In keeping with a session about using AI responsibly, here is an honest account of how the site itself was made. It was built over about two days of focused work by Miguel Guhlin working with Claude, an AI assistant. A person directed every decision: what to build, which quotes to use, and what is appropriate for a criminal defense audience. The figures below are rough estimates, offered in the same spirit of candor the session asks of any AI-assisted work.
Three figures
Estimated, not measured. The ranges matter more than the numbers.
| Whose time | Estimate | What it covers |
|---|---|---|
| AI assistant, active working time | ~4–7 hours | Aggregate generation and tool work: writing the code and content, web research, image processing, rendering pages to check them, and deploying, spread thin across two days. |
| Facilitator’s time | ~5–9 hours | Writing the requests, reviewing deployed pages and screenshots, making the judgment calls (which quote, which layout, what fits a TCDLA audience), and catching issues. |
| Equivalent human build, from scratch | ~250–400 hours | Roughly six to ten weeks for one skilled generalist, or about three to four weeks for a small team working in parallel. |
Why the human estimate is that large
The same work, broken down by workstream for one skilled person.
| Workstream | Estimated effort |
|---|---|
| Design system, brand, and two CSS frameworks (pages and decks) | 3–5 days |
| Deck framework: presenter view with cross-window sync, keyboard navigation, print/PDF export, exit icon | 4–6 days |
| Four slide decks: layout, copy, speaker notes, images | 5–7 days |
| Eight activity pages, including the interactive fluency self-assessment with scoring, a color chart, and PDF export, the ethics randomizer, and a fifty-scenario bank | 7–10 days |
| Hub, About, Presentations, Walkthrough, Quotes, and three handouts | 4–5 days |
| Translating all interface text into seven languages (Spanish, Vietnamese, Arabic, Hindi, Urdu, Chinese), including right-to-left support | 5–10 days |
| Content writing: defense-appropriate copy, ethics scenarios, and the prompt library | 4–6 days |
| Image production: generation, cropping infographics, and optimizing thirty-plus assets | 2–3 days |
| Quote research, verbatim verification, and sourcing (sixteen quotes) | 1–2 days |
| Testing, revisions, and deployment | 2–3 days |
That lands around forty to fifty-five working days for one person doing all of it; longer once you add coordination, or compressed with a team.
The takeaway
Roughly 250 to 400 hours of conventional human effort, delivered against about five to nine hours of a person’s time over two days, a rough thirty to fifty times leverage on human attention.
The honest caveats matter. The person still supplied all of the judgment: which quotes, which framing, what is appropriate for a defense audience. The translations are machine quality, where a human project would pay for professional review. And much of the time saved is grunt work (translating seven times over, optimizing images, renumbering slides, checking links) that compresses especially well, while the parts that need a real lawyer’s eye still rest with a human.