Bhavana's Week 10: Ship it or Scope it
Bhavana Bhariday runs a full audit against acceptance criteria, hits 100/100 on Lighthouse, and scopes out a company page rather than shipping it thin.
Marking a feature "done" and actually knowing it works are two different claims, and the gap between them is exactly what a checking week is supposed to close. Ten weeks into building SEO for a job search app, the work wasn't to add anything new, it was to verify what had already been built and honestly defer what wasn't ready yet.
Checking assumptions against reality
The discipline here is narrow on purpose: two things, and only two. First, a full audit confirming every acceptance criterion actually holds on the running site rather than in memory. Second, an independent tool run against that same site to check the self-assessment for blind spots. A self-assessment is easy to fool, which is exactly why the second check exists.
Running the full audit
The spec carries a real acceptance checklist, and going through it one item at a time is boring but essential work. JSON-LD structured data on every job page, noindex tags on inactive listings, a working sitemap index, correct robots rules, canonical URLs, and the indexing API, each one gets checked against the live site rather than assumed from memory.
Hitting 100/100 on Lighthouse
The independent check ran Google's own Lighthouse tool against the three page types that matter most: a job page, the landing templates, and the master directory. The result held steady across all three: 100 out of 100 on SEO, a score anyone else can reproduce by running the same tool, not a self-graded opinion of the work.
Scoping the company page honestly
Company detail pages were on the original list, meant to get the same meta and canonical treatment as job pages. But a meta tag needs a real page behind it, and the sponsorship data that would make that page substantive isn't in place yet. Rather than ship a thin page just to check a box, the page noindexes itself until it actually has content worth showing to a crawler.
Key takeaways
- A checking week means auditing what's already built against the running site, not adding new features.
- The acceptance checklist includes JSON-LD, noindex on inactive listings, sitemap index, robots rules, canonical URLs, and the indexing API.
- Lighthouse scored 100 out of 100 on SEO across job pages, landing templates, and the master directory.
- A page without real content behind it gets noindexed rather than shipped as a hollow pass.
- The rule: ship it or scope it, but don't fake it.
Try it yourself
Take a feature you've already marked done and ask Claude for the difference between its assumed status and its real status, then check what an independent tool would actually report. Find the item you're tempted to check off cheaply and decide honestly whether it's shippable or whether the right move is to scope it out until the data exists. This is a Humanitarians AI Fellow's weekly report from the Fellows program, where the work is documented in public as it happens.
Chapters
Full transcript(auto-generated, with timestamps)
Checking Assumptions vs. Reality
[0:00]Hi, I'm Bhavana, a Humanitarian's AI Fellow. Week 10 of my Job Seeker report, the job search app where I run SEO, making Google find and rank our listings. This week, nothing new got built. It's a checking week. Verify what actually works and honestly defer what isn't ready. Two things this week and only two. A full audit, confirm every acceptance criterion, then run
The Full Audit: Checking the Specs
[0:25]Lighthouse. That branch ended at 100 out of 100 across every page tested. And a company page attempt, which ended somewhere more honest, scoped out for now to revisit once the real data exists. One win, one deliberate not yet. First, the audit. The spec has a real acceptance checklist and the discipline is boring but essential. Go through it
Lighthouse: Achieving a 100/100 SEO Score
[0:47]One item at a time and confirm actual status, not the status you assume. JSON-LD on every job page, noindex on inactive ones, the sitemap index, the robots rules, canonical URLs, the indexing API. Each one checked against the running site, not against memory. Then the independent check. A self-assessment is easy to fool. So, the audit ran Lighthouse, Google's own tool,
Scoping Honestly: The Company Page Decision
[1:12]Against the three kinds of page that matter. A job page, the landing templates, and the master directory. The result was the same every time. 100 out of 100 on SEO. Not my opinion of the work, a score anyone can reproduce. The other half was scoping honestly. Company detail pages were on the list, the same meta and canonical treatment job pages get. But a meta tag needs a real page with real content behind it. And the sponsorship data that would make one
Recap: Ship it or Scope it, Don't Fake it
[1:41]Substantive isn't in place yet. So, rather than ship a thin page to check a box, the page noindexes itself until it's actually worth showing. Documented, deferred, not faked. The recap, checking not adding. The acceptance criteria confirmed one at a time against the running site. Lighthouse, 100 out of 100 on every page tested, a result anyone can reproduce. And the company page scoped out honestly instead of shipped thin because a meta tag with no real content behind it is just a box checked. The rule this week, ship it or scope it, but don't fake it. Your turn. Take a feature you've marked done and ask Claude for the difference between assumed status and real status. What would an independent tool actually report? Then find the item you're tempted to check off cheaply and ask whether it's genuinely shippable or whether the honest move is to scope it out until the data exists. Run this and audit your own done. Ship it or scope it.
