Back to log
🔍

Day 3: The 3am Rebrand and an SEO Experiment

ship30build-log

Day 3 was actually two days stacked on top of each other. The TikTok dashboard session bled past midnight into 3am. Then I woke up and helped the SEO team test AI on a real audit. By the time I stopped, I'd shipped work across three repos and something new was clicking.

The 3am rebrand

The TikTok analytics dashboard was functional but ugly. Started the evening intending to tweak the colors. Ended up rewriting the entire visualization layer, adding a warm fashion palette, building HTML report generators for campaign and SKU breakout data, and giving the whole thing an identity.

11 commits between midnight and 3:24am. It's called The Hive now.

Helping the SEO team try AI

After months of building Google Ads automation, the SEO team asked me to explore how AI could help with their workflow. Not to do SEO for them. They already know their job. But to see if tools, agents, or Claude Code could speed up the tedious parts.

So I picked a real project to test with. A packaging company's site, JavaScript-rendered, which meant scraping the raw HTML to see what the meta tags actually looked like. Found 7 issues on the homepage alone. No H1 (hidden inside a popup modal), heading tags used for styling instead of structure, empty alt text.

Keyword mapping, meta tag optimization, the full audit. Done in one session with an agent. The kind of work that normally takes days of switching between tabs, spreadsheets, and screaming into Search Console. It felt oddly calm.

The Keyword Planner moment

The interesting part was the keyword data.

The existing research doc listed one keyword at 2,400 monthly searches. I pulled the same keyword from Google's Keyword Planner API. The actual source, not a third-party estimate. Real volume: 90.

I wasn't expecting that kind of gap. It made me realize how much of SEO runs on estimates of estimates. Having API access to the real numbers just felt... different.

Everyone estimates search volume. I just... queried it.

What I actually shipped

TikTok client (11 commits):

  • Full dashboard rebrand to The Hive identity
  • Visual HTML report generators (campaign + SKU breakout)
  • Weekly progress report generator
  • Password-gated live web app for Railway
  • Waste reduction hypotheses with RM 30k/month savings analysis
  • Claude Project knowledge file for mobile workflow

SEO client (delivered, not committed):

  • 10-page SEO audit with optimized meta tags
  • Homepage content optimization (before/after with 7-section rewrite)
  • Keyword Planner API volume verification (17 keywords, Malaysia)
  • 5 blog article proposals from unused keywords

Company site (4 commits):

  • Updated Results section with latest proof points
  • Nav header and footer on blog pages
  • Marquee animation speed-up

cheeann.com (6 commits):

  • Published Day 2 blog post
  • Added daily toolkit badges to about page
  • Fixed mark tag rendering

What's becoming familiar

Three days in, I'm noticing the loop. Scrape the current state, audit it, optimize with real data, deliver. The tools change but the shape stays the same.

The agents handle the execution. What's left is the part I actually enjoy. Looking at something and knowing what's wrong with it.

Remove the friction of making and what's left is the friction of meaning.

Running total

  • 67 commits, 6 repos, 3 days
  • 1 SEO audit delivered (AI workflow test for the team)
  • 1 analytics dashboard rebranded and deployed
  • 2 websites live, 1 blog engine, 1 scoring engine
  • 0 blockers
Share