AI Mania Is Eviscerating Global Decision-Making

In one extreme case, I have seen an executive confess that they had never even used ChatGPT or any AI tool in their life, immediately after producing a technical strategy for an organisation with $2B+ in revenue which was entirely centered around AI.

Here's a report from an engineer at a company with a token leaderboard:

Checking out a parallel copy of our Go repository and telling the AI to rewrite the whole thing in Zig while I work on something else just so I can keep my job.

I particularly enjoyed this conversation with a skeptical executive at an over-enthusiastic company:

I asked why this was being repeated without opposition. Was it just sales fluff?

The answer was a lot more interesting. It was partially ridiculous sales material being delivered to an easily excitable audience, but this was not the dominant factor constraining honesty. Executives at their customers were saying absurd things about achieving 100x productivity, and this meant that if any executive at the vendor said that these gains were not plausible, it would undermine the credibility of the customer’s executive, be perceived as an attack (or heresy), and possibly result in an enterprise contract cancellation. And getting enterprise contracts cancelled because you wanted to opine on something that doesn’t really matter to your organisation’s mission is a great way to get fired.

Via Hacker News

Tags: ai, ai-ethics, ai-misuse

I’m announcing two new global feeds: a redesigned Global Shared Stories and a new feed called Good Reads.

Redesigned Global Shared Stories

Global Shared Stories has been in the sidebar for years, and for years it worked in a way I never liked. It was not the shares of everybody on NewsBlur. It was the shares of the accounts that the @popular account happened to follow, a list I put together by hand a long time ago and rarely touched. If you were on that list and you shared a lot, you decided what everybody else saw. If you shared one story a month and wrote a paragraph about why it mattered, you were drowned out by someone who shared thirty without a word.

So I rebuilt it. Every hour, NewsBlur now gathers every story shared across the whole site in the last few hours, caps each person at three so nobody can flood the pool, drops private blurblogs, and then picks a few worth reading. The picks accumulate, so the river stays deep enough to scroll back through.

The picking is done by Claude Haiku, once an hour, and it is worth being precise about what it does and does not do. It does not go looking for stories. It does not write anything. It only ranks stories that NewsBlur readers already chose to share, and the thing it weighs most heavily is the comment the sharer wrote, because a share with a few sentences attached is a share somebody thought about. It is allowed to pick nothing at all in a quiet hour, and in testing it regularly passes on half of what it is offered. If the API is ever unreachable, a plain heuristic takes over and the river keeps flowing.

Good Reads

The new feed in the sidebar is Good Reads, and it asks a question the other feeds do not: which stories did somebody finish and then do something about?

A story lands in Good Reads when at least two people read it closely, thirty seconds or more, and at least one of them then saved it, shared it, or trained it up. Finishing is not enough. Somebody has to have bothered to act. On top of that, the score is tilted toward feeds with few subscribers, so a story from a site with forty readers can beat a story from a site with forty thousand. That tilt is the whole point. The big sites do not need help getting seen.

Four feeds, four questions

There are now four curated rivers sitting together in the sidebar, and the reason there are four and not one is that they each answer a different question.

Global Shared Stories asks what people chose to hand to someone else. It runs on sharing, a deliberate human act.

Widely Read Stories asks what held the most attention across NewsBlur. It runs on reading time, not clicks, so a headline nobody read cannot buy its way in.

Long Reads asks what was worth an afternoon. Features and essays that readers gave real time to, rather than skimmed.

Good Reads asks what somebody finished and then kept, and leans toward the small sites you have probably never heard of.

Widely Read Stories and Long Reads have been around since April, and I wrote about how they work when they launched. None of the four ranks by clicks, and none of them is trying to keep you scrolling. They are all built out of what NewsBlur readers actually did with their time.

Because it was never obvious from looking at them which was which, each of the four now explains itself in a line at the top of its story list. It scrolls away with the stories, so it is there when you arrive and gone once you are reading.

Your classifiers still apply to all four. If you have trained a tag, an author, or a site, those green and red scores carry through, including on stories from feeds you do not subscribe to. And all four work as dashboard rivers, so you can park any of them next to your regular feeds.

Good Reads and the rebuilt Global Shared Stories are available now on the web. If a story shows up in one of these feeds that clearly should not have, I want to hear about it on the NewsBlur forum.

We built our companies to make software, because software used to be expensive to make. The teams, the roles, the handoffs, the sign-offs all exist for that reason. Now AI is making software cheap, and that raises a question nobody at the event was asking. If the thing you built the company to produce has suddenly become cheap, is the company still the right shape?

That was the one idea I took to Scott Logic’s “Adopting Agentic” evening for engineering leaders in Shoreditch. The rest of the night was rightly about tools and models. I wanted to talk about the organisation instead, because this is a design question, not a tooling one.

The Constraint Moved

An old idea from the 1980s explains what is going on. Goldratt’s Theory of Constraints says every system has one slowest part, and only that part sets the speed. A chain is only as strong as its weakest link. Speed up anything else and the whole thing does not get faster, the queue just moves somewhere new.

So think about coding, which we made faster. If you are not shipping faster, then coding was never the slow part. There is data on this: teams using AI agents wrote around eight times more code but shipped only about a third more.¹ That gap is the slow part showing up somewhere else. It has moved to the things AI cannot speed up, like deciding what to build, reviewing it, applying judgement, and getting people to agree.

You see it on the ground the moment the AI says “I can go faster if you like”, and then does. It floods you with pull requests, and every one of them still has to be understood, reviewed and judged by a human. The new bottleneck is deciding whether the code is any good, not writing it.

Two Paths

Every engineering org is now on one of two paths, and most have not noticed they are choosing. The two look identical at the start and end up somewhere completely different.

The first path is the easy one. Give everyone an AI, leave the org chart exactly as it was, and let each person get faster at their own bit. It feels like speed, but you have sped up every station except the constraint, so more code, more pull requests and more half-finished tools all pile up in front of the same review-and-decision queue. Over time it hollows you out, because when everyone is heads-down producing, nobody is doing the deciding, and that muscle wastes.

Taken to its logical extreme you get the “one-person team”, which is not a strawman. Coinbase’s CEO cut around 14 percent of staff and floated a future of one person being the engineer, the designer and the product manager all at once, with AI doing the rest.² It sounds bold, but a one-person team is oxymoronic and moronic. You have not built a team of superhumans, you have thrown away the one thing you now need most, which is people combining their judgement.

Amplify, Don’t Dissolve

The second path takes more thought. You redesign the team around the new slow part rather than bolting AI on. Teams get smaller, four or five people, and one person owns a whole job from problem to deploy rather than passing it down a line. Your platform team stops doing vendor plumbing and starts building your own agents, skills and guardrails, the intelligence layer no vendor can sell you. And you invest in the enabling teams who spread the working patterns, because they are the difference between adoption that compounds and adoption that stalls. I go much deeper on that team structure in my tortoises not hares write-up from Fast Flow Conf.

The common thread is to put your best people where the constraint actually is now, on direction and judgement. The developer stops being constrained by code, so what took six months takes six weeks, but they still need the designer and the product manager in the room. The product manager gets customer insight on tap and can prototype and validate in days. The designer is freed from fiddling with layout to do real design thinking, and keeps feeding the design system so it gets stronger rather than staler.

None of this asks for mythical people. Amplification asks your specialists to grow into more strategic, more demanding work, and you have to invest in that. The alternative asks for a unicorn who is brilliant at three jobs at once, and almost nobody is. Slow and steady, but amplified, beats fast and frantic every time.

Delegation, Not Abdication

Amplification only stays healthy if judgement stays in the loop, and that is easy to lose. The AI does an okay first job, so you let it do more, and more, until it is building everything and you are not even looking. That is abdication, not delegation, and you are still responsible for what the AI does.

For engineering specifically, engineers should happily let go of where the semicolons go and the latest syntax, but never give up architecture, simplicity, security, and being able to tell whether the code is any good. Let the mechanics go, and protect the judgement that made the craft worth doing.

So here are three things to do if you are the CEO of a product business. Find your real constraint rather than assuming it is coding, and measure shipped value, not lines of code. Choose amplification deliberately, because hybridisation is what you get by default when you hand out tools and walk away. And protect judgement as you scale, so that handing work to AI never quietly turns into nobody owning what it ships.

A hundred and fifteen years ago, Christian Larson wrote one of the first popular self-help manifestos. The Optimist’s Creed argued that it was a choice, and a useful promise. Not to promise the world, or the boss, or the market. To promise ourselves. Optimism is not a mood. It’s a discipline.

Last week, Reid Hoffman reminded us that the urge to build is also a choice. That we are homo techne, the species that shaped the tools and is shaped by them in return.

Each of these ideas argues that the future is not something that happens to us. It’s something we make, together, on purpose, or not at all. A potential promise, or a series of promises, that enable a better future.

In the words of Yoyodyne Propulsion Systems: The future begins tomorrow. Perhaps we can show up to make it better. In fact, we must.

Promise Yourself

1. To see optimism not as a prediction but as a choice. Pessimists are sometimes right, but they rarely build anything.

2. To remember that the future is not a place we’re going. It’s a thing we’re making. Every day, with every choice, whether we admit it or not.

3. To be so busy making things better that you have no time to explain why things can’t improve.

4. To understand that “it might not work” is not a reason to stop. Plan for the downside and commit to the contribution.

5. To trade the comfort of certainty for the possibility of contribution. Certainty is for spectators.

6. To be too generous for hoarding, too curious for cynicism, too committed for despair, and too busy shipping to permit the presence of Resistance.

7. To stop waiting to be picked. The world doesn’t care about your credentials. It only cares about what you create.

8. To begin. Before you’re ready. Because you will never be ready.

Promise the Work

9. To ship. Not because shipping is easy, but because unshipped work helps no one.

10. To make the tool serve the human, and not the other way around.

11. To remember that every tool is a teacher. The hand shaped the stone and the stone shaped the hand.

12. To fight and refuse “at scale” as an excuse for “without care.” Scale is a multiplier. It multiplies harm as it multiplies good.

13. To sign your work. Not to take credit but to earn trust.

14. To fix the thing, not the blame.

15. To honor the boring parts. Infrastructure is effort made invisible.

16. To know the difference between building something people need and needing people to want what you built.

Promise Each Other

17. To ask, every time, possible for whom? A lever that lifts only the people holding it is not a lever.

18. To be just as enthusiastic about the success of other builders as you are about your own. Scarcity is a story; possibility compounds.

19. To remember that four billion people got the phone before they got the library or the bank. The phone became both.

20. To teach what you know. Generosity is the only moat that makes the world bigger.

21. To take responsibility for the means as well as the ends.

22. To welcome the skeptic without becoming one.

23. To argue about the how without abandoning the whether. We can make things better. Let’s argue about how.

24. To measure what matters. It doesn’t matter how much money you raise, what sort of buzz you were able to generate, or which bridges you trolled under. What matters is the benefit created. Not engagement, but enrollment.

We are not users, we are people.

Promise the Future

25. To embrace the real choice between the possible and the likely. When your work has impact, playing the lottery is not a moral option. The downside may belong to other people.

26. To take the long view on the purpose and the short view on the action. Plant trees. Ship today.

27. To remember Socrates was right that writing would change memory. He was wrong when he insisted it would diminish us.

28. To notice that every era has its printing press, and every era has people who burn the books.

29. To hold the lever of possibility and technology with both hands. One hand for ambition, one for responsibility.

30. To remember the mistakes of the past, learn from them, and press on. Guilt is not a strategy. But experience, repair, and commitment are.

31. To realize that the whole world will never be on your side, and yet we must commit to building for the whole world.

32. To understand that this creed is not about technology. Technology is just the newest name for the oldest promise: that tomorrow can be better than today, and that it’s ours to make.

The stone is in our hands. It’s already shaping us.

What are we shaping back?

HT to Reid, Christian, and Kevin Kelly.