A Better Way to Use LLMs for Writing
After experimenting with several articles on writing with LLMs, I came up with what I feel is a better way to use LLMs without losing your voice.
Senior web developer with 15 years building CMS architectures, leading migrations, and shipping front-ends that hold up long after the original build. WordPress, Drupal, Sitecore, Jekyll, and the JavaScript in between.
After experimenting with several articles on writing with LLMs, I came up with what I feel is a better way to use LLMs without losing your voice.
Six checks that passed on broken code in a Next.js landing page build, including axe reporting zero violations with three accessibility defects.
How I designed a Sanity content model from a Figma spec, using seven typed blocks with no page builder and heading levels set by section order.
How I converted a Next.js 16 landing page from static export to Sanity, covering Presentation, CORS origins, draft mode, and four silent failures.
How to test a website against WCAG with axe-core, Lighthouse, WAVE, and Pa11y, what those tools miss, and where AI fits in manual review.
Which WCAG version each accessibility law names and who must meet it, covering the European Accessibility Act, ADA Title II and III, and Section 508.
Answers to a salesperson's questions on AI lead scoring. What the score looks like in your CRM, whether you see low leads, and who sets the cutoff.
Identity checks prove a lead is a real person at a real company. What bot detection, honeypots, and SMS verification cost an AI-equipped attacker.
An idea for scoring sales leads with AI checks on the email domain, the site behind it, and the person who submitted it, averaged into one score.
Prism's line-numbers plugin measured the gutter against the wrong line-height when the stylesheet loaded late. One rule in the critical CSS fixed it.