU.

This problem extends also to texture mapping, by doing affine texture mapping, as it’s called on the PS. This mapping of textures is rather flawed and leads to the constant shifting of textures as the camera’s perspective is not taken into account. Although this texture mapping can be improved, the developers of the game have to add more polygons for this, which of course reduces performance. This is the main cause of the shifting and wobbling of textures.

Another issue on the PS was a lack of mipmapping support, which means a sequence of the same texture, each with each a different resolution. This allows for high-resolution textures to be used when the camera is close, and low-resolution textures when far away. On the PS this lack of mipmapping led to many texture pixels being rendered to the same point on the display, with camera movement leading to interesting flickering effects.

When it came to rendering to the output format, the Nintendo 64 created smooth gradients between the texture pixels (texels) to make them fit on the output resolution, whereas the PS used the much more primitive nearest neighbor interpolation that made especially edges of objects look like they both shimmered and changed shape and color.

The PS also lacked a dedicated floating point unit to handle graphics calculations, forcing a special Geometry Transformation Engine (GTE) in the CPU to handle transformation calculations, but all in integer calculations instead of with floating point values. This made e.g. fixed camera angles as in Resident Evil games very attractive for developers as movement would inevitably lead to visible artefacts.

Finally, the cartridge-based games of the N64 could load data from the mask ROMs about 100x faster than from the PS’s CDs, and with much lower latency. All of these differences would lead to entirely different games for both game consoles, with the N64 being clearly superior for 3D games, yet the PS being released long before the N64 for a competitive price along with the backing of Sony would make sure that it became a commercial success.