Class 08 · Movement 01
Build & Present
Your App
Session 08 · AI Engineering
Session 08 · Why this site matters

After the AI screen, your site is the first human impression.

~7 in 10

companies used AI in hiring by 2025 — 72–83% of it for resume screening (2024–25 industry surveys). Your resume meets an algorithm before it meets a person.

47%

The one to remember: 47% of hiring managers are less likely to call a candidate they can’t find online; 20% expect an online presence (CareerBuilder/Harris, 2018 — dated, still the rigorous version of the claim).

The FDE angle

Forward-deployed hiring is portfolio-friendly — and a dense portfolio of real engagements is the closest substitute for the 2–4 years postings ask for. The site you build this week is where your CTSBDC engagement write-ups will live.

Resume Builder Oct 2024 (n=948) · HireVue 2025 (n=4,000+) · CareerBuilder/Harris 2018 · compendium Part II
Session 08 · The quality bar

Beat the professional web — on day one.

95.9%

of the top one million home pages fail WCAG 2 accessibility — the first regression in six years, averaging 56.1 errors per pageWebAIM Million · 2026 report · as of Aug 2026

The same four failures, every single year
low-contrast textmissing alt textunlabeled inputsempty links

Fix four things and your first website beats most of the professional web.

Core Web Vitals — the speed bar
LCPlargest paint≤ 2.5s
INPinput response≤ 200ms
CLSlayout shift≤ 0.1

A static export on a CDN passes these nearly by default — the lowest-rung choice from class 06, paying off visibly.

The procedure: DevTools → Lighthouse, run on your deployed URL — the four failures and all three vitals are in that one report. The checks are the floor, not the bar: taste and writing are yours (next slide).
webaim.org/projects/million (2026) · web.dev/vitals — thresholds at p75 · captured Aug 2026
Session 08 · What goes on it

Work first. Case studies, not tech lists.

The one section that matters

Work / Projects — each project a short case study: what it is, why it exists, how it’s built. “React · Tailwind · Supabase” is an ingredients label, not a story — write the story.

No work yet? You already have three: this site (its own build story), your class-06 drawing, and anything you’ve made for a club, a job, a hobby. “Work” means problems solved, not employment.

The rest

About — who you are, in your own voice · Contact — one reliable channel. A single page holding all three sections is completely fine — depth beats sprawl.

The case-study skeleton — same anatomy as the town hall
  1. Problem — who it’s for, what hurt
  2. Approach — the stack, and why the lowest rung
  3. Live link — show, don’t claim
  4. What I’d do next — scope, visibly held
Inventory TruthA parts counter kept losing track of stock. I built a scanner page that writes to one Postgres table — counts stopped drifting the first week. live ↗next: barcode search

One structure, learned once: your talk, your case studies, and — in weeks 9–14 — your CTSBDC engagement write-ups.

2026 portfolio-guide consensus (sources in research/notes/movement-m2.md) — convention, not survey data
Session 08 · The working loop

The agent does the typing. You decide.

1your drawing
2CLAUDE.md
3failing check
4agent builds
5you verify
a failing check = the thing that must become true before you accept the work — a Lighthouse score, zero broken links, a checklist row. Write it before the agent starts.
You own

Taste — what it should feel like · copy — every word is yours · judgment — what stays inside the cut line · verification — DevTools open, checks green. The parts that make it your site.

The agent owns

The typing. Components, styles, wiring — generated against your drawing and your CLAUDE.md, checked by your verifier. No ceiling — and no judgment either.

Next class it ships. Definition of done, a launch checklist that doubles as your verifier, and what to do when the demo gods get hungry.
Connect.AI logoConnect.AI
Class 08 · Build & Present Your App01 / 05
1 / 5
Checking access…