Dave A'Hearne's Blog

Musings on London Tech Week 2026 Some musings about the events and talked that happened this year at London Tech Week.

AI Opinion

29 Jun

Deep Due: Part 2 - Teaching the Graph to Think Part two of a series building an agentic UK corporate fraud investigator — building in the investigation loop, using LiteLLM and Qdrant db to perform detailed investigations with contextual memory.

LangGraph Python Agents LiteLLM Vector Databases AI Engineering

Coming Soon

Deep Due: Part 1 - Data Aquisition & Graph Investigation Part one of a series building an agentic UK corporate fraud investigator — exploring the Companies House API, mapping the data graph, and laying the foundations of the LangGraph investigation loop.

LangGraph Python Agents AI Engineering

27 May

F1 Results Predictor: Part 5 - Automation, Drift & Monitoring Part five of a series building a podium predictor from scratch - A system that needs a human at every step isn't really a system. We ingest live race data, detect drift, and automate the path from new results to a promoted model.

ML MLOps Python Data Science

18 May

Programming Isn't Binary The agentic vs. engineer debate isn't wrong on either side. It's arguing about two different categories of work without admitting it.

AI Opinion Critical Thinking

Coming Soon

F1 Results Predictor: Part 4 - Productionizing Part four of a series building a podium predictor from scratch - A working model in a notebook is just a promising start. We version the data, manage the model, and build the system around it.

ML MLOps Python Data Science

11 May

F1 Results Predictor: Part 3 - Feature Engineering Part three of a series building a podium predictor from scratch - We instrument our training pipeline with MLflow, engineer a richer feature set, and watch what happens when we give the model something worth learning from.

ML MLOps Python Data Science

1 May

Eating Well A look at the hidden costs of what we choose to believe, and who benefits when we stop asking questions.

Opinion Critical Thinking

29 Apr

F1 Results Predictor: Part 2 - Simple Models Part two of a series building a podium predictor from scratch - Selecting an initial feature set, choosing LightGBM as our modelling approach, and establishing a no-frills baseline before any feature engineering or tuning.

ML MLOps Python Data Science

28 Apr

F1 Results Predictor: Part 1 - EDA and Baselines Part one of a series building a podium predictor from scratch - EDA, metric selection, heuristic baseline, and the start of the MLOps pipeline around it.

ML MLOps Python Data Science

27 Apr

Cold Turkey: Why I'm Stopping Using LLM Code Gen Why the illusion of productivity from LLM-assisted coding is making us worse engineers, and what I'm doing about it.

AI Software Engineering Opinion

15 Apr

Where Did My Tribe Go On the slow disappearance of the engineer who cares loudly - the shift from craft to performance, and where the hacker ethos went.

Culture Software Engineering Opinion

12 Apr

The Last Language You Can Read What happens to code quality and the engineering commons when AI is generating more code than anyone can meaningfully review.

AI Software Engineering Opinion

9 Apr

The Thinking Is the Work The eras of software, Reader Response Theory applied to code, and what it means when the thinking gets outsourced.

AI Software Engineering Philosophy

31 Mar

You Don't Need an SPA, Really A history of how React and Angular came to dominate, why they solved real problems at companies most of us will never be, and why you probably don't need one.

Web Software Engineering JavaScript

20 Mar

AI Is Worth the Water but It's Not the Product Neither the doom nor the utopia is accurate. On where AI actually sits as a tool versus a product feature, and what the hype cycle gets wrong.

AI Product Opinion

10 Feb