Video summary
10 Websites That Can Find What the Internet Deleted
Main summary
Key takeaways
Main ideas & lessons (what the video conveys)
- The internet constantly “erases” itself through:
- Server outages
- Platform shutdowns
- Corporate edits
- Link rot
- Technical barriers (e.g., anti-bot systems)
- Internet archiving responds to “digital amnesia” by preserving websites and digital culture before they become inaccessible or untrustworthy.
- Different archiving services solve different problems, including:
- National/cultural legal preservation (government/library vaults)
- Searchability of the dead web (finding content without exact URLs)
- Machine-readable raw datasets (useful for AI training)
- Handling modern dynamic platforms (including JS-heavy and anti-bot sites)
- Legal integrity and citation stability (preventing link rot in courts/academia)
- Self-hosted personal preservation (reducing dependence on third parties)
- Preserving interactive behavior (not just page snapshots)
- Cross-archive aggregation (reducing fragmentation across archives)
- The Wayback Machine is presented as the most resilient baseline, but it still faces major cyberattacks and legal pressure.
Methodologies / approaches presented
UK Web Archive (Number 10) — national legal-deposit crawling
- Treats web content like historically important documents rather than disposable media.
- Built as a large-scale digital preservation project run by:
- A consortium of six legal deposit libraries
- Spearheaded by the British Library
- Works by:
- Crawling and collecting every UK website at least once per year
- Captures a wide variety of content (e.g., government announcements, local disputes, fan pages).
- Goal:
- Preserve the nation’s digital “memory” so digital rot isn’t accepted as inevitable.
Library of Congress Web Archives (Number 9) — curated “surgical” preservation
- Focuses on relevant cultural and governmental snapshots, not everything.
- Operates by:
- Maintaining over 180 event and thematic collections
- Actively crawling around 15,000 specific seed URLs at a time
- Examples of preserved targets:
- Congressional records
- National elections
- Supreme Court nominations
- Major cultural shifts
- Goal:
- Freeze key moments so future historians don’t face a “void” in early 21st-century evidence.
Archivo.pt (Number 8) — make the dead web searchable without exact links
- Addresses the limitation of most archives that require the exact dead URL.
- Operates as:
- A Portuguese web archive (operating since 1996) preserving billions of files
- A search engine for the past web using Google-like full text and image search
- Approach:
- Indexes text and images from long-dead HTML so users can search by:
- Phrases
- Historical events
- Product names
- Topics from specific time periods (e.g., 2004)
- Indexes text and images from long-dead HTML so users can search by:
- Goal:
- Turn inaccessible digital “graveyards” into an easily navigable library.
Common Crawl (Number 7) — open raw crawling dataset for machines/AI
- Operates as a nonprofit that:
- Crawls the web
- Scrapes data
- Freely releases archives/data sets to the public
- Timeline/scale:
- Doing this since 2008
- Accumulated ~10 pabytes of raw data (as stated)
- Emphasis:
- Optimized for raw text/data, not human-friendly browsing
- Claimed impact:
- Foundational dataset used to train many modern AI/large language models
- Goal:
- Serve as “raw material” powering algorithmic systems dominating the economy.
Ghost Archive (Number 6) — archiving dynamic, JS-heavy, anti-bot platforms
- Targets a key modern problem: traditional crawlers often capture blank pages or loaders.
- Designed to bypass limitations like:
- Heavy JavaScript
- Infinite scrolling
- Anti-bot protections
- Capabilities/constraints (as stated):
- Captures pages up to 50 MB
- Saves video from platforms like Reddit and YouTube up to 360p, under a 100 MB limit
- Targets:
- Social platforms where modern culture happens
- Content types vulnerable to deletion (videos, channels, purged subreddits)
- Goal:
- Provide “anti-rot” preservation so modern social media history doesn’t vanish after site changes.
Perma.cc (Number 5) — preventing link rot for legal/academic citations
- Problem/context:
- Legal and scholarly citations break over time (“link rot”).
- Claimed evidence of scale (as stated):
- A 2024 Pew Research study found 38% of web pages from 2013 are inaccessible.
- How it works:
- Developed by Harvard Law School Library Innovation Lab
- Creates a permanent, unalterable archived link
- Captures a page snapshot at a precise moment (“that exact millisecond”)
- Stores it in an institutional vault
- Goal:
- Ensure cited evidence remains stable and unchanged for courts, scholars, researchers, and journalists.
Archive Box (Number 4) — self-hosted personal archiving to avoid third-party risk
- Core idea:
- Don’t rely on third parties to preserve your specific content (to avoid loss from funding changes, lawsuits, server failures, or cyberattacks).
- Approach:
- Open-source, self-hosted web archiving system
- Decentralizes preservation by running it on your own computer or server
- Inputs:
- Feed it individual URLs
- Or connect it to browser history / bookmarks folders
- Outputs/formats mentioned:
- Saves local copies in standard readable formats such as HTML, PDF, WARC, and PNG
- Goal:
- Build your own “personal Wayback machine” on hardware you control.
Web Recorder (Number 3) — high-fidelity preservation of interactive sessions
- Problem:
- Many modern sites are interactive software; screenshots/crawlers can’t preserve functionality.
- Approach:
- Open-source tool focused on high-fidelity archiving
- Records the user’s interactive session
- What it records (as described):
- User actions like dropdowns, map expansions, complex script executions, and navigation through 3D spaces
- The exact behavior and network traffic in real time
- Goal:
- Preserve the experience so future users can “click through” the site as it actually worked.
Momento Time Travel (Number 2) — unify searching across many archives
- Problem:
- Archives are fragmented; finding a vanished page can become a time-consuming checklist across many providers.
- Approach:
- An aggregator using the Momento protocol (RFC7089) to add a standardized “time dimension” to web requests.
- How users interact with it:
- Provide a URL and a target date
- What it does:
- Searches multiple global web archives simultaneously
- Cross-references major archives in a fraction of a second
- Goal:
- Bridge gaps in the archiving ecosystem so the timeline can be stitched together.
Wayback Machine (Number 1) — continuous large-scale public web capture and resilience
- Role:
- Presented as the “heavyweight champion” and baseline of digital history.
- Approach:
- Continuous operation since 1996
- Methodically “photographs” the public web
- Scale milestone (as stated):
- Crossed 1 trillion archived web pages (October 2025)
- Over 100 pabytes of historical data
- Threats described:
- Massive cyber attacks and DDoS campaigns (2024–2026 timeframe mentioned)
- High-profile lawsuits from corporate publishers
- Goal:
- Preserve human digital culture despite attacks and legal pressure.
Overall takeaway
- No single tool fully solves web preservation.
- The video frames internet archiving as a spectrum of solutions spanning:
- Legal custody
- Curated historical capture
- Search over archived content
- Machine-friendly datasets
- Dynamic-content capture
- Citation-grade immutability
- Self-hosted control
- Interactivity recording
- Cross-archive time-based aggregation
- The Wayback Machine is positioned as central infrastructure, but all these tools together form a broader defense against the disappearance of digital history.
Speakers / sources featured
- Sources named (organizations/tools):
- UK Web Archive (consortium of six legal deposit libraries; British Library spearheading)
- Library of Congress
- archivo.pt
- Common Crawl
- Ghost Archive
- perma.cc (Harvard Law School Library Innovation Lab)
- Archive Box (open-source, self-hosted)
- Web Recorder (open-source)
- Momento Time Travel (Momento protocol RFC7089)
- Wayback Machine (Internet Archive)
- Pew Research (2024 study mentioned)
- No individual human speakers are identified in the subtitles.