Using Claude Code: The Unreasonable Effectiveness of HTML
- Source
- x.com
- Author
- Thariq
- Date

Thariq's argument to make HTML, not Markdown, your 's default output — because tables, SVG, CSS, and interactivity beat plain text once an agent is generating real artifacts. A small default change with outsized effect on what your agent can produce.
- The practical argument for HTML is readership. The author stops reading his own markdown past about 100 lines and cannot get colleagues to read it at all, so a spec's chance of actually being read is the thing HTML improves.
- Sharing is the mechanical difference. Browsers do not render .md natively, so markdown travels as an attachment; an HTML file uploaded anywhere becomes a link a colleague opens on any device.
- When you deny the model a visual format, it improvises badly: ASCII diagrams, and in one case approximating colors with unicode characters. The output format is a constraint on what the model can tell you.
- The author explicitly warns against packaging this as an /html . Just ask for 'an HTML file' — the skill worth building is knowing what you want the artifact to do, and that only comes from prompting it from scratch a few times.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- agent skill — A reusable instruction file that teaches an agent how to do one job well — the procedure, the tools, and what counts as done.
“I would go so far as to say that there is almost no set of information that Claude can read that you cannot fairly efficiently represent with HTML.”
“The chance of someone actually reading your spec, report or PR writeup is much much higher if it’s in HTML.”
“Making HTML documents with Claude is just more fun and makes me feel more involved and invested in the creation, and that by itself is enough.”
“I’m a little bit afraid that people will read this article and turn it into a /html skill or something.”
articleThariq's deep dive: how effort levels shape Claude Code's behavior
postClaude Code Ships Plugin Evals to Catch Skill Regressions
postWatermarking without quality loss is a bit unintuitive, doesn't feel like it sho
postAll Claude generated text will have this embedded watermarking. For example, you
Checking sign-in…
Loading comments…


