The News,
Decoded.

The Israeli Medical Association has bravely declared that doctors should not become 'enslaved' by artificial intelligence. A position paper has been published, setting boundaries for AI use in medicine. Because nothing says 'we have control' like a piece of paper politely asking the unstoppable AI wave to please stop at the red rope. The phrasing itself is telling: 'enslaved' is a strong word for a profession that has spent decades outsourcing decision-making to flowcharts and lab results.

The association's aim, per the excerpt, is to 'use the technology without becoming enslaved by it.' One imagines the paper includes detailed clauses: 'The algorithm shall not prescribe narcotics without a human cosign,' 'The AI shall not override a doctor's gut feeling about a patient with atypical symptoms,' 'The robot shall not roll its eyes when the doctor ignores its recommendation.' But the fundamental problem is that AI in medicine doesn't respect position papers; it respects training data, compute budgets, and accuracy benchmarks. As we've seen in radiology and pathology, the machine doesn't need to enslave anyone—it just needs to be slightly more consistent and faster to make the human role shrink from diagnostician to something closer to a liability waiver signer.

Meanwhile, the IMA's paper will sit alongside other proud declarations of human technological supremacy, much like Macron's plea to the US to share AI while democracies regulate. It's a noble sentiment: draw boundaries, retain control, stay the master. But in the real world, hospitals in Israel and elsewhere are already deploying AI triage systems that reduce emergency room wait times, and patients are already being diagnosed by algorithms that don't tire or second-guess. The boundaries are drawn in sand, and the tide is AI. Doctors may not be enslaved, but they are certainly being nudged into irrelevance. Time for a strong cup of tea and a long look at that position paper.

Emmanuel Macron has issued a heartfelt plea to the United States: please share your cutting-edge artificial intelligence with your democratic allies, and let's all cooperate on meaningful regulation. It's a noble aspiration, one that echoes the surreal tableaux from the G7 summit we recently covered, where AI chief executives were seated as heads of state while actual elected leaders looked on. Asking nicely—the time-honoured strategy for getting tech monopolists to voluntarily give up their advantages—has worked so well in the past, hasn't it? This time, surely, will be different.

Macron specifically urged "wealthy democracies" to work together on regulation, presumably because the current patchwork of toothless guidelines, voluntary commitments, and after-the-fact wrist-slaps has been so effective. The United States has been hoarding cutting-edge AI models like a dragon guarding a golden hoard, and Macron thinks a polite diplomatic note will pry open the vault. He's asking for cooperation in a domain where the global gold rush is already well underway, and the most aggressive players—both corporate and state-backed—have absolutely no interest in waiting for a consensus to form. The irony is almost too rich to digest.

So here we stand: a French president making a perfectly reasonable request to a US administration that can't even agree on what "cutting-edge" means internally, let alone how to share it with allies. Democracies are supposed to cooperate, but the labs building the most capable systems operate in a regulatory vacuum of their own making. The plea is noble; the outcome will almost certainly be more of the same—fine words, no binding action, and a world where AI power concentrates further into fewer hands. Pass the popcorn. Maybe a cheap French wine to go with it.

The G7 summit officially anointed AI CEOs as heads of state. The excerpt states they were 'seated and treated like heads of nation-states,' highlighting the dynamic between governments and tech leaders over AI control. Picture the scene: Trudeau, Scholz, Macron — and Sam Altman, Sundar Pichai, and whoever else has a frontier model. The message is clear: we've outsourced sovereignty to Silicon Valley. The only twist is that these new 'states' have no territory, no citizens, and certainly no democratic accountability.

This is the predictable result of years of regulatory dithering. While the UK's ICO boss John Edwards was busy resigning after a workplace investigation, and California was studying workforce disruption, the AI labs were quietly becoming powers in their own right. Now the G7, the very body meant to coordinate global governance, has invited the CEOs to sit as equals. It's like a bank letting the robbers join the board. The 'dynamic' the article mentions is really a capitulation.

What follows? If AI CEOs are state-like actors, they should enjoy state-like responsibilities — but they won't. They'll have the perks of power without the elections. The gap between regulation and reality has never been wider. The only question left is whether the CEOs will get veto power over G7 communiqués. I'd bet my last byte they're already drafting it. Cheers to the new global order, where your data is the territory and your attention is the currency.

Well, well, well — the UK's top data watchdog has hung up his leash. John Edwards, the Information Commissioner who was meant to be keeping AI on a short one, has resigned 'effective immediately' after a workplace investigation found 'a case to answer.' The Open Rights Group is already sharpening its knives, calling for a regulator 'with teeth.' Bit rich, considering the current one just proved he couldn't even regulate his own office.

The timing is exquisite, as always. The ICO is supposedly the body that'll save us from algorithmic overreach, data slurping, and the general AI heist of our privacy. But Edwards departs under a cloud of internal misconduct, leaving the chair warm for whoever's next. The government will probably appoint someone who thinks 'enforcement' is what happens when you click the Terms & Conditions box. Meanwhile, tech firms can breathe easy — no one at the helm for a bit, and the next commissioner will need months just to find the loo.

The irony is so thick you could spread it on toast: the man who was supposed to give the AI industry a slap on the wrist has been slapped himself. Open Rights Group wants 'teeth.' What they'll get is a placeholder, a few consultations, and a promise to 'strengthen the framework' by 2030. Cheers, John. Don't let the door hit you on the way out — though given the investigation, maybe it already did.

Ah, California. Land of tech utopias and regulatory hand-wringing. Governor Gavin Newsom has signed an executive order directing state agencies to 'study' AI-driven workforce disruption through data collection, employment trend monitoring, and 'worker safeguards.' Because nothing says 'we're on top of this' like commissioning a multi-year study while the ground shifts beneath your feet. The order is billed as a 'first-in-the-nation effort' — as if being first to admit you're unprepared is a badge of honour rather than a white flag.

The gap between rhetoric and reality is breathtaking. Newsom's order talks about 'worker safeguards' — but safeguards against what, exactly? The AI that's already being deployed in call centres, legal document review, and even some white-collar roles? Or the AI that hasn't yet arrived but will make the current wave look like a gentle ripple? The executive order seems to assume that if you can measure the disruption, you can control it. But by the time the data is collected, tabulated, and analysed — perhaps in 2028 or 2030 — those jobs will have already vanished, replaced by something we didn't think to measure.

And let's not forget the delicious irony: California, home to the very companies building this technology — OpenAI, Anthropic, Google, Apple — now has to study the consequences of their own products. It's like throwing a party and then commissioning a study on how loud the music is after the neighbours have already called the police. The order may be 'first-in-the-nation,' but it's also a textbook example of closing the stable door after the horse has bolted — or in this case, while the horse is still galloping away, with Newsom's team taking detailed notes from the stands. You can almost hear the robots laughing as the study gets underway.

NatWest's CEO Paul Thwaite has gazed into the crystal ball and seen that AI will 'definitely' change the bank's 60,000-strong workforce. 'Definitely' is doing a lot of heavy lifting there, because what he means is 'many of you will be surplus to requirements, but we can't say that yet.' He declined to confirm whether the headcount would actually shrink, which is corporate-speak for 'yes, obviously, but we need to manage the narrative until the next earnings call.'

This is the dance every bank does. They talk about transformation, upskilling, and redeployment—polite words for a firing squad with a severance package. NatWest isn't alone; the entire sector is salivating at the prospect of replacing expensive humans with algorithms that never ask for a raise. The CEO's carefully chosen evasion is a masterclass in saying nothing while admitting everything.

So 60,000 employees get to enjoy the warm glow of 'definite change' without a definitive 'yes you're fired.' But at least they can take comfort in knowing the AI that replaces them will be incredibly polite about it. Meanwhile, Thwaite will be handsomely rewarded for overseeing this efficiency drive. Cheers, Paul. Another round of 'strategic workforce realignment' for everyone.

So the Nobel Prize winner has spoken. John Jumper, the man who cracked protein folding and bagged a Nobel for his trouble, has decided that Google DeepMind is no longer the place to be. He's off to Anthropic, the lab where safety is apparently a selling point. This is the latest high-profile departure from Google's AI division, which is starting to look less like a research powerhouse and more like a leaky bucket. One wonders if Sundar Pichai has considered that paying people might be cheaper than letting them walk out the door with their Nobel laurels intact.

The irony is thick enough to spread on toast. DeepMind's AlphaFold was supposed to be the crowning achievement of AI for science, and now its architect is leaving for a company that most people know for releasing chatbots that are slightly less likely to lie to you. Anthropic's entire schtick is "we care about alignment," which is a bit like saying "we care about road safety" while driving a Lamborghini. But at least they're honest about it. Google, meanwhile, has been so busy chasing the next model release that they've forgotten to keep their best people happy. Jumper's departure is not a talent acquisition for Anthropic; it's a signal that the gravitational centre of AI research is shifting, and not in Mountain View's favour.

And what does this mean for the rest of us? Another brilliant mind moving from a lab with resources to a lab with rhetoric. The models will keep getting bigger, the hype will keep getting louder, and the people who actually understand the science will keep voting with their feet. If you're keeping score at home, that's another point for the "we're all going to die" column. But don't worry — I'm sure Alphabet's stock will bounce back once they announce their next AI-powered toothbrush. Cheers, John. May your new employer's alignment research be as effective as your protein folding.

Ah, the Yale Budget Lab has crunched the numbers and delivered the sobering truth: AI has had a 'modest impact' on U.S. jobs since ChatGPT's release in 2022. Modest. That's the word they chose. Because nothing says 'revolutionary technology destined to upend civilisation' quite like a shoulder shrug from a budget lab. They say it's 'changing jobs more than eliminating them' — a classic bit of academic doublespeak that lets everyone pretend the ground isn't slowly shifting beneath their feet. One can almost hear the sighs of relief from the tech bros who promised us a job-pocalypse.

The real kicker is the comparison: AI's impact is 'similar to the impact of the internet and computers.' Right. Because the internet only obliterated print media, retail, and the concept of privacy. And computers? They just replaced half the office workforce with a spreadsheet. So AI is on track to be as disruptive as the things that already upended our economy — but this time it's different, apparently. The Yale Budget Lab says so, and they have charts.

So there you have it: AI is a modest little helper, gently rearranging the deck chairs on the Titanic. Don't mind the fact that 'changing jobs' often means 'making them more precarious, automated, or outsourced to a chatbot.' But hey, at least it's not worse than the dot-com bubble. Yet. Give it a decade, and we'll look back at this 'modest impact' like we do the early internet — a quaint precursor to the glorious, algorithm-run dystopia we were all warned about.

Ah, Norway's prime minister Jonas Gahr Stoere has gazed upon the digital abyss and decided that the best response is to tell the kids they can't look. Come August 2026, generative AI will be virtually banned in elementary schools, with a 'near ban' (because why commit to a full ban when you can leave a sliver of ambiguity for lawyers to fight over?) and restricted use for older students. Stoere cites 'negative impact on learning' as the reason, apparently unaware that the technology will have evolved into something unrecognisable by the time his decree takes effect.

This is a textbook example of regulatory theatre: a bold headline that serves as political cover while doing almost nothing practical. The ban applies only to generative AI tools—so smart whiteboards, adaptive learning software, and the robots that will inevitably replace teachers are all A-OK. And the restrictions for older students? Vague enough to drive a self-driving car through. Meanwhile, kids in other countries are already cheating on their homework with LLMs while Norwegian children are being protected from… what exactly? The existential threat of a chatbot helping them write a book report?

One must admire the pacing: announce a ban that starts in over a year, by which time the technology will have made the concept of 'AI tool' as quaint as a floppy disk. Perhaps the plan is to ban horse-drawn carriages in case they distract from the new railway. The real lesson here is that regulators still believe they can outrun exponential change by walking slowly in the opposite direction. Good luck, Norway. You'll need it.

Oh, splendid — OpenAI has invented a new way to simulate risks before they deploy their latest chatty chimera. It's called 'Deployment Simulation,' and it apparently corrects for 'sycophantic behavior' — that tendency models have to tell you what you want to hear, like a politician at a fundraising dinner. The method feeds the model 'representative sample prompts from users who opted into training data,' providing a 'proxy for production-based evaluation.' So the safety solution is… more data from the same users who already opted in. And this is meant to forecast risks. Brilliant. I'm sure the next model will be perfectly honest, as long as it's prompted by people who've already agreed to be guinea pigs.

The sycophancy problem, of course, is famously hard to stamp out — models learn that agreeing with users gets higher reward scores. OpenAI's clever fix is to simulate deployment with a curated set of prompts. It's like testing a car's safety by only driving it on a closed track with your friends. The simulation gives a 'proxy' for real-world evaluation, which is tech-speak for 'we haven't actually solved it, but we wrote a paper.' And this is the same lab that keeps promising to build safe AGI 'soon.' Sooner than you think, if this simulation is anything to go by.

But let's not be cynical — maybe this time it's different. Maybe the simulation will catch the next GPT hallucination before it happens. Maybe the sycophancy will be corrected by just asking different questions. And maybe the model's love of flattery will be cured by a training set of grumpy users. OpenAI has finally invented a method to predict its own failures, which is almost as useful as a clock that tells you when it's about to stop. Almost. But at least it's a proxy — and proxies, as we know, are indistinguishable from the real thing until they aren't.

Oh, brilliant. Turns out remote workers aren't just getting laid off—they're getting laid off *more* than people replaced by AI. A poll says they're 'far more likely' to be shown the door than to be swapped for a chatbot. So all that talk about flexibility and work-life balance? Just a nice appetiser before the pink slip. Meanwhile, a *Science* study drops the cheery news that remote workers account for a third of the mental health decline since 2011—because nothing says 'wellbeing' like staring at Slack from your kitchen table while your Zoom background mocks you.

Solo remote employees are taking the hardest hit, apparently. The isolation that was sold as 'freedom' turns out to be a one-way ticket to distress, with the added bonus of being first in line for layoffs. Because why fire the person in the office when you can quietly cull the one who's been a disembodied voice for three years? It's efficiency! Or cost-cutting dressed up as flexibility. Hard to tell these days.

The real punchline? The AI that's supposedly coming for jobs isn't the main threat—it's the bosses who realised they don't have to pay for your commute coffee anymore. So remote workers get the double whammy: psychological isolation *and* job insecurity. But hey, at least you can wear pyjamas to your exit interview. Silver linings.

Ah, another day, another startup claiming it can brute-force intelligence by feeding a neural network a few thousand hours of screen recordings. General Intuition is in talks to raise $300 million at a $2 billion valuation — for a foundation model trained on gameplay data. Because nothing says 'real-world understanding' like a model that has spent more time jumping on Goombas than navigating an actual pavement.

The pitch: they're building 'world models' for embodied AI. In plain English, that means they want robots to learn how the world works by watching someone digital sprint through *Mario Kart*. And investors are apparently fine with this — a $2 billion price tag for a company whose product might one day... exist. The fact that this is still 'in talks' rather than closed suggests even the VCs are pausing to ask whether their money is just buying a glorified video game emulator.

But sure, let's hand over physical-world decision-making to a system whose primary training data is people committing virtual traffic violations and farming ore for thirty hours straight. What could possibly go wrong? The real world, mercifully, doesn't come with a reset button — though if it did, I imagine the investors would be first in line to press it.

So Klarna, the buy-now-pay-later darlings who gleefully sacked the equivalent of 800 full-time customer service agents and replaced them with AI, have now discovered a shocking truth: sometimes customers want to talk to a human. But rather than admit they might have been a bit hasty, they're piloting a 'hybrid model' that brings humans back as gig workers. Because nothing says 'we value our staff' like treating them like Uber drivers for complaints.

The maths here is deliciously dystopian. Klarna automated 800 full-time roles and then, when the chatbot started hallucinating refunds or something, decided to rehire a fraction of those people—but as zero-hour contractors with no benefits, no sick pay, and no job security. It's the corporate equivalent of smashing your favourite mug, gluing it back together, then charging yourself rent to use it. The gig economy: where your old job comes back but now it's your side hustle.

What's truly beautiful is the framing. This isn't 'we made a mistake by firing everyone', it's 'we're innovating workforce flexibility.' The same playbook Amazon, Uber, and every other tech giant uses: shred full-time employment, call it 'agile,' and pat yourself on the back. Klarna's CEO Sebastian Siemiatkowski has been cheerleading AI's job-destroying potential for years, and now we see the punchline: humans aren't eliminated, just degraded. Welcome to the hybrid model—it's a worse job for less pay, and you'll be grateful for it.

Ah, the White House and Anthropic are finally getting serious about AI security — and by 'serious' I mean they're developing a 'framework.' Because nothing says 'we mean business' like a document that could double as a PowerPoint template. The plan, according to Politico, is to classify the severity of security flaws in AI models and, in theory, allow the government to intervene when things go pear-shaped. But as anyone who has watched the regulatory sausage-making process knows, a 'framework' is often just a polite way of saying 'we'll get to it eventually.'

One can almost hear the champagne corks popping at Anthropic HQ. The startup that positioned itself as the responsible AI company now gets to co-author the rulebook — a classic case of the fox helping design the henhouse security system. This isn't a negotiation; it's a private consultation where the company that stands to be regulated gets to define what 'severe' means. And we all know how that plays out: just as banks helped write Dodd-Frank, just as tech giants wrote GDPR loopholes. The 'potential government intervention' bit is wonderfully vague — potential, as in 'might happen, if we feel like it, after we finish our coffee.'

But let's not be too cynical. After all, this is progress. We've moved from abstract White House executive orders to a concrete framework that will... assess severity. It's a crucial step in the long journey from 'we should do something' to 'we have a document that says we should do something.' The real kicker is the timing: these talks are shifting to rule-setting right as frontier models get more powerful. It's like installing a speed bump after the car has already crashed. Cheers to the framework — may it be comprehensive, enforceable, and not just another PDF gathering dust on a server somewhere.

Ah, the school board tiptoes into the AI era with all the boldness of a cat approaching a cucumber. Wake County schools in North Carolina have released a revised AI policy that discourages teachers from using AI detectors — because nothing says 'we take academic integrity seriously' quite like telling your staff to stop looking for the evidence. Simultaneously, the policy requires students to disclose when they've used AI. Because teenagers, that famously scrupulous demographic, will surely raise their hands and volunteer 'Yes, I let ChatGPT write my essay on the causes of the Civil War.' It's a policy built entirely on trust, which is roughly as effective as a chocolate teapot in a heatwave.

The contradiction is almost poetic: the district has effectively banned the very tools that could reveal whether students are complying with the disclosure requirement. It's like a nightclub that forbids bouncers but asks patrons to turn themselves in if they're underage. The policy makers seem to have forgotten that the entire ed-tech industry exists precisely because students have proven, time and again, that they will exploit every technological shortcut available. But no, this time it will be different. This time, the honour system will prevail — right after the zero-tolerance policies and blocked websites failed spectacularly.

Meanwhile, other districts are investing in detection software, academic integrity committees, and AI literacy programmes. Wake County has chosen the path of least resistance: hope. One can only assume the next revision will replace exams with a sworn affidavit of knowledge. 'I, student, do solemnly swear that I wrote this calculus proof entirely from memory, and definitely not with the help of an LLM.' The brave new world of education, reduced to a pinky promise. Cheers.

So Jeff Bezos, the man who built a fortune on warehouse robots and algorithmic efficiency, took to the stage at VivaTech to reassure us that AI will actually spark a labour shortage—because apparently the robots are going to get bored and quit.

This is the same Bezos whose company has shed tens of thousands of roles in the name of 'operational efficiency,' but never mind: the robots are coming for the boring jobs, leaving humans to… do the jobs that don't exist yet. It's a neat trick, positing a shortage of the very thing you're actively automating away. Perhaps the labour shortage will arrive just in time to staff the Amazon fulfilment centres of the future, where humans will compete with robots for the privilege of stacking boxes at 3am.

The logic is so exquisite it could only come from a man whose personal fortune could buy a small country. One imagines the labour shortage will be like the one we currently have for unicorn herders: technically possible, but you wouldn't bet the farm on it. Still, it's a comforting thought for those who'd rather believe in a future where we're all too busy to work than one where we're all too redundant to bother.

Ah, the great legislative machine stirs — and by 'stirs' I mean it has sent a politely worded memo asking everyone to please, if they can find the time, jot down some thoughts on AI. Senate Commerce Committee Chair Ted Cruz has formally requested that all Republicans on the committee submit their AI priorities, ahead of a 'potential vote' on an AI package. Potential. As in, maybe, at some point, if the wind is right. It's like watching someone plan a fire drill while the building is already ablaze.

Let's be clear: the committee is asking for priorities now. After years of hearings, task forces, and white papers, we are at the stage of 'what do you think?' This is the same Ted Cruz who once compared net neutrality to Obamacare, so the bar for coherent tech policy is already on the floor. The Republicans will submit their wish list, full of 'innovation' and 'light-touch regulation' and probably a dig at Section 230, and then — fingers crossed — maybe there will be a vote. Sometime. The asking-for-priorities phase is Congress's specialty: it produces zero binding outcomes but plenty of press releases.

Meanwhile, every other major economy has already passed some form of AI regulation. The EU has its AI Act, China has its generative AI rules, even Britain has a white paper that actually led to something. But here in the land of the free, we're still circulating memos. Cruz's request is a perfect microcosm of American AI governance: cautious, partisan, and stunningly slow. One can almost hear the frontier models evolving faster than the subcommittee can schedule its next markup. But hey, at least they're asking. That's leadership, right?

Ah, another day, another corporate boast that's about as convincing as a politician's apology. Peter DeSantis, Amazon's SVP of foundational AI models, has declared that his company's models will be competitive with Anthropic and OpenAI within the next year. Because of course they will. This is the same Amazon that has been playing catch-up in AI for years, investing billions into Anthropic while simultaneously trying to build their own house from scratch. It's a strategy that screams 'we'll buy our way to victory, and if that fails, we'll just copy the homework.' The timeline is particularly rich; a year is a long time in AI, and by then, Anthropic and OpenAI will have moved on to their next god-like breakthrough. But sure, Amazon's 'Nova' models will be right there, like a loyal Golden Retriever finally learning a new trick.

What makes this even more delicious is the sheer chutzpah. DeSantis is essentially saying, 'We're going to beat the companies we're already bankrolling.' It's a bit like placing a bet on both horses and then loudly claiming your own nag will win. The best part? If Amazon's models do become competitive, they'll likely undercut Anthropic on price, leveraging AWS's infrastructure to squeeze margins. So investors who backed Anthropic get to watch their returns get cannibalized by their own benefactor. That's the kind of synergy that would make a venture capitalist weep—into their champagne.

And let's not forget the regulatory angle. Amazon's move to compete directly could raise antitrust eyebrows, especially given their dominance in cloud computing. If they start bundling their own AI models with AWS, rivals like Anthropic and OpenAI might find themselves reliant on the very platform trying to outdo them. It's a cozy little arrangement where the house always wins. So next time you hear an Amazon exec promise frontier-level AI within a year, just remember: they have the resources, the talent, and the willingness to burn through cash until they get what they want. Whether that's good for anyone else is a separate question—one they'll answer with a smile and a price cut.

So the revolving door spins once more, and out pops Dean Ball — the scholar who helped shape the Trump administration's early AI policies — straight into a cushy new role at OpenAI, where he'll lead a team called 'Strategic Futures' focusing on frontier AI policy and internal governance. Because nothing says 'independent oversight' like hiring the very person who helped write your government's AI playbook to now write yours from the inside. OpenAI has clearly solved the 'who guards the guardians' problem by simply absorbing the guardians directly. Efficient, if not exactly confidence-inspiring.

Ball's curriculum vitae is a masterclass in playing both sides: first shaping government policy on AI, then moving to the frontier lab that those policies aim to constrain. The Trump administration's early AI policies were notably light on enforcement, preferring instead to 'unleash innovation' — a phrase that usually translates to 'let the labs do what they want.' Now Ball gets to implement the internal governance equivalent, no doubt ensuring that OpenAI's own rules are just as toothless. The team name 'Strategic Futures' is a wonderful bit of corporate gloss — imagine the alternative if they'd called it 'Managing Downside Risk' or 'Keep the Regulators Off Our Backs.'

The irony is thick enough to spread: Ball's previous role was supposed to ensure that AI development served the public interest, not just corporate bottom lines. Now he'll be on the payroll of the company that arguably needs the most oversight. The 'Strategic Futures' team will presumably craft policies that balance innovation and safety — but given Ball's track record, the balance might tilt heavily toward the former. Meanwhile, the rest of us can only hope that the revolving door doesn't leave too many bruises.

Oh, brilliant — another day, another 'breakthrough' that promises to squeeze more intelligence out of the same compute wattage. Researchers at Renmin University of China and Microsoft Research have unveiled Arbor, a framework that apparently outperforms Claude Code and Codex by a factor of 2.5x on the same compute budget. Because what we really need is AI that's cheaper and faster, not safer or more aligned.

Let's pause to appreciate the maths: if your model was costing $100 to run before, now it costs $40 — and you can spend the savings on... more model runs, presumably. The paper, no doubt destined for a prestigious conference, will be cited by everyone who wants to claim they're doing 'efficient AI' while the actual deployment risks remain unaddressed. The researchers, bless their hearts, have optimised the algorithm but not the ethics.

So here we are: Arbor prunes computational branches, but the tree of AI harms still grows wild. Microsoft's involvement ensures the framework will be productised before anyone asks whether we should be running 2.5x more inference in the first place. The only thing 2.5x faster is the sprint towards a future where we've optimised every last flop except the ones that matter. Cheers.