Essays, notes, and field reports on AI, engineering, and leadership.
AI Is a Car, Not the Destination
AI can take us further and faster, but humans still need to choose the destination, steer with judgement, and know when to take back control.
AI will not kill SaaS. It will sort it.
The companies best placed to kill SaaS are not doing it. That observation is doing a lot of work.
Sanding off the messy bits
Humans learn from AI. AI learns from humans. Somewhere in that loop, the texture that made our writing sound like us starts to disappear.
Your sprint planning is missing a column
Story points measure human effort. In an AI-augmented codebase, the bigger cost is model compute. Why we're still planning like it's 2022.
Ada Lovelace saw the whole argument in 1843
A passing quote at the Science Museum is the cleanest framing of the current AI debate, written before electricity was domesticated.
AI-assisted versus AI-aware
Using AI to get better at your current role might be the fastest way to make your current role obsolete. The difference is worth noticing.
What actually leaked when Claude Code leaked
The headlines said 'Claude source code is leaked'. The accurate version is narrower, more boring, and more useful. The distinction matters for how we think about AI moats.
Prompt injection is the risk we keep under-discussing
The model saying something wrong is a mild failure mode. The model being talked into doing something wrong is the one we're still not pricing properly.
The jagged shape of machine intelligence
AI writes strategy and fails at decimal comparison. The shape of that gap is informative, predictable, and mostly misunderstood.
Prompt engineering is becoming context engineering
The work has moved from writing the perfect instruction to assembling the right context. It's a plumbing job now.
Graceful degradation is the only AI strategy that survives contact with regulators
Why the first design question for any production AI system should be: what does this look like when the model is wrong?
You can hear which model wrote it
The stylistic signatures of major LLMs are now detectable enough that reading their output feels like reading a default voice. That matters more than most writers realise.
Late-night vibe coding has its own context window
I run Claude and Codex in tandem most evenings. Both models eventually hit their limits. So do I.
Don't build a business on AI getting worse
Cleanup businesses betting on the persistence of AI 'slop' are building on a shrinking market. The durable bet is the opposite.
The quiet moment when agentic AI starts working
Most new AI tools are incremental. A small number produce the feeling of crossing a capability threshold. The agentic release was one.
The person being referred is the one you want to hear from
Second-hand referrals through family or friends are well-meaning but structurally strange. The underlying question about agency is real, but it's complicated by cultural context.
Stop asking for status, start asking for evidence
The single change that unlocks speed in most large engineering organisations. A single page, five metrics, and the courage to defend it.
Back to sharing again
A short note on a quieter stretch, what it bought me, and what I plan to write about next.
A quieter end to the year
A short Christmas note on a full-on year and a few quiet days for family, food and conversations without an agenda.
Apple Intelligence as a layered bet
Apple isn't choosing an AI partner. It's building an architecture where no partner is critical and every layer is swappable. That's the whole strategy.
Leadership in twenty words
You can spend a year reading about leadership. The working definition I keep coming back to fits in a single sentence.
I got my hands-on coding back, with a bonus
Years of architecture and management had dulled my implementation speed. Copilots didn't just close the gap - they restored the flow, with experience attached.
The two coding copilots are teammates, not a winner
Comparing Claude Sonnet and GPT-5 on daily engineering work. The answer isn't a ranking. It's a division of labour.
Office gossip points the wrong way
People spend hours discussing departing leaders. The leaders who matter to your next two years are the ones just arriving. Redirecting attention is high-leverage.
Your organisation is already an agent. Your AI agents just have to fit inside it.
Most enterprise AI failures are not model failures. They are operating-model failures - the agent is smarter than the system it lives in.
The compliment nobody records
The best leadership feedback isn't in a performance review. It's what people in another team say about you when you're not in the room.
AI as collaborator, not mind reader
Users say 'that's not how I'd write it' and blame the model. The missing input is almost always context. Here's what context engineering actually looks like for your own voice.
Useful compounds. Important doesn't.
The most promoted people chase titles. The most irreplaceable ones don't. What looks like the fast track to seniority is often the slow way there.
Hiring smarter people is the easy part
The advice to hire people smarter than you is everywhere. The harder work is what comes after - and it's where most senior managers quietly fail.
Why I wrote The Decade Intelligence Changed
A note on the book - what it is, what it isn't, and why I wrote it outside work.
Live translation arrives at OS scale - six years later
Apple's WWDC demo of system-level live translation looks like magic. It's also functionally what we shipped at Zendesk in 2019. What changed is the platform, and what that tells us about timing.
I built a small app on my coffee break
Yesterday I skipped the usual scroll, opened Lovable.dev, and built the small thing that automates one of my errands. It was the most relaxed I’ve been all week.
What AI copilots can - and cannot - do for engineering organisations
Notes from embedding generative-AI tooling across a software development lifecycle: what worked, what didn't, and what we learned not to measure.
AI learning from AI: the photocopy problem
As AI-generated content fills the internet, new models are being trained on it. The photocopy analogy is real, but the timeline and impact are less uniform than the headline suggests.
When conversation starts sounding like PowerPoint
Catching yourself speaking in bullet points is diagnostic. Corporate communication has quietly trained us out of storytelling. The fix is harder than it sounds.
One bad meeting isn’t the end of your career
A short reminder I needed myself this week: a meeting that went sideways doesn’t outweigh the contributions stacking up quietly behind it.
The thing about past work is that it stays done
A year-end note on imposter syndrome and the one observation about past achievement that genuinely helps when the next thing feels too big.
Context engineering is the new management skill
The leaders who will thrive in an AI-native organisation are the ones who learn to manage context, not just people.
It's okay to put non-fiction down at halfway
Most non-fiction books peak around chapter six and elaborate for another six. Abandoning at halfway is often rational, not a failure of attention.
"AI-first" is a verb, not a slogan
Every team I've ever seen succeed with AI treats it as a verb - a practice applied to specific workflows - not a banner hung over the whole company.
On winning the Next CIO Award, briefly
A short note on the Next CIO Award, why it matters less as a trophy and more as a forcing function, and the people who made it possible.
Template first, content second (usually)
The choice between content-first and template-first presentation authoring affects more than productivity. It shapes your thinking and your defence against slipping deadlines.
Sora and the quiet economics of generated video
The headline is creativity. The actual impact is a sweeping shift in the economics of stock footage, training data, and visual credibility.
If you must use ChatGPT, prompt it like you wrote it
AI-drafted emails that arrive unedited are detectable, irritating, and gradually eroding trust in workplace communication. The fix is small. The discipline is harder.
Why a tyre company became the world restaurant authority
A tyre company invented restaurant criticism to sell more tyres. A century later it's the most authoritative dining guide in the world. The patience to play this long is rare.
Which language should I learn? The question beneath the question
The 'JavaScript for everything' joke lands because it's nearly true. The real question is what you're trying to build, and the answer has changed since 2020.
Plunge pricing: when the grid pays you to switch on the dishwasher
My energy provider just paid me to use electricity. The first reaction is that this is absurd. The second, more useful reaction is to ask what it tells you about how grids actually work.
Using ChatGPT as a 'shrink': useful, bounded, complicated
AI chat as low-friction emotional processing has real benefits and real limitations. The framing 'shrink on demand' oversells what it actually does well.
Casual Friday is a relic of an office most people no longer go to
A small note on the strange persistence of dress-code rituals when the office they were designed around has structurally changed.
Leadership communication isn't a soft skill - it's the work
Brisk, framework-laden communication beats vague substance only when the framework actually matches the work.
Open-book exams, AI, and what learning is actually for
The cognitive science of open-book learning has been clear for decades. AI extends the principle. The pedagogical establishment is still catching up.
If you're reading this on a weekend
A short note on the weekend-LinkedIn ritual, why it persists, and the more interesting question hiding underneath.
Leadership is a verb, not a title
The most useful thing I can say about leadership: it shows up in what you actually do, not in what is printed on your business card.
Staying yourself as you grow
On the quiet pressure to compromise who you are in pursuit of recognition, and why I think the trade is rarely worth what people are sold.
Transformation, the buzzword
A short rant about a word that has lost most of its meaning, and the small, unflashy work that the word is usually used to avoid.
What ChatGPT actually changed in week one
Reading my own week-one post about ChatGPT, three years on. The 'forever changed' framing was right in spirit. The mechanism was less obvious in the moment.
Are we re-solving live broadcast - or solving a different problem?
Broadcast TV solved one-to-many at scale 75 years ago. Streaming engineers keep re-solving similar-looking problems. The case for streaming is real but narrower than its enthusiasts claim.
A small plea to the people who build video calls
When someone raises a hand on a call, the person presenting almost never sees it. A short product complaint that has cost me more meeting time than I want to admit.
A small feature request for calendars
When you propose a new time for a meeting, the calendar should hold that slot tentatively. The current behaviour produces a small but persistent class of embarrassment.
The honest paradox of working from home
We say we are more productive at home. But we still want our kids in real classrooms and our holidays in real places. What does that contradiction mean?
Hiring senior engineering managers at BT
A short note from late 2021, when I was building out engineering management at BT digital. It is the sort of post that ages well precisely because it is plain.
Agility is not a vibe
In the summer of 2021 every leadership conference promised "agility". Most of them meant nothing more than "we are nervous". Here is what the word should actually mean.
When live theatre moved online
A short note from spring 2021, when an online performance event tried to bottle some of what 2020 had taken away. Worth keeping for the small reminder of how recent that all is.
My first flash sale: launching iPhone X in India
In November 2017, my Airtel team launched iPhone X in India alongside the global launch. It was my first flash sale. Here is what I learned that night and the morning after.