Complex problems. Simple solutions.
Systems architecture · Fraud · Risk · Security · Engineering leadership
Most read
what resonated the mostKnowledge inflation
LLMs stretched the scale. What you know stayed the same and it is worth less. On the trap of feeling close to expert while measuring against a bar that no longer exists.
Read → AI · QualityThe volume of code multiplied
Every week more code and more features ship, and the hours to review it are the same. The loop I use today, what I validate with, and what gets delegated and what doesn't.
Read → Career · BuildingLearning to build in public
In a room, confidence gets mistaken for knowing. On building in public, putting your work out into the world, and why doing it today costs almost nothing.
Read → AI · JudgmentThe model helps you build, but you decide
Anyone can build anything with an LLM, but not everyone builds something good. On the wave of generic products, why quality dropped, and why the model helps you build but doesn't decide for you.
Read → AI · CredibilityCredibility is the new diploma
Learning has never been this fast, so speed stopped being the problem. The new one is credibility: how you prove you actually know. An idea of where education can go.
Read → AI · VisionWhat matters now is your vision
Learning and building have never been this easy, so the scarce thing is no longer knowledge but vision and judgment. On learning like never before, crossing into other areas, and why what matters most today is knowing what to build and recognizing what generates value.
Read → AI · OpinionMy take on AI: notes from a paradigm shift
We're in the middle of a paradigm shift with AI. The technical excuses for resisting it keep changing, but the message underneath is always the same: fear that the time invested will lose its weight.
Read → Fraud · True Cost 8/8[The Real Cost of Fraud] For Every $1 of Fraud, You Lose Almost $4 (and You Only Measure the First) (Part 8/8)
The number your team reports as the cost of fraud is only a part. The real bill —false positives, team hours, fees, slow response, fragmented stack— is 3 to 4 times larger, and almost nobody adds it up.
Read → Fraud · True Cost 7/8[The Real Cost of Fraud] Rules, Lists, Velocities, Models, Graphs: The Problem Isn't Choosing, It's Combining Them (Part 7/8)
Why there's no Swiss army knife in fraud prevention — and how combining rules, lists, models, and graphs in a single system produces the quality data the engine needs to decide well.
Read → Fraud · True Cost 6/8[The Real Cost of Fraud] Chargebacks Are the Thermometer, Not the Disease (Part 6/8)
Why chargeback rate always reaches you late, why teams end up measured against the wrong metric, and what changes when you start thinking about fraud from the attacker's side.
Read → Fraud · True Cost 5/8[The Real Cost of Fraud] False Positives: The Damage Your Fraud System Does to Your Best Customers (Part 5/8)
Why false positives are the invisible cost of poorly calibrated fraud prevention, and how to measure them without sacrificing real fraud detection.
Read → Fraud · True Cost 4/8[The Real Cost of Fraud] Scaling the Analyst Team Doesn't Scale (Part 4/8)
Why 75% of the fraud analyst's day happens outside the fraud product, and what has to change for every team hour to land where it matters.
Read → Fraud · True Cost 3/8[The Real Cost of Fraud] Your Fraud Stack Has 5 Vendors and None of Them Talk to Each Other (Part 3/8)
Why having five fraud vendors that don't talk to each other costs you more than the fraud itself, and what a truly integrated stack actually looks like.
Read → Fraud · True Cost 2/8[The Real Cost of Fraud] If the Attacker Uses AI, Your Static Engine Has Already Lost (Part 2/8)
Why having a trained model isn't the same as having a fraud engine that learns from the adversary at the same pace the adversary learns from you.
Read → Fraud · True Cost 1/8[The Real Cost of Fraud] The Invisible Cost of Slow Response (Part 1/8)
Why the lag between detecting a new fraud pattern and blocking it in production is the metric that defines how much fraud actually costs you.
Read → Leadership · Culture[High-Performance Team] Culture, Technical Practices, and Reflections (Part 3/3)
The culture, technical practices, and principles I designed for a high-performance engineering team. Async first, radical candor, monolith first, and lessons learned.
Read → Leadership · Teams[High-Performance Team] Team Design and Tech Stack (Part 2/3)
How I designed the squad structure, defined the ideal developer profile, and planned the consolidation of a stack fragmented across 8 languages and 571 repositories.
Read → Leadership · Teams[High-Performance Team] The Diagnosis and Optimization Framework (Part 1/3)
How I diagnosed an engineering organization disconnected from the business and designed a 5-pillar framework to transform it into a compact, high-performance team.
Read → Leadership · Execution[Engineering Org Playbook] 90-Day Execution Plan for an Engineering Organization (Part 3/3)
Squad rituals, metrics-driven planning, prioritization system, DORA metrics targets, and career ladder. Part 3 of 3.
Read → Leadership · Metrics[Engineering Org Playbook] How to Build an Engineering Metrics System the Business Actually Cares About (Part 2/3)
Weekly KPIs, CTO alerts, connecting engineering metrics to business outcomes, and how to implement a dashboard without over-engineering it. Part 2 of 3.
Read → Leadership · Data[Engineering Org Playbook] How to Diagnose an Engineering Organization with Data (Part 1/3)
How to identify bottlenecks, calculate Revenue at Risk, and prioritize improvements in a fintech engineering organization. Part 1 of 3.
Read → AI · MCPMy POC with MCP: Connecting Claude with a Test API
Experimenting with Model Context Protocol to connect Claude Desktop with external APIs. A POC showing the potential of conversational interfaces.
Read → Open Source · DataspotBuilding dataspot: Lessons from Real-World Fraud Detection
The story behind creating dataspot - an open-source library born from real fraud detection challenges and sleepless nights thinking about data patterns.
Read → Fraud · GraphsGraph Contamination in Fraud Detection: The Problem Nobody Talks About
Why graph-based fraud detection ends up hurting legitimate customers and what the industry is missing about contamination by association.
Read →No posts found for your search.
Get new posts in your inbox
One email when I publish. No spam, no forced cadence.