Dominic J. ArrowArchiveContact

Software engineering and AI

by Dominic J. Arrow

start at (01)

(01)

Software Engineering

MAKE IT WORK. MAKE IT RIGHT. KEEP IT RUNNING.

Abstract ink figure for Requirements01Requirementswhat should exist, and for whomAbstract ink figure for Design02Designboundaries before codeAbstract ink figure for Implementation03Implementationwork, then right, then fastAbstract ink figure for Testing04Testingevidence, not hopeAbstract ink figure for Code review05Code reviewa second pair of eyesAbstract ink figure for CI/CD06CI/CDsmall releases, always readyAbstract ink figure for Deployment07Deploymentboring by designAbstract ink figure for Operations08Operationsread the live systemAbstract ink figure for Maintenance09Maintenancewhere the cost lives

A defect found in requirements costs a fraction of the same defect found in production.

Read the full lesson

(02)

Machine Learning

EXAMPLES IN. RULES OUT.

Abstract ink figure for Data01Datathe examples the machine seesAbstract ink figure for Model02Modela shape for possible rulesAbstract ink figure for Loss03Lossone number for how wrongAbstract ink figure for Gradient descent04Gradient descentdownhill, step by stepAbstract ink figure for Training05Trainingdescent at scaleAbstract ink figure for Generalization06Generalizationright on unseen dataAbstract ink figure for Evaluation07Evaluationthe exam it cannot memorizeAbstract ink figure for Inference08Inferencefrozen weights at work

One idea trains everything, nudge every weight slightly downhill, millions of times.

Read the full lesson

(03)

AI Engineering

YOU DO NOT TRAIN IT. YOU ENGINEER AROUND IT.

Abstract ink figure for Model01Modelrented, frozen, behind an APIAbstract ink figure for Prompt02Promptstate the task and doneAbstract ink figure for Context window03Context windowits entire working memoryAbstract ink figure for Context engineering04Context engineeringdecide what it seesAbstract ink figure for Tools05Toolsfunctions it may callAbstract ink figure for Agentic loop06Agentic loopact, observe, decide againAbstract ink figure for Harness07Harnessthe engineering around itAbstract ink figure for Subagents08Subagentsfan the work out

The same model in a different harness produces different results.

Read the full lesson

(04)

The Corpus

  1. Software EngineeringA plain language tour of requirements, design, implementation, testing, review, delivery and maintenance.
  2. Machine LearningA plain language tour of data, models, loss, training, generalization, evaluation and inference.
  3. AI EngineeringA practical map of the parts around a model that turn a request into dependable work.
  4. An AI found a WordPress zero-day for $25. Read the fine print.A researcher used GPT-5.6 to build a working WordPress Core exploit for about $25 - a bug brokers value in the hundreds of thousands. What is verified, what is his own account, and what actually changed for defenders.
  5. The closing cyber gap: open-weight AI is four to seven months behind the frontierThe UK's AI Security Institute has, for the first time, measured how far downloadable open-weight models trail the closed frontier on offensive-cyber tasks: four to seven months, down from six to ten a year ago. Why a measurable, shrinking gap changes what defenders should assume.
  6. WAICO: China launches a global AI governance body - what it can and can't doTwenty-nine states founded a China-led AI cooperation body in Shanghai on 16 July - without the US, EU, UK, Japan or South Korea. What a governance bloc with no binding power actually changes.

All 54 essays

(05)

Systems Thinking

Systems thinking treats a technology estate as one connected system: applications, data, infrastructure, and the teams that run them, tied together by dependencies and feedback loops. Behaviour emerges from those interactions, not from any single component. That is why a failure rarely sits where its symptom shows; it travels along the connections. Read the whole system first, and the fix is usually small.

Connect on LinkedIn

Dominic J. Arrow
Dominic J. Arrow