King's College Sarsen Lecture

 

https://www.kcl.ac.uk/events/silcretes-stone-tools-the-origins-of-the-sarsen-megaliths-at-stonehenge

It's Not Ritual, Mike — It's a Content Farm

I wondered about the phenomenon of AI-produced videos that Mike Pitts has highlighted, the unasked question of why. Incentives matter. So it seemed appropriate to ask an AI the question.

Follow the money (such as it is)

These are content farms. Pitts himself identifies one Instagram account as based in the Philippines — not incidentally, one of the major hubs for AI content farming alongside India and parts of West Africa. A Kapwing study from late 2025 identified 278 YouTube channels producing nothing but AI slop: collectively 63 billion views, an estimated $117 million in annual ad revenue. The top channels pull $4 million a year.

The channels Pitts found aren't in that league — low views, low subscribers. But the marginal cost of each video is close to zero. Stitch together AI clips, overlay a synthetic voiceover, bolt on SEO tags ("Stonehenge mystery," "how was Stonehenge built"), and you have a lottery ticket that costs nothing to print. YouTube CPM for history content aimed at US/UK viewers runs $5–12 per thousand impressions. A few hundred views multiplied across dozens of daily uploads across disposable channels adds up — not to a fortune in London, but to real money in Manila or Lagos. As the Manila Bulletin noted, the strategy is simple: flood the platform, earn from volume.

That's why the channel names — BRIGHT SIDE, The Strange Vault, Chewy Playz, PyramidCoin — are disposable nonsense. They're fishing nets, not brands. The four Easter Island videos posted on a single day aren't an algorithm responding to Pitts' curiosity. They're a production line scaling up on a trending topic cluster.

Is the apocalypse nigh?

Pitts' concluding vision — AI cannibalising itself until original research vanishes — deserves pushback. The Stonehenge video has about 500 views. YouTube's own algorithm now penalises low-effort AI content with up to a five-fold reach reduction, and CEO Neal Mohan used the phrase "AI slop" in his January 2026 letter, pledging enforcement. The platform has already terminated or wiped at least 17 channels.

The real loss is narrower: the middle ground once occupied by decent popular television, where a casually curious viewer might have stumbled on something approximately true. That's worth worrying about. But the web has been through content-farm infestations before — Google's Panda update crushed the keyword-stuffing epidemic of 2009–2011. The cycle of exploitation and platform response doesn't end in civilisational collapse. It ends in better filtering, with damage done in the interim.

What can be done?

For individuals: source literacy. Check the channel, the subscriber count, the production style. If the voiceover sounds like a machine reading a script written by a machine summarising Wikipedia, trust your instincts.

For platforms: enforce the existing rules. Mandatory AI-content labelling under YouTube's 2026 transparency policy is a start, but only if permanent demonetisation for non-disclosure is applied at scale. Provenance standards like C2PA need wider adoption. Raising monetisation thresholds for new channels would make the throwaway-channel model less viable.

For creators: keep making the real thing. Structured metadata, transcripts, institutional collaboration, and presence on platforms that reward quality over volume all help. So does writing sharp blog posts pointing out that Neolithic Wiltshire did not, in fact, have windfarms.

Pitts is right that there's a problem. But it's not a metaphysical crisis. It's low-grade commercial fraud enabled by cheap tools and a platform that hasn't caught up with enforcement. 

Stonehenge, the Neolithic Olympic Park

 

The press is full of reports this morning that the original Stonehenge build was basically the Neolithic London 2012 Olympics: wildly over budget, chaotically organised, drug fueled frolics to glorify dodgy leaders and authoritarian regimes? At least I think that is the implication of Win Scutt's theory, maybe.

The press reports are to celebrate the completion of the modern Kusuma Neolithic Hall. My friend Luke Winter ran a tight, respectful ship — no nonsense, just proper experimental archaeology and a brilliant volunteer team who built the whole 7-metre beauty using only authentic methods. It came in on time, on vision and full of heart. London 2012 could’ve learned a thing or two.

A sample of the Press:

• English Heritage’s official page on the Kusuma Neolithic Hall
• The Independent’s first look inside
• The Times piece that ran with Win’s theory

Devizes Sheela na gig

Just down the road from the Museum in Devizes is St John's Church, I always like to check on the corbels when I pass.

"These figures were discovered by Dr Theresa Oakley and Dr Alex Woodcock who published their findings in the paper “The Romanesque Corbel Table at St John’s, Devizes and its Sheela na gig” (The Wiltshire Archaeological and Natural History Magazine Volume 99 2006). The Sheela na gig of the title is one of a pair of exhibitionist figures on a corbel. This makes this corbel the third exhibitionist pair discovered so far in the UK, the other two being at Kirknewton and the window lintel at Whittlesford." - More, not for the fainthearted, at https://sheelanagig.org/devizes-st-john/

Sheela na gig corbel, Devizes by Kenneth Lymer on Sketchfab

An AI cleanup by Grok of the pair.

In Search of the Giant Stones | Stonehenge Access All Areas Ep 2

New Sidebar Tools: Fact Checker, Second Reviewer & Site Search

I’ve added three new widgets to the sidebar to support reading and research on sarsen.org.

These tools are designed to help evaluate claims, review documents, and find relevant content on the site. They all use Grok, with prompts written to maintain high standards of evidence. Other AI tools are available and I would urge you to use them as well, Grok just produced the cleanest widgets and is very good at fact checking. Having detailed prompts prewritten improves the ease and accuracy of using AI agents.

1. Archaeological Fact Checker

This widget helps assess the strength of individual claims.

You can paste a URL or a block of text, and it opens a structured analysis that evaluates claims according to a clear hierarchy:

• Peer-reviewed papers as the highest standard
• Academic consensus
• Single-author assertions, which are examined using logical and Bayesian-style reasoning

It’s particularly useful when encountering claims in articles, social media, or discussions about Stonehenge, bluestones, sarsens, or Neolithic archaeology.

2. Second Reviewer

This tool is designed for reviewing papers and drafts.

It offers two options:

• Review Pasted Text — Paste a section or excerpt for a structured review covering the referenced science, logic, inferences, strengths, and weaknesses.
• Open Grok for PDF / Document Upload — Opens Grok with a detailed reviewer prompt already loaded. You can then upload a full PDF or document directly for a more thorough review. This version also provides suggestions for improvement when reviewing drafts.

It applies the same evidence standards as the Fact Checker, with a focus on peer-reviewed sources and logical rigour.

3. Search sarsen.org with Grok

This is a site-specific search tool.

Instead of a general search, queries are directed toward content on sarsen.org. It’s useful when you want to find posts or information on particular topics (such as specific monuments, papers, or debates) without leaving the site’s context.

How to use them

All three widgets are available in the sidebar. They are intended as aids rather than definitive answers. The quality of the output depends on the prompt and the material provided. As with any AI-assisted tool, it remains important to check original sources, especially peer-reviewed papers.

These tools are experimental and will likely be refined over time. Feedback on how they perform is welcome.

Thinking Three Dimensional

 Most glacial reconstructions are drawn in two dimensions. Ice limits appear on maps as lines advancing across the landscape, with arrows showing flow directions. This creates a comforting sense of certainty: the ice reached here, stopped there, and therefore must have overridden everything in between.

But glaciers do not advance across flat maps. They move through three-dimensional space, and height matters enormously.

North Devon sits at the southern limit of contested Irish Sea ice extent, which makes the elevation question here more than academic. The Fremington Clay series and the scattered erratics along the Croyde–Saunton shoreline contain far-travelled material; some of it almost certainly arrived via ice-rafting or ice-marginal processes when Irish Sea ice occupied the Bristol Channel. That much is uncontroversial, and it is not what is at issue.

What is at issue is whether that ice actually came onto the land at any significant elevation above the contemporary shoreline — whether the ice limit drawn on a map translates into ice that physically overrode the ground behind the coast.

The evidence does not support that conclusion. Credible far-travelled material in north Devon is concentrated at or near modern sea level. The Fremington Clay sequence is largely confined below 30 m OD. The scattered erratics along the Croyde–Saunton coast are coastal features. Above that level, the record goes quiet. There are no well-documented glacial deposits on the north Devon mainland at meaningful elevations, no striated pavements, no till sheets, no unambiguous trains of transported material that would indicate ice moving across the land surface.

The Ramson Cliff boulder at around 80 m OD on Baggy Point has been repeatedly cited as the exception — proof that Irish Sea ice reached significant elevation on the north Devon coast. It deserves to be examined on its own terms rather than accepted by inference from two-dimensional ice maps. A peer-reviewed re-examination published this year concludes that it does not function as reliable evidence of glacial emplacement at that elevation: there are no supporting glacial deposits at that height, no coherent glaciological mechanism that would place it there, and no record of the boulder's existence before 1969 despite earlier surveys of the area (Daw, Ixer & Madgett 2026, Quaternary Newsletter 167: 13–19, doi:10.64926/qn.20517). An exotic stone in an anomalous position is not self-evidently a glacial erratic. Further discussion and statistical analysis are available at sarsen.org (February and October 2025).

The logical point here is simple and applies well beyond north Devon. A glacier can reach a coastline without surmounting the ground behind it. Ice-rafting can deliver boulders to a shoreline without the ice sheet having climbed the hills. The two-dimensional line on a map — the ice limit — tells us where the ice margin sat, not how high it reached into the interior. Treating a mapped ice limit as a guarantee of inland overriding conflates the plan view with the vertical reality.

Until robust, well-contextualised evidence appears — glacial deposits, striated surfaces, or unequivocal erratics at meaningful elevations on the north Devon mainland with documented discovery contexts — the most defensible reading on current evidence is also the most conservative one: Irish Sea ice influenced the coastal zone of north Devon at low levels. The case for significant inland incursion at elevation remains unproven.