OBSERVATION I want to tell you about two phones in my house. Mine and my wife’s. Same household. Same wifi. Same algorithm logic running silently in the background. Two completely different traps. My wife wanted to improve her English. So I curated her Facebook feed — followed vocabulary pages, joined English learning groups, flooded her algorithm with educational content. A week later I checked her phone. Not a single English lesson. What I found instead was a masterclass in mutton curry. The algorithm had ignored everything I set up and served exactly what her behaviour had been quietly telling it for months. Not what I intended for her. Not what she consciously asked for. What she had already shown it she wanted — through watch time, scroll speed, every pause on every video — long before I arrived with my good intentions. I laughed at the time. Then I unlocked my own phone. I didn’t find food. I found Nepal. Years of engaging with political commentary had built a feed that was essentially a 24-hour stream of faction arguments, blame cycles, and debates that circled the same ideas with slightly different vocabulary each week. I thought I was staying informed. Looking back, I was cognitively congested. The algorithm didn’t know who I was trying to become. It knew exactly who I had been — at my most engaged, most reactive, most habitual. And it served that version of me every single day. WORKING THEORY Here is what I got wrong when I tried to hack my wife’s feed. I assumed that following the right pages and joining the right groups would signal to the algorithm what she wanted to see. It doesn’t work that way. The algorithm doesn’t listen to declarations of intent. It watches behaviour. Watch time. Scroll speed. What you pause on. What you click. What you return to at eleven o’clock at night when your energy is low and your resistance is gone. Liking an English learning page tells the algorithm nothing meaningful. Spending forty minutes watching a mutton curry video tells it everything. And here is the part that made me uncomfortable when I sat with it honestly. By the time my wife reaches for her phone, her Cognitive Capital is already spent. She has managed the baby, the house, the cooking, the cleaning, the garden — everything. The phone comes out in the small pockets between all of that. And in those small pockets, TikTok requires nothing from her. It reads her passive behaviour and serves accordingly. Facebook, where the English content actually started appearing after my curation effort, requires slightly more activation energy to open. Comfort wins. Not because she chose it over growth. But because the choice happens at the moment when there is nothing left to choose with. This is not a discipline problem. This is a design problem. The platform is specifically engineered to catch you at your lowest energy moment and serve you exactly what requires the least from you. MY OWN AUDIT When I started the Algorithm Audit I snoozed every political profile on Facebook. Deleted my YouTube history. Cut the feed at the root. The void arrived immediately. But then something real happened in Nepal that I genuinely needed to follow. I re-engaged — but differently this time. Deliberately. The significant moments only. The things worth actually watching. But when I returned to the commentators I used to follow closely, something had shifted. Not in them. In me. The commentary that once felt sharp now felt like a drama series I had seen too many episodes of. Same ideas. New events. Same circular conclusions. I realised I had my own position. Not just an opinion — a blueprint. A vision built not around any leader or party, but from the ground up. Starting at the local level. Let communities identify their own problems and solutions. Let those small units shape the larger vision — the way small water sources form a river, not the other way around. I had written it. Published it. Shared it where I could. The response wasn’t rejection. It was something quieter and harder to sit with. Everyone was busy with their own vision. The previous version of me would have kept fighting for attention in the feed. Kept shouting until someone listened. This version placed the work where it belonged and walked away. Not because I stopped believing in it. Because I did the maths. Building an early career in the UK, managing a full-time job alongside a toddler, running Saurav Insight — every unit of Cognitive Capital has somewhere more important to go than fighting for space in a feed designed to reward noise over clarity. That is not the season I am in. The previous version of me shouted on social media. This version knows the difference between a fight worth having now and a vision worth protecting until the conditions are right. That shift — that is what the audit actually produced. THE LINE THAT MATTERS The algorithm doesn’t know who you are trying to become. It knows exactly who you have been. EXISTING RESEARCH This pattern has a name in the research. Eli Pariser’s work on filter bubbles shows that algorithms create sealed information environments built entirely on past behaviour. The feed doesn’t show you the world. It shows you a version of the world constructed from your own history. Research on identity anchoring in recommendation systems shows that platforms optimise for engagement above everything else. And the content that generated the most engagement historically — your most reactive, most emotional, most habitual responses — is what the algorithm weights most heavily. Behavioural science adds another layer. Habits change slowly — far more slowly than intentions. Recommendation systems don’t just reflect who you were. They actively amplify existing habits because the algorithm rewards consistency of behaviour, not declarations of change. The platform and the habit reinforce each other until something deliberately interrupts both. I’m calling the result of all this digital identity lag — within this framework, it refers to the gap between who you are now and who the algorithm still believes you are based on everything you did before. My wife is not the person who spent hours on food content before we started paying attention. I am not the person who needed Nepali political commentary to feel informed. But the algorithm is still serving those versions of us. Because we haven’t fully shown it who we are becoming. And showing it requires consistent, intentional behaviour — not declarations, not page follows, not good intentions. Behaviour. Repeated. Over time. OPEN QUESTION If the algorithm only knows who you were — if it is built entirely from your most engaged, most reactive, most habitual past self — then the question isn’t how do you clean your feed. The question is: how do you show it who you’re becoming? And underneath that, a harder question. What happens when who you were is still partly who you are? My wife still loves mutton curry. I still care deeply about Nepal. The audit doesn’t ask us to pretend otherwise. It asks us to decide — consciously, deliberately — how much Cognitive Capital those older identities deserve in this season of our lives. And whether the algorithm is making that decision for us. From our cohort, the pattern is already visible. Participants who showed the strongest protocol compliance on Day 1 showed the highest professional alignment by Day 8. Participants who struggled to contain the home-country content pull on Day 1 disengaged from the study entirely. The feed knew their older selves too. Interesting pattern. Not proof. A pattern worth investigating further. INVITATION The Algorithm Audit Beta is still open. A seven-day field study measuring exactly where your Cognitive Capital goes before it reaches anything that matters. The Day 1 baseline takes ten minutes. If your feed feels like it knows a version of you that no longer quite fits — You already know what we’re measuring. Because you cannot update an identity the algorithm has already decided for you — until you show it something different. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit sauravinsight.substack.com
