All of my long-form thoughts on AI, programming, product development, and more, collected in chronological order.
What changes when agents own strategy and execution? Legal structure, accountability, pricing models, failure modes, and the minimum governance to ship safely.
From UK home sales stuck in 12–20+ week conveyancing cycles, to hiring that drags ~40 days, to finance teams drowning in reconciliations, some sectors are structurally overdue for AI efficiency gains. We'll start with conveyancing/housing, recruiting discovery/matching, and accounting close/reconciliation - then add insurance claims, construction ops, healthcare admin, and public-sector back-office - showing where agentic workflows (document intake, contract checks, entity matching, claims triage) can cut weeks to days, and why the timing finally makes sense.
A tight, repeatable 10-day pilot that shows financial impact, defines stop/go cut-offs, and ends with a clean handover.
A production template for WhatsApp support automation: compliant flows, human takeover, full audit trails, and routing that respects SLAs and customer tiers.
A practical playbook to protect gross margin in e-commerce using rule-based pricing, automated invoice checks, and anomaly detection with clear audit trails.
A pragmatic, vendor-agnostic LLMOps architecture that ties traces, evaluations, guardrails, and hard cost budgets to releases - so you can ship safely and keep control.
A practical pipeline for high-quality Retrieval-Augmented Generation: remove duplicates, split semantically, fuse lexical + dense search, rerank, and measure.
If AI does most work, what does "income" become? This piece explores plausible UBI designs for an AI-heavy future - cash dividends from AI windfalls, "universal basic compute" credits, and data-as-labor royalties - grounded in what we've learned from real basic-income pilots and existing dividend models. It sketches how such systems might emerge, who funds them, and how access to powerful AI could be rationed fairly without stifling innovation.
What the EU AI Act requires in 2025–2027, how it lines up with the UK ICO's AI & Data Protection Risk Toolkit, and the exact outputs your team should ship.
Why skepticism improves ideas, why timing dominates outcomes, and why shipping many rough attempts increases your odds of a win.
