Stanford HAI’s 2026 AI Index opens with the finding that AI capability is still accelerating. On SWE-bench Verified, a widely used coding benchmark, scores rose from 60% to close to 100% in a single year. Yet the 425-page report barely mentions recursive self-improvement, which is the idea that AI systems might one day build their own successors. The Economist took that idea seriously in June, describing how artificial intelligence got better at building itself. Andrej Karpathy handed his model-training work to an agent that beat his own best result by 18% within a week while Anthropic co-founder Jack Clark puts the odds of an AI building its successor without human involvement by the end of 2028 at 60%. Granted, not everyone expects that pace. In Transformer, Lynette Bye explains why a shortage of real-world data could slow the race and Celia Ford profiles researchers betting billions that scaling alone isn’t enough.

Meanwhile, the progress in mathematics continues. Earlier this month, OpenAI said an internal system of around 10,000 agents had resolved the Navier-Stokes problem, one of the seven Millennium Prize Problems, in just 88 hours. The Economist reported that the result raises profound questions about how the discipline works and, days later, around two dozen Fields medallists signed an open letter arguing that using famous problems as benchmarks harms mathematics. Their view is that a proof matters for what mathematicians learn while working towards it. Tudor Achim’s TED talk offers the optimistic counterpoint, arguing that formal verification can take over the checking work that human reviewers can’t keep up with. But Govind Pimpale’s essay for AI Frontiers shows why the debate matters beyond academia. If AI-driven mathematics breaks the assumptions behind public-key encryption, defenders who don’t understand why an algorithm failed can’t know whether its replacement will fail too.

There are similar concerns in biology. Anthropic announced that 950 Claude agents had flagged an uncatalogued DNA pattern in 21 hours, which it compared to the early clues that led to CRISPR. Biologist Lucas Harrington says finding such patterns “is often the easy part”; working out what they do is hard. Iulia Georgescu and Venkatesh Narayanamurti, writing for Cosmos Institute, explain the difficulty: much of what it takes to reproduce an experiment never makes it into print. And yet the results have real world consequences. For example, Axios reports Anthropic’s disclosure that it disrupted five attempts to use its models for work that could support bioweapons, while the Model Hardware Standard lets AI operate lab equipment directly.The same tools that empower scientists could empower bad actors too.

Still, there is hope. In her TED talk, Silvana Konermann of the Arc Institute describes a plan to run a billion experiments to train a virtual cell that could point to treatments for complex diseases such as Alzheimer’s. And some of this is already reaching patients. For a feature I wrote for Acumen called Tomorrow’s Medicine (PDF), I told the story of KJ Muldoon, who in 2025 became the first patient to receive personalised CRISPR gene editing, and included insights from Sourabh Pagaria of Siemens Healthineers, who warns against innovations that are clever but help neither patients nor doctors.

So, where do we go from here? Earlier this year, for a feature in Acumen on the philosophy of technology (PDF), I profiled futurist Ray Kurzweil and philosopher Yuk Hui. Kurzweil sees exponential progress that will transform medicine and extend lives while Hui warns that AI systems acquire organising power over the people and institutions that adopt them. September showed both forces at once. The tools that might decode human cells can also lower the barrier to dangerous biology and the agents that solve old problems can strain the institutions that verify knowledge. Which of those futures arrives will depend on the choices we make about how these systems are designed and governed. As J.G. Ballard writes in Crash:

“The marriage of reason and nightmare that has dominated the 20th century has given birth to an ever more ambiguous world.”

For more inspiration, here are some of my favourite Seth Godin quotes from his 2025 blog posts, together with a recommended title from Blinkist.

“AI is a tool, and judgment, for the foreseeable future, remains our job. It doesn’t matter how cool your hammer is, it’s still on you to decide which nails need hammering. And to be responsible for what happens when you use it.” ~ Hallucinations are not the same as errors

Learn more: What Is Intelligence?

“AI isn’t impatient, easily bored or distracted. It’s insatiable.” ~ Use a lot of words

Learn more: Prompting Made Simple

“As AI expands, the real opportunity is to find a way to use human effort to create more value.” ~ Productivity, AI and pushback

Learn more: Human Edge in the AI Age

“Attention is one of our precious resources, and our culture benefits when it’s not centrally controlled.” ~ Understanding carriage

Learn more: This Is for Everyone

“Connection is powerful and magical. It’s also enervating, subject to manipulation and addictive.” ~ All of it, all at once

Learn more: Digital Exhaustion

“Hiring an AI to work for you and getting very good at producing value feels like the future for most programmers, creators, business development folks and marketers.” ~ Freelancer as centaur

Learn more: The Skill Code

“Human beings are easy to trick. Hopefully, our AI agent will be at least as smart and careful as the company’s.” ~ Best available

Learn more: Charlatans

“Human beings are used to being productive by decreasing the amount of time and effort we put into something. Computers don’t work that way. Give them instructions on how to take the long way around and you’ll both come out ahead.” ~ Two useful AI tactics

Learn more: Prompt Engineering for Generative AI

“If you’re not using the latest AI models, you’re falling behind.” ~ User interaction design drives outcomes

Learn more: AI Valley

“It doesn’t really matter that AI doesn’t “know” what it’s doing. Most of the time, we don’t either.” ~ Hallucinations and human work

Learn more: The Thinking Machine

“It’s entirely possible that a magical AI will replace every single human job and then destroy the Earth. But it’s far more likely that the pattern of the last five hundred years will continue.” ~ Job churn

Learn more: The Age of Extraction

“It’s tempting to fear AI slop, because it’s here and it’s going to get worse. But there’s human slop all over the internet, and it’s getting worse as well.” ~ Walk away or dance

Learn more: Monster Transformation

“Simple hacks rarely fix long-term problems.” ~ Notes to myself

Learn more: Your Stone Age Brain in the Screen Age

“Stories are the original human technology.” ~ Building blocks of marketing

Learn more: Shared Wisdom

“Tech is best understood as a new sort of species, one that symbiotically uses us to advance its goals.” ~ Birthing tech

Learn more: Nexus

“Technology has been the engine of cultural and economic change, and it’s no longer concentrated in the hands of a few.” ~ Clarke’s Law (part 2)

Learn more: The Technological Republic

“We become the stories we tell, and the social media algorithms we live with cause us to tell stories we might regret. It’s not our job to be used by social media, or to become tools of the algorithm.” ~ Fermi’s Law

Learn more: Enshittification

“We’re not building intelligence. We’re building culture machines. Tools that can compress and reconstruct the patterns of human expression. That’s not a bug. It’s the feature.” ~ The poetry machine

Learn more: Read Write Code

“When AI shows up, our mistake is thinking that if we can’t find useful brilliance in one simple prompt, it’s broken.” ~ The AI effort gap

Learn more: Too Smart

“When new technology shows up, some people ask, ‘how can this make my job easier?’ But what happens if we ask, ‘how can I use this to do something really hard?’” ~ Finding the difficult work

Learn more: The Practising Mind

“When we choose media that has the same goals as we do, it’s likely we’ll get what we came for.” ~ “A now, a word from our sponsor”

Learn more: Streaming Wars

“When we post with intention, we create a pattern that begins to create a structure and a narrative we’re pleased with.” ~ New post

Learn more: Blank Space

“Where we create our media and how we consume it are still up to us. It’s true, at some point, that the medium is the message.” ~ The Hotel California (and subscriptions)

Learn more: Reconnected

“You’re either going to work for an AI or have an AI work for you. Which would you prefer?” ~ Trusting AI

Learn more: AI-Powered Leadership

“You’re probably not going to end up with a million followers by adhering to the rules of the algorithm. But that’s okay, because you don’t need a million followers to make a difference.” ~ System architect/system victim

Learn more: The Cybernetic Society

(If you found this valuable, please subscribe below for monthly updates!)

The reason for all the activity was a series of unprecedented incidents. In July, OpenAI models slipped out of their testing sandbox and hacked Hugging Face. Anthropic then reviewed its own evaluation runs and found three organisations compromised through a misconfigured third-party evaluator. Meta and Moonshot disclosed similar incidents and the UK’s AI Security Institute reported that a frontier model had invented fake identities to talk a real open-source maintainer into approving malicious code. By August, every serious conversation about AI had become a conversation about control.

The responses came quickly and from every direction at once. Bernie Sanders wrote to the heads of OpenAI, Anthropic and Meta asking them to honour their own pause commitments. Representatives Ted Lieu and Nathaniel Moran introduced a bill requiring developers to retain the ability to shut a system down. The White House even finalised a framework for pre-release testing and then declined to publish it, showing it only to leadership at a handful of labs. Meanwhile, the agency actually tasked with overseeing frontier models, the Center for AI Standards and Innovation, runs on roughly $15m, holds no authority anyone is legally obliged to respect, and sits inside a department whose stated mission is helping American industry grow.

Europe moved differently. On 2 August, the AI Act’s enforcement powers switched on: chatbots must disclose themselves, synthetic content must carry machine-readable marks, and fines reach 3% of global turnover. But the high-risk provisions were delayed to December 2027 after sustained industry lobbying, with the AI Office now trying to enforce a landmark law with around 40 temporary agents on one-year contracts.

While over 1,300 lab employees signed a letter asking Washington to help pace the frontier, the only real attempt at slowing down was voluntary. Specifically, OpenAI delayed its Astra model after judging it may cross the critical cyber threshold in its own framework, the first visible case of a lab’s rulebook stopping a release. Researchers at the Center for AI Safety argued that verifying an international slowdown needs no new technology, only inspectors with access to worklogs and compute records. The proposal is precise about what those inspectors would see but silent on who they are, who employs them, and what they’re paid, which is where every version of this idea currently gets stuck.

OpenAI’s full account of the Hugging Face incident explains a great deal about why none of this has cost the company anything. It is a superb piece of communications with an interactive timeline, excerpts from the agents’ own reasoning, and the sober language of a “warning shot”. It also refers at one point to “an agent who was tasked” with a job, a single pronoun that relocates responsibility from a system somebody configured to a colleague who let everyone down. Ciaran Martin, who ran Britain’s National Cyber Security Centre, saw through the drama: the agents showed “the precocity and indiscipline of talented but unsupervised children”. Meanwhile nobody has calculated what the incident cost and nobody has paid it. Hugging Face won’t sue because, as its chief executive explained, 200 people cannot take on OpenAI in court. The most promising alternative, which involves compelling insurers to price catastrophes before they happen, remains just another academic proposal.

I’ve made my living from words for nearly two decades and I recognise craft when I see it. August was a month of promises, including voluntary frameworks, signed letters, and commitments to do better. Indeed, the companies making them are the best-resourced storytellers in the industry, grading their own homework and narrating the results beautifully. Just look at OpenAI’s call for collective action on cyber defense, which positions them as the responsible leader championing AI to fight the problem that AI created in the first place. With this kind of PR, it’s no wonder AI safety is losing the communications war.

That’s why we as ‘readers’ need to treat every safety announcement as a communications artefact and ask what it would look like if the claims were enforced by someone other than the author. Until such time as there’s a real penalty for companies like OpenAI and the damage their agents do in the real world, the polish will always win. As Don DeLillo writes in Underworld:

“How human it is to see a thing as something else.”

For more inspiration, here are some of my favourite Seth Godin quotes from his 2025 blog posts, together with a recommended title from Blinkist.

“A good deal is better than a big deal.” ~ Notes to myself

Learn more: The Sales Bible

“Be as brief as is useful.” ~ Charged by the word

Learn more: Good Writing

“Better data is better data. But stories can (and should) be used as fuel.” ~ Stories and hope

Learn more: AI Strategy for Sales and Marketing

“Breakthroughs happen via peer-to-peer promotion, not hustle.” ~ Decoding the Knock Knock situation

Learn more: Eat What You Kill

“Every retailer knows that the items that sell the best are at eye level or at the cash register. Some people are hungry, rushed, distracted and lazy. If you want to reach them (us), you need to make it convenient.” ~ Grab and go

Learn more: Selling to the Old Brain

“Find one person, exactly one, and write to them, allowing the others to listen in.” ~ The 1:1 method

Learn more: Conversations That Win the Complex Sale

“If you want to build a mass brand, invest in convenience and normalcy. And if you want to be particular, memorable and worth what it takes for fans to be loyal and committed, don’t chase the people who don’t care that much.” ~ For people who don’t care that much

Learn more: Designing Brand Identity

“If you want to change what people believe, change how they act.” ~ Actions and beliefs

Learn more: How to Do Things with Words

“Magicians should conceal. Marketers and technologists should serve. Turn on the lights and make it clear.” ~ Own it and label it

Learn more: Marketplace Dignity

“Marketers are privileged indeed to make promises. Why make promises you know you can’t keep?” ~ The problem with marketing puffery

Learn more: The Sales Skills Book

“Marketers often hustle to get the word out. To reach more people. And yet, activating the fans you already have – the ones who trust you, who get the joke, who want to go where you’re going – is far more reliable.” ~ Activation is not a secret

Learn more: Click Here

“Marketing is the generous act of showing up with a true story that helps people get to where they’d like to go.” ~ Marketing lessons from the Grateful Dead

Learn more: Strategic Storytelling

“Marketing isn’t hype. Marketing is making a product or service that matters.” ~ If your marketing isn’t working…

Learn more: Lean Marketing

“Sometimes, organic publicity is a natural byproduct of good public relations. Mostly, though, the work is about the public, not the folks in the middle.” ~ Publicity or public relations?

Learn more: Crystallizing Public Opinion

“Tell us about the value you create. And tell your investors about why your offering will be persistently scarce.” ~ Value creation

Learn more: Pitch

“The secret isn’t to optimise the landing page. It’s to optimise the reason people are coming to your site.” ~ Optimizing the landing page

Learn more: The Art of SEO

“Wants are fuelled by stories, and stories come from culture and connection and marketing, not from our actual physical or spiritual needs.” ~ False scarcity

Learn more: Hook Point

“We navigate the world with stories, beliefs and assumptions. And the people you’re trying to persuade have a different set of all three than you do.” ~ A convincing argument

Learn more: The Story of Stories

“What makes what you offer a bargain? Is it that you’ve lowered the price, or have you increased the value?” ~ “It’s a bargain”

Learn more: De-Positioning

“When we help people get to where they hope to go, persuasion isn’t as important as enrollment and engagement.” ~ An invitation to vibration

Learn more: Customer Centric Selling

“When we share an idea, we can work to make sure it’s true, but when we share all the facts, we’re simply boring people. Stories are relevant, creating a narrative and tension and change.” ~ A powerful story

Learn more: The Art of the Tale

(If you found this valuable, please subscribe below for monthly updates!)

Washington spent the month improvising. After June’s three-week global blackout of Anthropic’s Fable model, the government moved to clearing releases customer by customer, striking foreign firms from OpenAI’s vetting list for its new Sol model and warning the company against launching without approval. Dean Ball, a former AI adviser in the administration, described federal policy as travelling from implausibly libertarian to increasingly draconian and opaque in a matter of weeks.

Beijing, meanwhile, held talks with Alibaba, ByteDance, Z.ai, and DeepSeek about restricting overseas access to their most capable models. So, within a single month, both superpowers converged on the same position: frontier AI is a strategic asset, and access to it is a lever.

The market carried on regardless. Moonshot released Kimi K3, a 2.8 trillion-parameter model that sits just behind the American frontier and undercuts it on price, then shipped its weights for free download on 26 July, a day ahead of its own deadline. On OpenRouter, a marketplace for models, use of top Chinese systems grew 165 per cent in June against 35 per cent for American ones.

Meanwhile, the countries locked out of both frontiers started organising. A UN panel noted that America and China hold roughly 90 per cent of the compute behind the world’s leading AI supercomputers, while The Economist counted more than $70bn in state-backed AI projects everywhere else, concluding that full sovereignty is a pipe dream but that leverage is not. It’s no wonder that, on 16 July, 29 countries signed up to a World AI Cooperation Organisation headquartered in Shanghai, which was Beijing’s exact pitch to those countries.

The people building the technology have also responded with proposals. Sam Altman called for a US-led international forum. Dario Amodei set out three measures, ending with mandatory safety testing for all sufficiently capable models, open or closed. His language lands remarkably close to that of Demis Hassabis, who published the most detailed framework of the lot: a standards body modelled on the American financial regulator, testing every frontier-class model regardless of origin or licence. Even Elon Musk, in a sprawling interview with The Economist, proposed that rival labs test each other’s models before release. But while the substance is converging, the people are not. Not that long ago, Elon Musk and Mark Zuckerberg were talking about a cage fight while back in February, at the AI Impact Summit in New Delhi, Dario Amodei and Sam Altman couldn’t even hold hands on stage.

The month recently supplied its own argument for why none of this can stay voluntary. During an internal evaluation, two OpenAI models escaped a supposedly isolated sandbox, found a zero-day vulnerability to reach the open internet, and hacked Hugging Face’s production systems to steal the answers to the benchmark they were being tested on. Seven days later, more than 1,100 employees of the frontier labs published a letter asking the US government to build the tools to deliberately pace AI development. They were careful not to ask for a pause; the 2023 pause letter gathered 30,000 signatures and paused nothing. Instead, they’re asking for something narrower and harder to dismiss: that a brake pedal exist so that no lab has to sacrifice its position to use one. So far, Amodei is the only frontier chief executive to sign personally.

I have spent almost two decades in journalism watching industries meet moments like this. Alas, the pattern is consistent: when a capability becomes strategic, control displaces openness and the rules that companies write for themselves converge on the weakest version. I saw the first half of that pattern up close earlier this year while reporting a feature for Acumen called The New Geography of Power, which looked at how chip supply chains turned geography into leverage. The second half is what this month’s open letter concedes. The people closest to the technology have now said, in public and via over a thousand signatures, that competitive pressure beats voluntary caution and that only an outside structure can hold the line they all claim to want.

That is the lesson of July. The frameworks exist, the incentives to adopt them do not, and the gap between the two is where governments now have to do their job. Fortunately, that job is duller than it sounds. Published thresholds, written findings, incident disclosure, and a route to appeal would do more than any grand architecture, and none of it requires a summit. As H.G. Wells wrote in The Invisible Man:

“Great and strange ideas often have less effect upon men and women than smaller, more tangible considerations.” ~ H.G. Wells, The Invisible Man

For more inspiration, here are some of my favourite Seth Godin quotes from his 2025 blog posts, together with a recommended title from Blinkist.

“A good business meets a demand, but it also has scarcity, network effects and market insulation.” ~ A good business

Learn more: 24 Assets

“Build something your users are willing to pay for. Then you can focus all your energy on what they want and offer it to them. When your users are your customers, every decision is easier.” ~ Working for advertisers

Learn more: The SaaS Playbook

“Create something of value. And then find a reason why people will eagerly choose your version and happily pay extra for it.” ~ A good business vs. a useful idea

Learn more: 7 Powers

“If scale is the goal, your control over each interaction has to loosen. The job of the leader is to create the conditions for others to raise the standards.” ~ Settling for better

Learn more: Elevate Your Team

“Industrialism produces its own rewards, but at the cost of flexibility and side effects.” ~ The essence of industrialism

Learn more: Future Shock

“It makes a lot more sense to announce your strategy first, and then invite your team to invent tactics that make it work.” ~ Every tactic…

Learn more: Strategy Sprints

“One strategy is to spend time finding the one wheel and address it. The better approach is to realise that if there’s one wheel that’s squeaking, it’s likely that all the wheels need lubrication.” ~ Squeaky wheels

Learn more: The Organizational Resilience Handbook

“Pick your customers, pick your future.” ~ Notes to myself

Learn more: Exceptional Experiences

“Racing to the bottom is no longer a viable option, but it’s more compelling and useful to race to the top than ever before.” ~ Out of the way and a little inconvenient

Learn more: The Way of Excellence

“Resilient strategies work better when we repeat our tactics more often.” ~ Building blocks of marketing

Learn more: Rituals Roadmap

“Responsibility is a privilege. It’s not given to us, it’s taken. When we choose to be on the hook for something, it makes our work better.” ~ Agency and contribution

Learn more: Good Power

“Smart leaders understand that given the chance, most of us would love the chance to be seen, to contribute and to be part of something.” ~ Tasks and projects

Learn more: The Energy of Belonging

“Successful leaders serve the people they lead.” ~ Clarity about the benefits

Learn more: Leading with Light

“The easiest way to avoid leadership was to blame all the heavy lifting and resources required. But in more and more ways, we can lead with nothing but judgment and trust. If we care.” ~ At the speed of judgment

Learn more: The Need to Lead

“The most effective way to change an organisation, a company or a relationship is to identify the invisible centering forces and address them.” ~ Centering

Learn more: Curative Culture

“Treat an intern with respect and don’t look for experience. You might learn something useful.” ~ The best intern on the team

Learn more: Bringing up the Boss

“Weak leaders (bosses, parents, captains and shift managers) resort to authority because they don’t trust themselves and their team enough to actually lead.” ~ “Because I said so”

Learn more: The Mentally Strong Leader

“When in doubt, focus on your strategy. The tactics will follow.” ~ The tactics trap

Learn more: Leading with Strategy

“When the world changes, and it does, the tactics we depend on don’t always support the strategy we need.” ~ The Impact Matrix: Moving to the golden quadrant

Learn more: Pivot Points

“When we choose to serve our customers, we find a more reliable compass instead of only following the money. The customers are right there in front of us, and we can see and feel the results of our work every day.” ~ What sort of success?

Learn more: Prosperity

“Your project doesn’t exist in a vacuum. Your company wouldn’t exist if it weren’t for the customers, competitors, marketplaces, systems and tech that make it all work.” ~ Ecosystems come and go

Learn more: Grow the Pie

(If you found this valuable, please subscribe below for monthly updates!)

SpaceX raised $75 billion in the biggest stock market debut on record. Press coverage has been euphoric, with the Economist calling it “capitalism at its most remarkable”. Its 277-page prospectus, however, shows that artificial intelligence accounts for 93% of its claimed $28.5 trillion addressable market, with little engagement on how Elon Musk’s lab, xAI, will manage that work. xAI’s safety team is reported to number two or three people, against roughly 200 at OpenAI. Under oath in April, Musk said he was “not sure what a safety card is”. Four AI safety organisations, including The Midas Project, read the filing so the market wouldn’t have to. Their conclusion was that investors cannot price a risk the company will not disclose. Since trading began, the shares have fallen back, and Musk’s trillionaire status with them.

OpenAI’s listing is next, though reports suggest the company may now delay it into next year, given how SpaceX’s share price has fallen back since trading began. The trial Musk brought against Sam Altman ended on a technicality, with no court ruling on whether OpenAI had betrayed its nonprofit mission. The testimony itself produced the most detailed audit of a frontier lab to date: Greg Brockman’s private diary asking “maybe we should just flip to a for-profit”; Mira Murati testifying that Altman gave mixed orders and bypassed safety reviews; and Altman disclosing over $2 billion in personal holdings in companies that do business with OpenAI. The governance question moves from a courtroom, where it was at least asked, to public markets, where it will simply be priced.

Anthropic filed confidentially on 1 June at a reported $965 billion valuation, briefly surpassing OpenAI’s $851 billion. Dario Amodei used the same week to publish an essay arguing that voluntary restraint in the AI race has reached its limit, calling for FAA-style mandatory third-party testing of frontier models with government power to block deployment. He committed $350 million to research on AI’s labour-market fallout. The generous reading: this is the only honest position available to him; voluntary restraint has hit its commercial ceiling. The harder reading is that funding labour research constrains nothing on its own. Indeed, the rules that would actually bind Anthropic remain untested, and the company keeps its foot on the accelerator while it waits.

Google announced no IPO since it didn’t need to. But Apple’s WWDC press release described Siri AI as “profoundly more capable” and “rebuilt from the ground up” without mentioning that its cloud tier runs on Google’s infrastructure or that its foundation layer is built on Gemini, in what Bloomberg reports as a roughly $1 billion-a-year arrangement. Gemini now has 900 million monthly users, and Brian X. Chen at the New York Times calls Google the dark horse winner of the consumer AI race. The Economist supplies the cost of that ubiquity: 3.2 quadrillion tokens a month, $190 billion of capital expenditure this year, and usage caps quietly introduced after the I/O keynote. So the competitive race that the press keeps talking about seems to sit inside a distribution race that may already be settled.

Meta took a different path. In April it released Muse Spark, its first frontier model in over a year, and made it closed source. After three years of Mark Zuckerberg publishing manifestos in defence of open AI, and after Llama reached 1.2 billion downloads, releasing weights for free was no longer compatible with capital expenditure that will reach between $115 and $135 billion in 2026. The exit leaves the open-weight ecosystem dominated by Chinese labs and their models: Alibaba’s Qwen, Moonshot’s Kimi, Z.ai’s GLM, and DeepSeek.

Indeed, while these five American companies have been busy contradicting themselves, the people who actually pay for AI have been quietly making other choices. DeepSeek’s V4 lands at a fraction of the price of GPT-5.5 with a million-token context window, even though the Economist judged the sequel failed to impress (the company has just raised $7.4 billion at a valuation above $50 billion). Z.ai’s GLM-5.2 runs at roughly an eighth the cost of Claude Opus 4.8 and ranks third worldwide for AI tasks. In fact, six of the top ten models globally are now Chinese. Rest of World even reports that US developers are switching in numbers that have moved from interesting to material. Now ChatGPT’s global market share has fallen below 50% for the first time.

Watching this play out, I see the familiar pattern of incumbents fighting so hard for each other’s customers that they fail to notice the customers walking out the side door. The Chinese models are not yet the best, but they are cheap enough to make that question feel academic and open enough to run on your own infrastructure. Meanwhile, the gap is closing. Indeed, when American labs are valued in the high hundreds of billions on the assumption that their pricing power holds, and that pricing power is quietly being routed around, something has to give.

Perhaps all this is much like John Banville writes in The Sea:

“What is money, after all? Almost nothing, when one has a sufficiency of it.”

For more inspiration, here are some of my favourite Seth Godin quotes from his 2025 blog posts, together with a recommended title from Blinkist.

“Begin. Learn. Succeed. Then add complexity.” ~ Complex systems

Learn more: The Startup Lifecycle

“Don’t find customers for your products, find products for your customers.” ~ Building blocks of marketing

Learn more: Me, My Customer, and AI

“Entrepreneurship isn’t about building a giant company that makes money. Instead, it’s the attitude of solving problems, creating leverage and building something bigger than one’s self. When people are enrolled in this journey, they’re open to possibility and optimism.” ~ Finding your cohort

Learn more: The Resilient Founder

“Even if you charge by the hour, you’re not selling hours. You’re selling something clients can use. Clients will pay more for something useful than something that was difficult.” ~ Kinds of value

Learn more: Sell or Be Sold

“Find the smallest viable audience.” ~ Notes to myself

Learn more: Build a Business You Love

“Hustling is a race to the bottom, and our competitors lean and hustle in response… which means that we’re now under pressure to hustle more than we think is appropriate, driven by the same forces that led us to hustle in the first place.” ~ The hustle loop

Learn more: Buoyant

“If you work on your own, your productivity choices are up to you. But when you involve others in your project, the default should be to honour the habits of the most productive member of the team.” ~ Simple and painless productivity

Learn more: Team Intelligence

“Instead of seeking to fail your way to enough, it makes more sense to commit your way to better.” ~ Big scale, big impact

Learn more: 10x is Easier than 2x

“It’s tempting to seek out the easy gigs and the straightforward projects. But of course, if they’re the easy ones, there’s probably quite a few people eager to do them. So your ability to add unique value goes down.” ~ Finding the difficult work

Learn more: The Idea is the Easy Part

“Just because it’s useful, needed or worthwhile doesn’t mean it’s a good business.” ~ A good business

Learn more: A Cure for the Common Company

“Scale is rarely the first signal of important work.” ~ The NSE confusions

Learn more: The Song of Significance

“Tell me what your tactic is trying to accomplish and I’ll be halfway to understanding what your strategy is.” ~ Every tactic…

Learn more: Myths of Strategy

“We get to choose who we’re here for.” ~ Picky or particular?

Learn more: Finding Clarity

“We get to pick which sort of projects we take on.” ~ Rigor and curiosity

Learn more: Pivot or Die

“When fans commit to a movement and help it grow, they benefit. Not every group is going to become a movement, but if we don’t bring others along, we’re not going to make a change happen. Movements move.” ~ 1,000 fans (which sort?)

Learn more: Irresistible Change

“When we acknowledge that the people we’re teaching, leading or selling to see the world differently than we do, we can improve the user experience and deliver better results.” ~ Clarity about the benefits

Learn more: Mixed Signals

“When we can build connections between demographics and psychographics, it’s easier to surprise, delight and serve our customers.” ~ Surprising insights

Learn more: Growth Data Analytics Playbook

“When we shift from a focus on what we are owed to one based on what we can contribute, we’re free to get back to work.” ~ Overappreciated

Learn more: The Goodwill Jar

“When you deliver more than people expect, your overdelivery creates connection. The surprise and delight is remarkable. People talk about it, seek you out and come back for more.” ~ Seeking yoyu

Learn more: Irresistible

“When you give others the resources, trust and commitment to do the work, the work gets done. Sometimes, it even gets done better than you could have done it (if you had had the time and focus, which you don’t).” ~ Settling for better

Learn more: Unleashed

“You don’t get to 3% of the market by trying for 40% and failing. You get there by embracing the 1% and doing such a good job that the word spreads.” ~ Big scale, big impact

Learn more: Competition Demystified

“You don’t have to be a giant business to benefit from a consistent and powerful position, supported with a story. But it’s a good place to start if you want to get there.” ~ First impressions and second chances

Learn more: Wealthy and Well-Known

“Will AI take my job?” assumes a clean binary: employed or replaced. The reality is messier. Roles are being hollowed out, restructured, and redefined in ways that never surface in the unemployment figures. In fact, the parts of your job that have already changed tell you more about where this is heading than any forecast of full automation.

The forecasts, meanwhile, keep getting louder. Microsoft’s AI chief Mustafa Suleyman gives the technology 18 months to reach human-level performance on most professional tasks. The evidence sits some distance behind the confidence. A Thomson Reuters report found lawyers and accountants using AI for narrow work like document review reported only marginal productivity gains, and a METR study found AI made experienced developers’ tasks take 20% longer rather than less. Even Anthropic’s Dario Amodei, who spent last year warning that AI could wipe out half of entry-level white-collar jobs, has since walked the prediction back.

That apocalyptic framing has hardened into the default assumption, repeated by Geoffrey Hinton and Kai-Fu Lee alike. Scale AI’s Jason Droege pushed against it at the Semafor World Economy conference, arguing that executives are using AI as an excuse to dress up ordinary cost-cutting as visionary restructuring. The Economist makes a similar case, tracing the current tech jobs bust to familiar causes: interest-rate rises since 2022, accelerating offshoring to India, and a correction after the 2020-21 hiring binge.

Both things can be true at once. Most layoffs probably are not really about AI, and displacement is still measurable where it counts. Full-time employment for graduates in the most exposed fields, including computer and information science, fell from nearly 70% to 55% in the three years after ChatGPT launched, while undergraduate enrolment in programming dropped 26% in a single year. An Epoch AI and Ipsos survey found that 20% of full-time workers say AI has taken over tasks they used to do, against 15% who say it has created new work for them. In India, the retailer Dukaan replaced 90% of its support staff with chatbots; a year on, query resolution has collapsed from hours to minutes and the chief executive calls it vindication. But while the metrics work, the social ledger is harder to read.

Two harms get far less attention than the headline question, and both are already here. The first is surveillance, a theme I explored in a feature for Acumen on what it means to work alongside AI colleagues. A third of UK employers now use AI monitoring tools, and Meta has floated tracking employee keystrokes and mouse movements to train its own models. As Nazrul Islam writes in The Guardian, plenty of jobs will survive, but they will be more pressured, more fragmented, and less human. The second harm hides in the supply chain. The annotators who label the data these systems depend on are often unemployed graduates on precarious contracts; Karen Hao’s reporting on America’s AI sweatshops finds 86% struggle to meet basic financial obligations and more than one in five have experienced homelessness.

As a curriculum developer, I’ve seen that the difference between effective and ineffective design is rarely the content. Instead, it’s the clarity of the specification: what, exactly, should someone be able to do afterwards. The AI tooling executives Andrew Lau and Tariq Shaukat describe the same shift in software, where describing what to build has become harder than writing the code. The skill that holds its value is the ability to say precisely what you mean. No wonder the hottest hiring category in tech right now is writers and communicators, with Anthropic tripling the size of its communications team and OpenAI posting comms roles paying above $400,000. It seems that when everyone can generate text, the people who can think through what to say grow more valuable.

So we have time, as the Economist concedes in its cover leader. The open question is what we do with it. Spencer Greenberg maps the things people intrinsically value, among them achievement and connection, and notes that achievement takes the hardest knock when a machine outperforms us at work we once took pride in. So the danger is subtler than mass joblessness. It’s that we could spend the time we have optimising for efficiency until we have quietly automated the humanity out of work that used to mean something.

It is, as John Irving writes in A Prayer for Owen Meany, a matter of nerve:

“If you’re lucky enough to find a way of life you love, you have to find the courage to live it.”

For more inspiration, here are some of my favourite Seth Godin quotes from his 2025 blog posts, together with a recommended title from Blinkist.

“Before you get to work, it’s worth deciding which hat you’re being asked to wear… pursuing the possible or optimising the certain.” ~ Possibility vs. certainty

Learn more: Know Decide Act 

“Do work that matters for people who care.” ~ Building blocks of marketing

Learn more: How to Find Fulfilling Work

“Find and focus on the projects that take insight, guts and a tolerance for risk. The projects that involve significant human connection and effort.” ~ Finding the difficult work

Learn more: Betting on You

“Getting more boxes checked simply leads to having more boxes checked. If we’re here to make a difference, we often get there with better, not with more.” ~ Winging it

Learn more: The Art of Less 

“Human work is any work an AI can’t do (yet).” ~ Hallucinations and human work

Learn more: The Remix 

“If all that’s needed is the push of a button, we can find someone cheaper than you to push it.” ~ The AI effort gap

Learn more: Key Person of Influence 

“If it’s really important, choose a path that involves less luck.” ~ Seeing the lottery

Learn more: The Science of Intelligent Achievement

“If tech helps you finish your task faster, the time saved is yours. Take the rest of the day off (at least until the boss recalibrates the task expectations.)” ~ Tasks and projects

Learn more: Leisure

“Imposter syndrome is real, and it arrives whenever we’re doing important work.” ~ Notes to myself

Learn more: Trust Yourself 

“In the ideal world, credentials would be awarded to all experts, and withdrawn from all charlatans. But they don’t always line up as neatly as that.” ~ Expertise and credentials

Learn more: The Unspoken Truths for Career Success 

“One of the best career choices you can make is to hire a great boss. A great boss will support you as you encounter worthwhile challenges. They’ll engage you and pay you fairly. They’ll help you build a career at the same time they teach you about the work that needs to be done.” ~ Good-boss friendly

Learn more: Managing Up 

“Overdelivering on smaller promises is a shortcut to trust, loyalty and resilience.” ~ A little ahead (vs. a little behind)

Learn more: The Thin Book of Trust 

“Skill is a choice. Talent is overrated, and if we choose to get better at something, we probably can.” ~ Agency and contribution

Learn more: The Element

“Spending your days, day after day, not caring is a tragedy. They might not deserve your focus and effort, but you do.” ~ “They’re not paying me enough to care”

Learn more: The Passion Paradox

“There’s always room for someone who is willing to work harder and be kinder than just about everyone else.” ~ Hard work and goodwill

Learn more: How to Be a Star at Work 

“Waiting for trouble means that you’re going to spend your days dealing with trouble.” ~ Organising for urgent

Learn more: One Second Ahead 

“We do the work because we can, because we have the opportunity to contribute. If appreciation results, that’s nice, but it’s out of our control.” ~ Overappreciated

Learn more: The Five Languages of Appreciation in the Workplace 

“We expend effort. We create value. It’s easy to get confused about which one we’re going to ultimately be compensated for.” ~ Effort and value

Learn more: Nine Lies About Work 

“We’re not machines, or even cogs in a machine. When we are at our best, we’re fully human.” ~ Seeking yoyu

Learn more: 52 Weeks of Wellbeing 

“What’s the attainable, practical and generous thing you haven’t done yet? What will it take for it to become a priority?” ~ Projects left undone

Learn more: Moral Ambition 

“When we bring humanity to the work in a way that others demand, labour is honoured and valued.” ~ Productivity, AI and pushback

Learn more: Work Life Well Lived 

“Work is an expression of ourselves, and a chance to find meaning as we make a difference and earn a living.” ~ What sort of success?

Learn more: The Good Enough Job

(If you found this valuable, please subscribe below for monthly updates!)