Bigger Dreams: Refactoring Our Ambitions for the Age of AI

(Image generated by ChatGPT)

AI was imposed on us

There was no referendum. Nobody asked us, and we never got a chance to say no.

The frontier models are being imposed on us whether we welcome them or not... they arrive, they get embedded into institutions, markets, and daily habits faster than any prior technology, and the world after them will never be the same as the world before.

This is not a trend to be debated or a product to be adopted or rejected. It is an imposition, and the first thing to do is stop pretending otherwise. The question is no longer will this happen? The question is how will we live with it, and what will we build in it?

An interesting comparison

Let’s compare the “The same ordinary Tuesday” in 2019, 3 years before ChatGPT, and in 2027, 5 years after

2019, 9:04 a.m. A mid-sized company:

• The marketing department plans its year: Eight campaigns, each a month of agency meetings.
• The engineering team of nine is booked for the next two quarters on one feature.
• The CEO’s assistant asks for a market summary; the analyst has all week.
• The legacy invoicing app breaks again; nobody rewrites it. The estimate is “several millions.”

2027, 9:04 a.m. The same company.

• The marketing department is iterating on forty campaigns a quarter, and the senior creatives’ job is to make ten of them exceptional.
• Two engineers, directing a fleet of code-generating systems, are shipping the feature that was “booked for two quarters.”
• The market summary is a draft before the coffee is cold, and the analyst now spends the day on the questions, not the data.
• The invoicing app was rebuilt last month, in six weeks, for a fraction of the 2019 estimate.
• The CEO has stopped asking what is possible within budget. He asks what is worth doing.

The distance between the two Tuesdays is not an upgrade. It is a new world.

Many domains are impacted forever

The imposition reaches far more domains than is usually admitted, and the impact is not cyclical but structural. We are heading toward a world split in two:

• on one side, a stratum of high-end intellectual human craft;
• on the other, AI output for most things.

The machine handles the volume; humans retreat to the summit — and the summit is narrower than it looks.

Art is threatened. Not that art will die, but that the everyday production of images, music, and words will be flooded by machines. The object (the painting, the track, the manuscript) is no longer scarce. The hand that made it is.

Software engineering will probably disappear, as a profession as we know it. Not because software will not exist anymore, but because the act of writing it by hand will no longer be the bottleneck.

Science will be transformed radically, in method and in pace, because the cost of running an experiment, being computational, theoretical, or in the lab, is collapsing.

Imagining where we fit

In the domain of Art, AI will be able to generate endless videos, images, scores, and texts at near-zero cost. Humans will still have original creations and live performance.

In the domain of Software, all the process is being replaced by AI: from specification completion to architecture to code generation and testing. In a way, new people can ship software (they already do) without having the software engineers in the loop. Humans will stay with the framing of the products, the right questions to ask, the things not to do (really valuable in a world where everything is possible).

The Science world will also be put upside down and things have already started. Diagnostics support, drug screening, protocol drafting, are being taken in charge by AI, humans taking the bedsides, the difficult cases and the ethical decisions.

The Law & consulting domain also suffers a lot: Drafting, research, due diligence at scale, market analysis, everything can and is done by AI. But humans are still required for strategy, negotiation and, for sure, accountability.

Notice the pattern: in every domain, what is devalued is the production of the artifact; what is revalued is the judgment behind it, prior to the generation and after. That is the shape of the coming world.

We are going through the five stages of grief

So it is only natural that a great many people are now going through the five stages of grief, the Kübler-Ross model:

1. Denial,
2. Anger,
3. Bargaining,
4. Depression,
5. And finally acceptance.

Most of us are not yet in the fifth. Most of us are now in one of the four phases before it, and the four phases are easy to spot in the street, in the office, in the press, in the media, and in ourselves.

A fictional illustration of the “4 voices, 4 phases.”

1. Denial: the marketing director

   • It’s a bubble. The models are just autocomplete with a big vocabulary. Give it two years, the novelty wears off, and we’ll be back to normal. We already survived the ‘end of the web’ and the ‘end of the newspaper.’

2. Anger: the senior developer

   • Ten years I spent learning this codebase, the history, the traps, the people who wrote each module. Now a stranger can do in a weekend with a chat window what my team did in a quarter. And the people who should be embarrassed are the ones getting promoted.

3. Bargaining: the gallery owner

   • If the governments regulate it, if we can keep a core that is human, if there’s a label like ‘handmade’, ‘human-made’, the way there is for food, then it’s fine. We just need the rules to protect the craft. If that, we can live with it.

4. Depression: the illustrator

   • I opened my portfolio last week and I couldn’t tell which of these were mine. I used to know my lines the way you know a friend’s face. Now I look at my own work and I feel like I am an imitator of myself.

These are not irrational reactions. They are the grief of a world that is actually disappearing; and grief, by its nature, has a sequence.

We will enter the fifth stage — but it will be active, not passive

We will pass through these phases, and we will enter the fifth: acceptance. And here the model breaks. Kübler-Ross described the acceptance of the dying — a settling with the inevitable, a quieting. That passivity fits a grief that ends. Ours does not end. Ours is the grief of a transition, and when it is worked through, what comes out is not resignation but decision.

Our acceptance will be active, already is for people that digested the event. To accept the new world is not to resign to it; it is to look at the terrain that has opened up and to decide, with open eyes, what to build in it. And because the terrain has expanded, we will not merely cope with it: we will boost our ambitions.

Update of the four voices

Let’s come back for our 4 voices: “The same 4 voices, 6 months later.”

1. The marketing director (out of denial)

   • We stopped producing eight campaigns a year and started producing a hundred. The machine does the volume. My team of six is now a team of six directors: their job is to make ten of the hundred exceptional. I have more ambition for this department now than I ever did when I had forty people.

2. The senior developer (out of anger)

   • I stopped defending the codebase. It was never mine; it belonged to the company. Now I am the person who decides what the machines write, what gets tested, what deserves a full rewrite, and what never gets built at all. The job got smaller in lines of code and bigger in judgment.

3. The gallery owner (out of bargaining)

   • I stopped waiting for the rules. I sell the hand now: studio visits, the process, the person, the live unveiling. The object was never the point. The machine has done me the service of making that obvious.

4. The illustrator (out of depression)

   • I stopped illustrating. I direct images the way a director directs actors — I have a vision, I set the constraints, I reject ninety percent of what comes back. My portfolio is no longer a stack of drawings. It’s a body of decisions.

The pattern: each voice left the grief by finding its summit. The high-end human craft remains in its domain, and then, having found it, we can aim higher from there. That is active acceptance.

Our ambitions are currently the ambitions of the available means, and that is terrible

The tragedy of the present moment is that people’s ambitions are shaped by the means available to them.

We dream only of what is realizable.

We have internalized the budget: the estimate becomes the horizon, the headcount becomes the ceiling, the “it would cost too much” becomes the end of the sentence. We are a species defined by its capacity to want and we have reduced ourselves to wanting only what we can afford.

That is terrible. Not because the means are small (they are enormous, and growing) but because the ambition has stopped leading.

Ambition should be the variable that moves first; the means catch up. If much more is now realizable, then our ambitions should be much bigger: not 10% bigger, not a bigger version of the old ambition, but a different order of magnitude.

The ceiling of the possible has moved. The ceiling of the wanted must move with it.

The budget of dreams: the same €3 million company, twice.

A company with 3M€ in revenue and a tight budget. Same money, different era:

In 2019:

• Customer-facing software: One marketing website, €80k€, 4 months.
• Main application: ”We could rewrite it, but the estimate is 1.8M€… Hmm, we’ll wait”.
• Business processes: The digitization pilot was rejected in 2019 and again in 2022.
• Data: ”A data platform is for companies 10× our size”.
• Ambition level: One dream, deferred.

In 2027:

• Customer-facing software: Full customer portal + 5 internal tools, 90k€, 6 weeks.
• Main application: Rewritten, with test suite, 150k€, 8 weeks.
• Business processes: 23 processes digitized, 210k€, expected productivity gains > 1M€/yr.
• Data: A single data platform across the company, 120k€.
• Ambition level: Five dreams, done!

The budget barely changed. The ambition changed by an order of magnitude. Because the cost per unit of realization collapsed. This is what “the ambitions of the available means” looks like when the means change: it is not a correction. It is a reorientation.

Illustrations: software and science

The software industry

Instead of crying that software engineering is dead, ask what happens when the marginal cost of software falls very, very low. The profession may disappear, but the capacity it existed to serve does not: it multiplies.

The perspectives that open:

• I can change all my old apps, the ones I have no means to rewrite manually because it was simply too expensive;
• I can rewrite the applications that carry a high level of recurring costs because they were badly designed from the start;
• I can digitize many business processes that I could never afford to digitize; each one opening a new source of productivity for the company;
• I can optimize my full IT system and reconsider the systems architecture as a whole, instead of patching around it decade after decade;
• What I can do, small businesses that never had the money to digitize their business can now do it;
• And everything else that used to end with “not worth the cost.”

The graveyard of the legacy era is the launchpad of the next.

Sample

Let’s imagine the fictional “Halden Industries”, a mid-sized manufacturer, 1,400 employees.

2019 baseline: * Fourteen internal applications, three of them over 15 years old. Annual IT maintenance: 3.1M€. * A rewrite estimate for the core application, produced in 2016: 4.6M€, 22 months. * The decision, repeated in 2016, 2019, and 2023: "not this year".

2027, with near-zero marginal cost. * All three legacy apps rewritten: 380k€, five months, delivered by two engineers directing code-generating systems, including full test suites that never existed in the original. * The badly designed invoicing system rebuilt: it cost 180k€/year in patches and workarounds; the rebuild cost 60k€ once, and the recurring cost is gone. * 23 processes digitized that were always “not worth it”: shift handovers, quality checks, supplier on boarding, returns triage. Cost: 210k€. Expected productivity gain: about 1.2M€/year. * The full IT system re-architected as a whole: fourteen silos consolidated into a single data platform. The architecture was reconsidered from the top, something that was unthinkable while each silo cost a fortune to touch. * Its 40 small franchise partners, who could never afford digitization at all, now run on the same stack for under 5k€ each.

Result. The rewrite that was rejected three times was done in one quarter. IT headcount did not vanish: it moved from maintenance to product work. And the company’s ambition document for the next decade is now written in a different register: it no longer begins with what the budget allows. It begins with what the business should become.

Science

The same cost curve will boost science in both directions, and the boost is not incremental, because science is the one domain where the cost of an experiment is the binding constraint.

Theoretical sciences

AI as a tireless research partner: exploring vast hypothesis spaces, generating and checking proofs, formulating conjectures, running simulations that would take human lifetimes.

The bottleneck in theory was never intelligence; it was throughput. The throughput is no longer the bottleneck.

Practical sciences

Where each trial, each compound, each composition used to cost time and money, the cost falls, the number of trials multiplies, and the map of the possible is drawn faster.

Sample: “three laboratories, one cost curve.”

Theoretical: “the conjecture factory”

• An AI system explores 10,000 structured variations of an open problem in additive number theory. It proposes three new conjectures and candidate proofs for two of them. A human team of four spends 90 days verifying, repairing, and extending one of them into a publishable result. In the 2019 workflow, that 90 days was one postdoc’s year of work, assuming the path happened to be clear. The path-finding itself, the part that used to be luck, is now machinery.

Practical, drug discovery: “the compressed screening”

• A target protein with a plausible disease mechanism. 40,000 candidate molecules screened: 11 days of computation. 200 candidates sent to synthesis. 14 advance to in-vitro testing. 2 become lead molecules. The 2019 workflow for the same step: three to four years, roughly 60M€, and a much smaller initial set; because the set was limited by what the lab could physically handle, not by what chemistry could offer.

Practical, materials: “the cheaper cathode”

• A battery program. The system proposes 1,900 cathode compositions ranked by a model of energy density, cost, and supply-chain availability. 40 are synthesized in a pilot line. One delivers +17% energy density with a cheaper, more available supply chain than the incumbent — a result the 2019 program would have needed a decade of combinatorial screening to find, and which it was not funded to attempt.

The pattern in all three: the ambition of the laboratory was always bigger than its throughput. The throughput has caught up.

We need to refactor our ambitions

So the conclusion is not comfort, and it is not a strategy memo. It is a refactoring — in the software sense, the very sense this whole argument started from: the structure changes, the behavior must be preserved and improved, and the code is rewritten rather than patched.

We need to refactor our ambitions, because ambitions and dreams are the real characteristic of humans. Not our tools, not our techniques, not our degrees or our habits: our will. Every prior leap of the species happened when the wanted outran the realizable: the flight dream outran the glider, the moon dream outran the rocket, the connectivity dream outran the telephone. The law has never changed. Only the direction of the gap has.

For a long time, the realizable outran the wanted; and we shrank to fit our means. That era is ending. The realizable has jumped ahead, and the wanted is lagging behind it, still shaped by the old budget, the old estimates, the old “not worth it.”

“Refactoring: old ambition → new ambition.”

The old ambition (shaped by the old means):

• The engineer: “Land a job in a good team and keep my skills sharp.”
• The company: “Digitize our customer service this decade.”
• The artist: “Sell enough work to keep the studio open.”
• The nation:“Protect our industries from disruption.”
• The species: “Sustain our way of life.”

The refactored ambition (shaped by the new means): * The engineer“Be the person who decides what a team of ten thousand machines is allowed to build.” * The company: “Run an operation where every process is software and every system can be rebuilt in weeks.” * The artist: “Own a named hand in a world flooded with images, and make the name mean something.” * The nation: “Become the place where the most ambitious applied science is done, because the cost of doing it is no longer the barrier.” * The species: “Solve the problems we have always deferred: energy, disease, the climate; because we can no longer afford to defer them.”

The world has been imposed on us. We will grieve, and the grief is legitimate. But the grief has a shape, and the shape has an exit — the fifth stage, the one that does not end in silence. Acceptance will not be passive; it will be active. We will look at the world that was imposed on us, and we will answer it with the thing that has always defined us:

Bigger dreams, bigger ambitions.

Note

This article was generated by AI on a very precise script of ideas chaining, fruit of a reflection that lasts for some weeks now. It was corrected by me, a human. Most samples were provided in the prompt and they were developed and put in context by the AI.

In a way, this article is exactly the demonstration of benefiting from AI. In normal times, it would have taken me half a day to one day to put all that in something meaningful. Considering I am not an English native speaker, the English would have been painful to read for English natives. The saved time enabled me to work on other topics while having this message to deliver: important today but a triviality in some months.

(October 12 2026)