Every era has a culture, and every culture is a marriage. The last century married science to industry. This one is marrying art, science and code, and for the first time the machine is a participant in all three. I have worked in all three for twenty-eight years, usually at the same time. Marain is where that gets applied to a business, and then built.
That is the thesis. The rest of this is what it is built on, including the part I got wrong.
What a culture is
The word has been worn thin by people using it to mean office snacks and a tone of voice. I use it in the older sense. A culture is what a group has agreed to count, and the language it has agreed to count it in. Everything else follows from those two agreements.
A newsroom that counts page views produces one kind of journalism. A newsroom that counts corrections produces a different one, with the same reporters. A company that counts booked pipeline builds a different sales organization than one that counts the time between a customer's problem and its resolution. Once the counting is settled and the vocabulary around it is settled, most of the behavior is settled too, whatever the values poster in reception says.
This is why I have never been able to treat art, science and code as three separate careers, although my CV reads as though they were. Working out what is worth counting is an empirical problem. Making the count legible to the people who have to act on it is a storytelling problem. Building the instrument that does the counting reliably, without you in the room, is an engineering problem. Drop any one of them and you get a failure I recognize on sight. A finding nobody acts on. A story nobody can verify. A machine measuring the wrong thing to four decimal places.
I am not going to lead with the triad as a claim, because a claim like that sounds like range and range sounds like shallowness. Here is what each one actually was.
The reporting years
I was a journalist from 2001 to 2007, as reporter, producer and editor. I launched the first science and technology section in a mainstream Indian newspaper, at the Hindustan Times. At CNN-IBN I anchored and produced my own weekly science show, Connect, and worked the national desk, covering politics and the political economy alongside science and technology. Some of the reporting came from places where conflict was the story. Some of it came from a dog show. The beat did not matter very much, because the subject underneath every one of those stories was the same one, which is what people do under pressure and what they tell themselves about it afterward.
Then I helped plan and build a television channel, NewsX, and became one of the youngest executive producers in Indian television. In February 2008 the owners let five senior people go without notice, and I was one of them. I wrote a letter of protest that questioned where the channel's money had come from and said an investigation was called for. The letter was published. Within days the information and broadcasting minister asked the finance ministry to look into the channel's funding. Eleven years later that question was at the center of one of India's largest corruption cases. I was one of several people who raised it, and I raised it while it still cost me a job.
I bring that up because the same instinct runs through everything since. I name the uncomfortable thing at full strength, in public, while it still costs something. I do not know another way to do the work.
The code that runs at the statue
I have been writing software since I was thirteen. It started with Pascal, then C and Clipper for pocket money while I was still at school. In 1998 I built a content management system by hand for the first company I started. The system was acquired and the dot-com crash took the rest of the company. It was the first of four companies I have founded. Two of them were sold.
The piece of my own code I think about most often is not from any of them. When the Statue of Unity opened in 2018, the statue of Sardar Patel near Ahmedabad that is the tallest in the world, its ticketing and gates ran on software I had written in two days: C, embedded on a Raspberry Pi at every gate, each one talking to a central hub that held the financial records and logged every transaction. I do not know whether it is still deployed. It was small software. What interests me is where it sat, at the point where a person arrives to look at a work of art and a piece of code decides whether the gate opens. Most software that matters is like this. It is unglamorous and load bearing, and it is where an intention meets a crowd.
The models
The third strand came out of advertising and consulting, and it is really applied anthropology. What do people believe, what are they afraid of, what do they want to be seen wanting, and which of those can you move with a message. For some of the world's largest nonprofits and for governments I built mathematical models that mapped human behavior against communications. At the level of an individual person this is not science and I would not pretend otherwise. At the level of a population it produces enough signal to make fairly accurate assessments of which messages move people toward the choices you want them to consider, if not make.
The best example of it working is a stakeholder audit I ran for Malaria No More in India. The finding that mattered was not a health finding. It was a counting finding. Official records put malaria deaths in India in 2013 at 440. The government's own medical certification of cause of death report put them at 5,428. An Indian Council of Medical Research committee put them at 40,297. The Lancet put them at 205,000. The disease was being under-reported by somewhere between ten and a hundred times, and because state funding is allocated against reported incidence, the broken count was causing the funding failure, which was causing the delivery failure. The strategic problem was an instrument problem the whole way down.
The sharpest single piece of evidence in that study came from an accredited social health activist in Junagadh Block, Kalahandi, in Odisha. She was paid between twenty-five and seventy-five rupees for malaria work and better rates for maternal reporting, which meant she was financially rewarded for spending her time elsewhere and quietly penalized for finding malaria. A national policy adviser and a village health worker were interviewed for the same study, and the village health worker had the finding. I have never seen a reason to weight a source by its seniority when the evidence can be checked.
Now the one I got wrong.
In 2019 my company published seat assessments for Uttar Pradesh from models that scored hope, fear, aspiration and resentment. The assessments were badly wrong. The models had been signaling that fear and resentment should be weighted much higher than I set them, and I overrode them because I did not believe anger was running that hot. The signal was in the data and I discounted it because it did not match my read of the country. That is why every system I build now logs a human override with a written reason.
I am wrong in public when I am wrong. That is the whole point. The failure that actually costs you is never an instrument that is broken, because a broken instrument gets found. It is an instrument that is right in a direction you find implausible, which is exactly the moment it is easiest to talk yourself out of.
What changes now that the machine is in the room
Today's AI is a precursor of something larger, and I think the transition we are in is from the internet age to an age defined by the melding of human and machine. But the shape of the working relationship is already visible, and it is older than the current wave.
In 2019 I delivered a study of narratives and attitudes across twenty-nine European countries, 211 pages of it, using sentiment scoring, translation, dimension reduction and a concept engine, with human analysts reading the resulting maps to produce the actual insight. All the processing happened in the reader's own browser with no data transport to a server, which cost the user some performance and which I disclosed rather than hid. Machines read at a scale no team can match. People decide what it means. Where the data goes is a design decision you make on purpose.
None of that has changed. The machines have got very much better at the reading. What has changed is the price of the override, and I know that price because I paid it. When your instrument is fast, confident and fluent, the discipline that matters most is being able to write down, in a sentence, why you are choosing to ignore it, and keeping that sentence where someone can read it later.
What this means if you hire me
You get the judgment and the system, and they arrive together, because separating them is how most advice dies. I will tell you what I think is actually driving your business, including the version you did not want, and then I will build the thing that measures it so that neither of us has to take my word for it in six months.
There are people I am wrong for, and I would rather say so here than in a first meeting. If the decision is already made and what you need is a document that supports it, I am the wrong hire. If the answer has to arrive looking the way your board expects it to look, I am the wrong hire. If you want the recommendation but not the system that runs it, we will both be frustrated. And if the number you report cannot be questioned, then there is nothing for me to do, because questioning the number is the job. If you already know the answer, you do not need me. If you suspect the usual answer is wrong, we should talk.
The name
Marain is the language of the Culture in Iain M. Banks's novels. It is a designed language, built to be understood across every kind of mind the Culture contains, so that differences between them are understood rather than erased, and conflict gets resolved through a shared language instead of through everyone being made the same. I took the name because that is the work. Standing between the people who fund a thing and the people who build it, and between art, science and code, and making each of them intelligible to the others without flattening any of them.
That is also the only sense in which I am in the culture business.