For every project in my portfolio, there are three that didn't make it. Ideas that excited me for a day, a week, sometimes a month, and then quietly died. Some were killed by technical limitations. Some were killed by market reality. Most were killed by my own taste - that nagging feeling, usually arriving around 2am, that the thing I was building wasn't quite right. Not bad enough to be obviously bad. Just not good enough to deserve existing.
Nobody shows you this part. Every portfolio is a highlight reel. The projects that shipped, the ones that looked good, the ones that generated the right kind of attention. But the graveyard behind the portfolio is where the real story lives. It's where you can see taste at work - not as an abstract concept, but as a practical editing tool. The ability to kill your own ideas is at least as important as the ability to have them.
The ones that almost lived
I'll tell you about five. Not because they were the best ideas - because they illustrate different reasons to kill a project, and each reason is its own lesson.
The Daily Digest. This was going to be a personal daily briefing tool - one page, every morning, consolidating everything I needed to know from across the internet. News from my RSS feeds, social signals from Slack, weather, schedule, tasks, all in one clean interface. I got quite far with it. The design was lovely. The concept was sound. But the execution ran into a wall: APIs. Every data source required its own API integration, and half of them were unreliable. The product would only work if every source was working every morning, and that felt fragile in a way I couldn't solve. I killed it not because the idea was bad but because the reliability was never going to be good enough. A daily briefing that works six out of seven days is worse than no daily briefing at all.
The Restaurant Ranker. Inspired by the fact that I eat out constantly in London and have strong opinions about every meal. The idea was a simple rating system - not a review site, just a personal ranking of every restaurant I visited, with a single score and a one-line verdict. The problem was me. I'm deeply uncomfortable publicly ranking restaurants because I know how hard the industry is and how much effort goes into every meal. Even a mediocre restaurant represents someone's dream. Ranking them felt mean. So I pivoted to Oishii London, which is about celebrating Japanese food rather than ranking all food, and that felt right. The edit was in the angle, not the format.
The Taste Graph. This one still haunts me. The idea was to build a visual map of how different cultural interests connect - showing that if you like this type of music, you probably also like this type of design and this type of architecture and this type of food. A network graph of taste. I still think the concept is brilliant. The problem is that building it requires either a massive data set I don't have or a series of subjective judgements that would be contentious. Who decides that Brutalist architecture connects to minimal techno? Me? That felt too arbitrary for a tool that was supposed to reveal objective patterns. I killed it because I couldn't find the right balance between editorial judgement and data-driven discovery. But I've not deleted the notes.
The Agency Autopsy. A blog series analysing famous advertising agencies that had closed - their best work, their worst decisions, what killed them. This one died because of tone. Every version I wrote sounded either smug ("I would have done it differently") or morbid ("and then they all lost their jobs"). There was no way to write about agency closures that felt generous and interesting rather than gossipy and mean. The subject was fascinating. The angle was impossible. Sometimes the problem isn't the idea - it is the voice.
The Micro-Museum. A series of tiny, highly curated digital exhibitions - five items each, on hyper-specific topics. "Five Tube Station Signs That Are Accidentally Beautiful." "Five Corporate Logos That Were Better Before." I loved this concept and I still think about it. But it needed photography I didn't have, and stock images would have killed the authenticity. A micro-museum with generic photos is just a listicle. I killed it because the gap between the concept and what I could execute was too wide. This one might come back when I'm better at photography.
Why killing matters
In advertising, the best creative directors weren't the ones with the most ideas. They were the ones who could kill ideas fastest. Who could look at a concept that was 80% good and know that the missing 20% made it not worth making. Who understood that a mediocre execution of a great idea does more damage than no execution at all, because it proves that the idea doesn't work when actually the problem was the execution.
Building products alone has taught me this lesson in the most visceral way possible. When you're the only person deciding what to build, you need a ruthless internal editor. Not a perfectionist - perfectionism is its own kind of paralysis. But an editor. Someone who can look at an idea honestly and ask: "Is this good enough to justify the time it'll take? Is this going to be something I'm proud of? Is this solving a real problem or just scratching an itch?"
Most of my killed projects failed the pride test. Not "is this technically working?" but "would I show this to someone I respect and feel good about it?" If the answer is no, it doesn't ship. This sounds obvious but it's genuinely hard to implement when you've spent a week building something and the sunk cost fallacy is screaming at you to keep going.
What the graveyard reveals
Looking at my anti-portfolio as a whole, I notice patterns. The ideas I kill tend to be the ones that are too broad. The ones that try to be everything for everyone. The ones that survive are specific, opinionated, and narrow enough to be excellent at one thing. Modern Retro isn't "AI art." It's a specific aesthetic applied to specific brands in a specific era. First Out isn't "a transport app." It's one piece of information - where to stand on the platform - for one city. Specificity is what survives the editing process.
I also notice that the killed ideas are often the most ambitious ones. The Taste Graph. The Cultural Interface Index. The Micro-Museum. These were the ideas that excited me most. But excitement isn't the same as execution. The distance between a thrilling concept and a finished product is enormous, and some concepts are simply too far from what one person with Claude Code can actually deliver. Ambition without execution is just daydreaming.
This doesn't mean those ideas are dead forever. Some of them are dormant. Some of them are waiting for the right moment, the right tool, the right collaborator. The anti-portfolio isn't a graveyard - it's a seed bank. The ideas are preserved, waiting for the conditions that'll let them grow.
But for now, the ones that shipped are the ones that passed every filter. The taste filter. The execution filter. The pride filter. The specificity filter. Fourteen products out of thirty-plus ideas. A 40% hit rate. That feels about right. That feels like taste working as it should - not preventing creation, but ensuring that what gets created is worth the attention it asks for.
Show me your anti-portfolio and I'll tell you more about your taste than your portfolio ever could.
