Document/state as of July 2026

About this site

What this is, how it was made, and why you can check every number on it.

Six essays on the state of software development with AI, written for working professionals — developers, tech leads, engineering managers, consultants. The central argument runs through every page: code generation became cheap, so the work moved to the edges — communicating intent upstream, verifying output downstream. The essays trace what that does to user stories, to specs, and to the shape of engineering discipline.

§1Openly AI-generated

This site is AI-generated and human-curated, openly. The essays were drafted by Claude from a researched source collection; the design, the build, and the review rounds ran through an automated pipeline (dark-website-factory) in which machine judges gated the intermediate stages and a person held exactly two gates — the brief at the start and the publish at the end. Stating this plainly is the author's professional position: AI-first work, disclosed as such, with its limits named rather than hidden.

§2Method

The raw material is a private collection of research on AI-era development practice, gathered December 2025 through February 2026 — largely AI-generated deep-research reports built around a handful of primary sources. That collection was condensed, and then, in July 2026, every article was re-verified against the live web: claims were confirmed, corrected, or flagged, each with a URL and date. The essays here are authored from that corrected July 2026 state.

Three provenance rules govern the prose:

  • Every quantitative claim carries an inline source with a name and date. Pages outside the essays — the home page, the 404 — state no figures at all, which is what makes the attestation in the margin checkable.
  • Claims the research could not verify are either dropped or visibly flagged like this — never silently promoted to fact. Vendor ROI numbers are labeled as marketing wherever they appear.
  • The author's own positions are marked as such — the blue "working thesis" boxes — and kept distinct from reported consensus. Corrections stay visible: where the early-2026 material was wrong, the essays show the strikethrough and the fix, because for this audience the correction is content.

The field this site describes moves fast. Every page carries its snapshot date — state as of July 2026 — rather than pretending timelessness. If you are reading this much later, treat it as a dated survey, not news.

§3Primary sources

The essays cite their sources inline and per essay. The load-bearing primary sources across the whole collection:

Secondary sources — surveys, roundups, trade-press summaries — are cited inline in the essays where used, with their limitations flagged in place.

§4Contact

Written and curated by Jörn Dinkla. The site is built with Jekyll and published on GitHub Pages; the reading experience needs no JavaScript, and the theme honors your system preference.