← Back to Lately
Lately · 4 October 2026

One evening, three projects, and an afterthought.

This wasn't planned. I spent an evening half-watching TV and pottering around the house, asking Claude for help on a few side projects whenever something occurred to me: a fix here, a new feature there, a change of mind or two.

Only near the end did the question pop into my head: how much did we actually get done tonight, and how long would that have taken one experienced person working on their own? So I asked Claude to go back through the evening's sessions and work it out.

~3¼ hrs
Elapsed time, with my attention very much split
~75–80 hrs
Estimated for one experienced professional (range 62–94)
3Projects
28Requests
6App releases
6Site updates shipped
24Distinct skills

What got done

Three projects, worked on side by side. Times are approximate (UK time).

A desktop app for tidying duplicate photos

≈ 22:30 – 23:50

Petunia and I — the music site

≈ 00:25 – 01:10

WEYER.TV — this site

≈ 00:45 – 01:45

The estimate

How long one skilled, senior person might take to do the same work properly: design, build, test on desktop and mobile, release and support.

WorkstreamLowHigh
Duplicate-photo app: six releases, Windows fixes, packaging, support20 h30 h
Petunia and I: news post, release-page repairs, 20-post fix, advice, docs8 h12 h
WEYER.TV fun mode: concept, theme, copy and jokes, illustrations, code, testing20 h30 h
WEYER.TV follow-ups: switch redesign, favicon, projects, notes feed, social links11 h18 h
Research, reviews and releases3 h4 h
Total (best estimate ≈ 75–80 h, about two working weeks)≈ 62 h≈ 94 h

The skills it called on

24 distinct skills, which would normally mean around 12 different roles.

Desktop developerPython app development, duplicate-matching logic, packaging and updates
Windows supportPowerShell, system folders, shortcuts
Front-end developerHTML, CSS and JavaScript, data-driven content, responsive layout
UI / visual designerTheme and visual system, interface and layout design
Illustrator / brand designerMemes and illustrations, icon design, image production
CopywriterProfessional web copy, comedy writing
Content editorNews writing with source credit, basic legal and IP literacy
SEO specialistShare previews, metadata, AI-crawler indexes
AccessibilityScreen-reader labels, reduced motion, contrast
QA / test engineerAutomated browser testing, regression root-cause analysis
DevOps / releaseVersion control and pull requests, deploy and cache troubleshooting
Product / documentationTrade-offs and recommendations, keeping project notes up to date

What I took from it

  1. My job wasn't typing; it was directing. Deciding what mattered, saying it clearly, looking at the result and correcting course. Most of my messages were a sentence or two, often half-formed.
  2. That's product ownership. Knowing what good looks like and communicating it well has always been the core of agile delivery. With AI doing the making, it becomes most of the job.
  3. Judgement still matters. Several of the evening's best moments were course corrections: “that switch isn't part of the navigation”, “the font looks comical”, “it wasn't a planned experiment”. The work got better because someone was paying attention, even if only half of it.
  4. Multitasking isn't the lesson. I'd normally coach people out of it. The interesting part is how far a little clear direction goes when the doing is cheap.

How the estimate was made. Claude reviewed the evening's session transcripts, listed every request and change, and estimated the hours one experienced professional would need for each, including design, testing and release but not meetings or handovers. It's an informed estimate, not a measurement, so treat the numbers as a range, not a fact.

It covers only the coding sessions Claude could see that evening. Anything I did in other chats isn't counted.

Talk to me about AI-assisted delivery → More from Lately