Websites Change. Your Automation Shouldn't Stop.

April 4, 2026 · Leon Ting · 5 min read

Websites change. It's not a bug — it's the nature of the web. Designs get refreshed. APIs get versioned. Selectors shift. Components get renamed.

The question isn't "how do I stop websites from changing?" It's: "how do I keep my automations running when they do?"

The Stories We've All Heard

"I built Site Spy after missing a visa appointment slot because a government page changed and I didn't notice for two weeks."

— vkuprin, Hacker News

"I'd use it to track when state wildlife agencies update their regulation pages — those change once a year with no announcement and I always miss it."

— Hauk307, Hacker News

"I currently use Wachete, but since over a year, it triggers rate limits on a specific website and I just can't monitor German laws anymore."

— nicbou, Hacker News

Three different users, three different needs, the same problem: websites change, and there's no reliable way to know when or adapt automatically.

Two Types of Change

Not all website changes are the same:

1. Data changes — new content, updated prices, new posts. This is what you want to track. Your automation should detect and report these.

2. Structure changes — redesigned layout, renamed CSS classes, new API version. This is what breaks your automation. Your automation should survive these.

Most tools conflate the two. tap watch and tap doctor handle them separately.

Watch: Track Data Changes

$ tap watch hackernews hot --every 10m
2026-04-04T10:00  +added   "Show HN: Tap"  score=342
2026-04-04T10:10  +added   "Rust 2.0 announced"  score=128
2026-04-04T10:10  -removed "Old post fell off"  score=12
2026-04-04T10:20  ~changed "Show HN: Tap"  score: 342→487

tap watch runs your program on an interval, diffs the results, and outputs only what changed. It's built on Unix primitives — while + sleep + diff. No database. No scheduler service.

Pipe it anywhere:

# Append to a log file
$ tap watch github trending --every 1h >> ~/trending.log

# Send to Slack via webhook
$ tap watch reddit hot --every 30m | curl -X POST -d @- $SLACK_WEBHOOK

# Feed into another program
$ tap watch competitor prices --every 6h | tap filter --where "change > 10%"

Doctor: Detect Structure Breakage

When a website redesigns and your selectors break, watch won't show changes — it'll show nothing. That's where health contracts come in.

# Every tap has a health contract
health: {
  min_rows: 5,           // must return ≥5 results
  non_empty: ["title"]   // title must never be empty
}

$ tap doctor
hackernews/hot    ✔ ok     30 rows  (245ms)
bbc/news          ✘ fail   0 rows   min_rows: expected ≥5, got 0
github/trending   ✔ ok     25 rows  (1.2s)

bbc/news is broken. The site changed its layout. Before your data went bad, doctor caught it.

Auto-Heal: Fix and Continue

$ tap doctor --auto
bbc/news: 0 rows (min_rows: 5)
  ☉ Re-inspecting https://www.bbc.com/news...
  ☉ AI analyzing new page structure...
  ✔ Healed: bbc/news.tap.js updated
  ✔ Verified: 12 rows, score=1.0

AI re-inspects the page, writes a new program, verifies it works, and saves. One AI call. Then $0 per run again.

The Self-Healing Loop

# Set it up once
$ tap forge "track competitor pricing"      # AI writes program
$ tap watch competitor prices --every 6h  # monitor changes
$ tap doctor --schedule "0 6 * * *"       # daily health check

You configure it once. The loop runs itself:

• Data changes? → watch reports them in real-time
• Site breaks your program? → doctor detects via health contract
• Need a fix? → doctor --auto re-forges with one AI call
• Back to normal → $0 per run until the next change

You sleep. Your automations don't stop.

Read more: Your Scraper Is Broken Right Now · Your Automation Costs $1 Per Run

curl -fsSL https://taprun.dev/install.sh | sh
tap update && tap hackernews hot
tap watch hackernews hot --every 10m
tap doctor