Three months, 604 commits and one live options book - building a production algorithmic trading system with Claude Code. Including, especially, the parts that did not work.
In May 2026 I set out to build an algorithmic options trading system, alone, using Claude Code as the engineering team. Three months later it holds 604 commits, 1,152 passing tests and four trading books - one of which places real orders on the National Stock Exchange every weekday morning without asking me anything.
This series is the honest version of how that happened.
It is not a tutorial and it is not a victory lap. Two of the books are dead, deliberately. The strategy the entire specification was written around was retired after seven weeks. Sixty-seven research studies mostly concluded “change nothing”. A measurement error flattered every result for months. Three separate bugs reached live money - each of them elegant, idiomatic code that was wrong about one fact of the physical world.
What emerges is a picture of where AI genuinely helps - implementation speed, exhaustive testing, tireless investigation - and where it does not: deciding what to build, knowing which numbers cannot possibly be right, and being willing to delete your best work.
If you lead product or engineering at a senior level, or you are trying to work out what to expect from these tools, the useful material is in the failures. They are roughly half of what follows.
This series was written with Claude, the same tool it describes. I supplied the project, the judgement calls and the arguments; Claude supplied the drafting, and dug every figure out of the repository’s own commit history so I could not flatter myself from memory.
Declaring that seems the least I can do given the subject. It would be a peculiar hypocrisy to publish eight posts on harnessing AI while implying I typed them all by hand. If the writing is good, that is partly the tool. If the judgement is sound, that part is mine. Distinguishing between those two things is, as it happens, what the entire series is about.
Day one produced three commits and zero trading logic. Why an agent without a constitution writes forty excellent files that quietly disagree - and the three fences that held for 604 commits.
Read this part → PART IIFifteen indicators in a week, all correct, and a conviction score of 91 on a flat market. What happens when a moving average, a MACD and an RSI are allowed to vote independently on the same fact.
Read this part → PART IIIFirst contact with the real tape. Partial fills, the seconds where a two-legged spread leaves you standing on one, and the unglamorous ninety days between “works” and “works unattended”.
Read this part → PART IVReal money, capped. Three live bugs in six weeks - an unwind that went briefly naked, arithmetic that produced a buy limit below the ask, and a restart that placed a trade nobody asked for.
Read this part → PART VThe keystone. Two months of backtests ran on prices I had partly invented. Buying 177 million rows of real option data inverted the answers - and showed the entire signal layer was worth nothing.
Read this part → PART VIEven with real prices underneath, the instrument was lying. The exchange changed how the close is set, and every historical result was quietly flattering by 6.5%.
Read this part → PART VIISixty-seven studies, mostly concluding “change nothing”. Three tests that kill good-looking results, and the eleven rejected improvements that produced the most valuable finding in the project.
Read this part → PART VIIIThe honest ledger: what the AI did superbly, what it got confidently wrong, and the decisions no model could have made. Seven things to take away if you lead product or engineering at a senior level.
Read this part →By day I run product for data strategy and operations at Condé Nast, where the brief is customer identity: the unglamorous business of establishing that the person reading on a phone in Mumbai and the one subscribing on a laptop in London are the same human being. Essentially the ‘slow work’ of turning unknown into known, in various stages. The glamorous parts of my day: developing the single customer view, and harnessing that data to optimise for amplified engagement and revenue across multiple lines and brands.
Twenty-four years of it now, across product, data and technology - client side and agency side, in media and publishing, CPG, insurance, automotive, FMCG and telecom, across North America, Europe and Asia. Enough time in front of CXOs to have learned that a business case travels further than an architecture diagram, and enough time behind them to know the diagram still has to be right.
This project was my evenings. It brings together the triumvirate - my love for the world of finance and markets, my drive to build a production grade system using the latest AI toolset, and the itch to discover first hand what these tools are truly capable of. And the only honest way to find out what these tools can carry is to hand them something that can lose real money.
Questions, disagreements, war stories from your own build, or a conversation about senior product leadership and AI-delivery roles - all welcome.