What took so long

The bottleneck was never ideas. It was time, and more precisely, the cost of converting thinking into writing. For anyone in the trenches (or, the hamster wheel depending on your perception) - building software, churning out code and making product calls, writing consistently loses the prioritisation battle. Every hour spent drafting is an hour not spent on decisions that ship products or unblock people. Something had to change before this would ever happen.

Fifteen years across three companies gives you an accumulation. At Autodesk I learned what building software at scale actually means, the unglamorous mechanics of it, and that those mechanics matter. For someone with a non-computer science background looking to pivot to software, Autodesk was an incredible place to start off at. At Zendesk I got into product thinking properly, grew and led large teams across timezones, and spent six years figuring out how to keep a distributed organisation moving in the same direction. Now at Airwallex, I am working on Fintech infrastructure that runs across 70-plus countries, where data-discipline is not a principle you aspire to but a requirement you operate under. Every request you serve and every record you write needs to be 'correct'. The sum of all of that and a squeeze of B-School is a reasonably large stockpile of perspectives on building products, leading engineers, and scaling organisations, and for the first time I have a realistic way to get them out.

Here is a loose signal of what I expect to write about, not a programme.

• Platform and systems thinking: what building software at scale actually requires, not the textbook version
• Engineering leadership: the messy realities of scaling teams, hiring, and the org design decisions nobody documents
• AI in software development: what is actually changing in the craft versus what is noise
• Data-driven decisions: what it looks like in practice versus what people claim it looks like
• Occasionally: career, advisory, the business side of engineering

None of those topics would have appeared without a shift in what writing costs.

"Expensive" is not just the time it takes to type. It is the internal negotiation that happens before a single word is committed: is this idea worth two hours? Is the argument tight enough? Will the editing overhead justify it? For most ideas, across most weeks, the answer was no. The idea sat in a notes file or dissolved entirely. The bottleneck was not motivation or ability. It was the overhead cost of converting a half-formed view into something fit to be read by another person.

What changed is that I now have a collaborator on the execution side. The ideas still have to come from somewhere real: many years of specific decisions, specific frustrations, specific patterns noticed across top tier Software companies. AI is unlikely to be able to generate that. It may not have well-formed opinions about distributed team design or data-discipline or what actually changes when you do the hard yards in the business. What it does is lower the cost of articulating those opinions once they exist.

The better frame for this is leverage trade-off, not quality trade-off. Removing the writing tax means perspectives that would otherwise stay permanently in draft now have a way out. The constraint has shifted. The question is no longer whether you can absorb two hours of drafting overhead per idea. It is whether you have something to say. AI-generated content is easy to spot when there is no real perspective behind it. The ideas remain the hard part.

If something here resonates, or if something is wrong and you can show me why, either is worth hearing. No promises on how often this will appear.