21 posts
The capstone: an AI analyst covering the top public and private AI companies, assembled from the series' disciplines — a deterministic pipeline, graded evidence, an agent that answers only from tools, and a judge that must prove it can fail. Part 6 of six weeks building AI systems by hand.
Evidence for AI agents: a correct answer scored as a failure, four searches that never found the right shelf, and why the filing, not the model, decides what your system can know. Part 5 of six weeks building AI systems by hand.
Assurance for AI agents: answer keys written before the run, judges that need judging, the audit trail that caught an agent manufacturing its own evidence, and why every number needs a reference class. Part 4 of six weeks building AI systems by hand.
The bill of materials for a working AI agent — model, skeleton, traces, evidence, protocol, memory, harness — and the two layers no vendor can sell you. Part 3 of six weeks building AI systems by hand.
Six weeks of building AI systems by hand, written up for operators. Part 2: what an agent may actually do, what actually binds it, and the five levers that decide how much autonomy to grant.
I'm a COO. I spent six weeks of late nights building AI systems alongside frontier practitioners, and caught my own agent fabricating evidence inside a perfectly formatted report. What the hours bought, and why the bar for 'technical enough' has moved.
I built a desk robot on a Saturday and gave it a brain made entirely of small open-source models. Then I named it badly. What a home robot taught me about open models, defaults, and the gap between demo and daily use.
I used Claude Code, a Mac Mini, and an Apple Ultra to build a personal health operating system — and an AI assistant called Zeus that messages me on Telegram when my body needs attention.
Six years between posts. Here's why I'm back.
Insights from 2 weeks at Stripe exploring best practices for building top-performing teams. Building teams is equal parts art and science.
Total addressable market can be a big challenge for startups. Here is how to approach TAM with a bottom-up methodology.
Everyone's first management role comes with growing pains. Here are 9 concepts every people manager should master.
A blueprint for how Europe can compete with Asia and the USA to build world-beating startups, focusing on venture capital, talent, and a single market.
Practical tools and frameworks for planning, goal setting, and measuring progress -- drawn from a decade of applying startup operating models to personal growth.
A look at the 16 books I finished in 2019, with a bias towards startup and philosophy books.
Reflecting on a year of change, growth, and the lessons learned along the way.
After nearly ten years managing partnerships and engineering teams at Twitter and Stripe, I enrolled in a coding bootcamp. Here's why a senior operator decided to finally learn to build, and what the experience taught me about engineering, learning, and the gap between managing builders and being one.
I read a lot of books in 2018. The majority on management and strategy; these books are the ones which challenged my existing mental models of the world the most
In any organisation you will encounter many different types of performers. There will always be a mix of A+, B and C type people. Something I've been preoccupied by is how to maintain originality in organisations. I believe this is made possible through malleable top-performers and organisational planning and foresight on how to nurture 'Originals' in cultures should be a focus for every leader.
App marketing has distribution and scaling challenges as a system. There are a tonne of supply constraints and demand scarcity economics which will impact the ability for newcomers to launch and distribute mobile applications. This is an except from a talk I gave at PGC Connects analysing the forces at play and potential routes to circumvent these.
A founder and operator guide to scaling tech companies courtesy of Andy Grove. Andy Grove's management philosophy is a long-time favorite of Silicon Valley founders. During my time managing teams at Twitter this management framework was launched in the company by Dick Costolo as part of Twitter's people management and operations philosophy.