SearchResearch (10/29/25): The 1 trick you need to know to use AI for deeper, better reading

 I absolutely adore... 

P.G. Wodehouse.  P/C Wikimedia

... the writings of P. G. Wodehouse.  Whenever I need a lift in the old spirits, I pluck a volume from the bookshelf of Wooster and Jeeves, I read a bit, and in the blink of an eye, all is right with the world.  As Wodehouse might say, God is in His heaven and the celestial choirs sing again.  

If you don't know Wodehouse, drop what you're doing and read a short story or two. Better yet, pick up a Wodehouse novel and dive in.  

I'd recommend Right Ho, Jeeves, which is an excellent place to start.  

The writing is droll and the language--especially the language--just tickle my humorous bones.  

BUT, Wodehouse is satirizing the language and behaviors of the early 1900s upper class.  They are a rich vein to mine, but roughly once each page, there is a phrase or word that escapes my understanding or offers up a nuance that completely misses my brain.  

For instance: 

Butter-and-egg man (An investor with a lot of money)

Absquatulate (To depart suddenly or abscond)

Cattywampus (Used to mean something that was directly across from something else, as opposed to its modern meaning of being askew or in disarray)

Those are fairly easy to look up.  But the more tricky phrases are things like: 

"Only that she’s a blister.”

Or... 

"Deprived of Anatole’s services, all he was likely to give the wife of his b. was a dirty look."  

I know what a blister is, but the obvious definition makes no sense here.  And what is "...the wife of his b."?  That's clearly not the end of a sentence, but feels like an abbreviation for something--but what? 

Here's where your friendly, local LLM comes in handy.  Here's what I did to figure out each expression:  I asked an LLM (Gemini in this case) to explain it to me in the context of the book... 

And when you need to be even more specific, give the name of the story in the context you provide to the LLM.  

In both of these cases, it's not clear that any amount of contextual reading would have taught me these meanings.  

This is a brilliant use of an AI to augment your ability to deeply read a text.  

On the other hand, use caution:  AI still makes mistakes, and they can be subtle. 

Here I asked a question about the mention of a device in a book written about the same time as Wodehouse: 

This completely checks out.  (Of course I double check everything.  Don't you?) The Veeder box is indeed a type of odometer made at the time.  

However... see this next part of the explanation: 

That mention of "By the time Evelyn Gibb and her husband were bicycling the West Coast in 1909..." is completely made up.  The book is NOT about Evelyn Gibb and her husband, but is about Vic McDaniel and Ray Francisco, friends who cycled 1,000 miles from Santa Rosa, California, to Seattle, Washington, for the Alaska-Yukon-Pacific Exposition. The author (Evelyn Gibb) is Vic's daughter, not his wife.  

SearchResearch Lessons 

1. Using an AI to give insights into obscure texts can be incredibly handy.  By virtue of having ingested so much text, an AI can often give you a perspective about a fragment of text that you don't understand. 

2. CAUTION:  Check everything--there are still hallucinations about!  Double check everything!

Hope you find this useful SRS method!  

Keep searching.  

SearchResearch (10/24/25): The shifting of SearchResearch

 I've noticed a subtle shift.  

The longer I write this blog, SearchResearch, the more changes I see.  Content on the web changes, the tools we use change--the whole ecology of writer, reader, producer, and consumer has dramatically shifted since I began writing back in January, 2010.  That was 15 years ago.  In Internet years, that's about 1500 years.  (I figure Internet years are about 100-to-1 with Human years.)  I've written 1478 posts and we've had 6.23 million reads. You have written around 10 comments / post, for which I thank you.  

In my first blog post I wrote: 

I have to warn you before you start reading: In the back of my head, I want something tangible to emerge from this. Ideally, a book, or a series of books, about how people search... how they research... and how they get good at doing this.

Congrats. We've done that.  The Joy of Search came out in 2019 to reasonable success.  I'm happy about how well it worked as a book.  

And I see that this blog is shifting a bit too.  

As you've seen, our typical pattern is that one week I'll pose a Challenge--usually a question about some interesting aspect of the world that requires using a particular research skill that you, dear reader, need to figure out.  Good news here, you figured out some deep skills.  

Some of my favorites have been skills like knowing Control-F (the skill of finding text),  using site: restriction (to search just within a particular website), or using deep resources like Google Books or the Newspaper archive.  

But with the rise of AI tools to help out with doing deep online research, it seems that our skills need to shift as well.  You still need Control-F, but I find myself using tools like site: less-and-less these days.  

So... I think we need to shift the way the SRS blog will operate.  As I wrote in 2010:  

When you think about it, search is not something you're born with--there's no inherent, latent skills for research (the way there is, say, for walking or spitting). Some people are really good at it, others just never quite get the basics. 

That's still really true--but more people know Control-F these days, and AI is doing a lot of the search-specific skill. 

HOWEVER... I still find myself using somewhat more subtle online research skills. The technical problem for this blog is that it's hard to frame the skills in terms of motivating Challenges.  So the blog is shifting a bit as well to try and communicate those sensemaking and deep research skills.  

I WILL pose interesting Challenges from time-to-time when I just can't resist their siren call, just not every other week as we've been doing.  

Instead, I want to point out some of the deep research skills we need to cultivate.  And that will require me telling stories, rather than posing a research Challenge.  

Bear with me as we try to figure out the new format.  I'm confident that we'll find something that's deeply interesting and fun.  Stay tuned as SRS starts a few experiments.  

In my next post I'm going to point you to some people who are writing about this new, AI-linked research methods.  That will be entitled, Key skills you need to have to be an effective online researcher and will be a collection of some posts by other folks who have good things to teach us as well.  

Stay tuned.  Keep reading, keep leaving comments.... 

And keep searching.  

 

Answer: How can the same locust look so different?

 

It's difficult to understand... 

Rocky Mountain locust. P/C Wikimedia

... how variable the appearance of an animal might be.  In this case, how can this particular insect--the Rocky Mountain locust--be so variable in appearence that biologists thought that the two different forms of the insect were actually different species?  

So how could biologists mistake the two different looks of a locust for two different species?  

1. How often has it happened that biologists have seen two (or more) species when it was really just one in different clothing?  Can you find another case of two (or more) species being reconciled into one? 

This seems like a great question for an LLM.  When I copy/pasted this into Gemini, I learned that this happens more often than I thought.  The Gemini answer talked about "lumping" and "splitting" a species definition--by "lumping together" two organisms that were thought to be different, and "splitting apart" organisms that look the same, but are actually genetically different.  

That sounds right, but the words "lumping" and "splitting" are probably NOT what biologists call this process.  

A quick query of: [what do you call it when biologists find that two different appearing animals are found to be the same species and they reclassify them in a new species name] taught me that biologists refer to this process (which happens a LOT), as synonymization. ("Synonymization" is the process of identifying and combining different scientific names that refer to the same organism. This happens when a species is described multiple times by different scientists, or when a species is reclassified into a different genus, or when different organisms are discovered to be variants of a common organism.)

I revised the question to learn about "splitting" and learned that this is just regular old speciation, which then leads biologists to a taxonomic revision in the textbooks.  

This is often the result of cryptic species, which are species that appear identical but are reproductively isolated.  

To find examples of "splitting" I asked the obvious query: [can you give me examples of organisms that seem very similar and were once thought to be one species, but are now understood to be multiple species?]  and found several examples.  The elephant in the room is obviously the African elephants...  

African elephant (Loxodonta africana).  P/C K. Russell

Historically classified as a single species, the African elephant, has now been distinguished as two separate species: the African bush elephant (Loxodonta africana) and the African forest elephant (Loxodonta cyclotis). They have nearly identical appearances, but DNA analysis revealed them to be genetically distinct and reproductively isolated, with the forest elephant being slightly smaller and having straighter tusks.

And the converse:  [can you give me examples of organisms that seem very different and were once thought to be different species, but are now understood to be just one species?] 

Leptocephalus, the larval form of Anguilla anguilla. Yes, they are transparent.

A lovely answer:  For centuries, the transparent, ribbon-like leptocephalus larva was believed to be a separate species from the adult eel (Anguilla anguilla). It was only in the early 20th century that scientists realized the leptocephalus is the eel’s larval stage and not a different organism.  

And when I asked about locusts in particular, I learned that for centuries, naturalists thought that the grasshopper and the swarming locusts were entirely different insects. (For the record, they also wrote that caterpillars and butterflies were completely different insects as well...)  

The solitary form of the locust lives alone as a grasshopper, while the gregarious or swarming form appears during outbreaks.  It's larger, brightly colored (often yellow and black), with longer wings, stronger flight muscles, and completely different behavior, preferring to fly in massive swarms.  

They were so different in appearance and habits that early entomologists gave them different scientific names. 

Then, around one hundred years ago (1921), Sir Boris Uvarov recognized that two locust species are one species but appearing in two different phases, a solitarious and a gregarious phase.  This phenomenon of phase polymorphism, is now called polyphenism.   (See a nice review paper on Uvarov's discovery, "One hundred years of phase polymorphism research in locusts.") 

It turns out that under crowded conditions, young locusts experience tactile stimulation on their hind legs and undergo phase transformation, triggering massive physiological and behavioral changes. This transformation affects color, size, brain structure, metabolism, and social behavior, switching them from a solitary to a gregarious form — leading to the famous locust swarming behavior.  

2. It's clear that organisms can have multiple shapes / patterns / colors (we've discussed this before in the context of plant mimicry).  Can you find an organism that has a huge number of different appearances?  Any idea WHY they have such variability?  

 

I put this question to Claude as [Can you find an organism that has a large number of variable appearances?  That is, what is the most polymorphic organism?]

All of the AIs--Claude (and Gemini and ChatGPT)--gave variations on a good answer, pointing out that both the Great Mormon Butterfly (Papilio memnon) and certain snails (e.g., the Grove snail, Cepaea nemoralis) are famous for their polymorphism, leading biologists to classify the different forms as different species.  

Polymorphisms in Papilio memnon.  P/C Wikimedia

The Grove snails have widely varying shells, which can be different colors (brown, pink, yellow) and have various banding patterns. These different morphs can look so distinct they puzzled early researchers, and the variety is controlled by a complex of closely linked genes.

Polymorphism in Grove snails (Cepaea nemoralis). P/C Wikimedia

And there's the answer: there are many organisms with widely varying morphs--ants, bees, fish, snails, and locusts.  

WHY this is so can be seen in the many shapes of dogs around the world.  How can one species be SO variable in size, shape, and color, yet all be one species?  

Another LLM query:  [what are the genetic factors the cause extreme polymorphism in some species?]   You can do that query yourself and read the details, but it boils down to this: there are a small number of genes that control a LOT of the variation in coat kind, coat color, size, muzzle shape, etc.  With a great deal of selective breeding over the eons (by people), the variation has been amplified into the great number of dog varieties that we see today.  (For lots of insights, see:  Boyko, Adam R., et al. "A simple genetic architecture underlies morphological variation in dogs." PLoS biology 8.8 (2010).  

Search Research Summary

1. The AIs worked well.  One of the nice surprises of this Challenge is how well the LLMs answered each question.  This is largely due to this being a not-especially controversial area--nobody bothers to push out pathological content about the genetics of insects or dogs.  

2. Double check everything.  HOWEVER... for each result I write about here, I double-checked each claim.  In some cases I triple checked. It's just what you have to do these days. 

On the upside, most of the explanations were quite good.  (The business about "lumping" and "splitting" aside--those are common terms that work well, but are not terms of art.)  

A few times I had to dive a little deeper into the topic area to fully understand what was going on.  But that's a big part of The Joy of Search.  

Hope you enjoyed this week's Challenge.  

Keep searching. 

SearchResearch Challenge (10/8/25): How can the same locust look so different?

It's difficult to understand... 

Rocky Mountain locust. P/C Wikimedia

... how variable the appearance of an animal might be.  

Sure, people look very different around the globe, and both dogs and cats have wildly variable appearances.  But in every case, you'd say that they're all of one species. 

So how could biologists mistake the two different looks of a locust for two different species?  

A bit of background here.

I've been reading Jeffrey A. Lockwood's brilliant book Locust: the devastating rise and mysterious disappearance of the insect that shaped the American frontier. (Basic Books, 2009.) 

Part of the book tells the story of the Locust Plague of 1874.  Locusts swarmed over an estimated 2,000,000 square miles (5,200,000 square kilometers) of the plains states in North America, causing millions of dollars' worth of damage. 

Residents described swarms so thick that they covered the sun for up to six hours. The swarms of Rocky Mountain locusts (Melanoplus spretus) were larger than the state of California and comprised some 12.5 TRILLION insects.

They would eat grass, trees, even the clothes off people's backs.  

But less than 30 years later, the entire species was extinct. Gone.  Vanished.  

That's the subject of Lockwood's book--how is it possible for such a vast number of insects to simply disappear?  

A cartoon of the locusts arriving in Nebraska

Laura Ingalls Wilder’s book, On the Banks of Plum Creek  has a description of what it was like to live through the literal plague of locusts arriving on the farm:  

Plunk! something hit Laura's head and fell to the ground. She looked down and saw the largest grasshopper she had ever seen. Then huge brown grasshoppers were hitting the ground all around her, hitting her head and her face and her arms. They came thudding down like hail. 

The cloud was hailing grasshoppers. The cloud was grasshoppers. Their bodies hid the sun and made darkness. Their thin, large wings gleamed and glittered. The rasping whirring of their wings filled the whole air and they hit the ground and the house with the noise of a hailstorm.

You might think of this extinction as the most spectacular “success” in the history of economic entomology — the only complete elimination of an agricultural pest species.  But it seems as if it was a total accident.  

(For all the details, I encourage you to read Lockwood's book--a fascinating detective story of a past extinction. Also check out the Wiki articles Locust Plague of 1874 and Rocky Mountain locust. For more details, Lockwood has a short article about his sleuthing, The Death of the Super Hopper.)  

But that's not our Challenge for this week.  Instead, I want to focus on that first question I raised earlier--So how could biologists mistake the two different looks of a locust for two different species?  

1. How often has it happened that biologists have seen two (or more) species when it was really just one in different clothing?  Can you find another case of two (or more) species being reconciled into one? 

2. It's clear that organisms can have multiple shapes / patterns / colors (we've discussed this before in the context of plant mimicry).  Can you find an organism that has a huge number of different appearances?  Any idea WHY they have such variability?  

It's fascinating stuff--hope you enjoy reading about it as much as I did.  

Be sure to tell us HOW you found the answers to this week's Challenge.  Regular search?  AI?  If so, what prompts did you use... and how well did it work for you?  

We want to hear about successes as well as disasters! 

Keep searching.  

Answer: What's the story with the greenhouses?

 Greenhouses... 

... come in two types--decorative (the ones with beautiful orchids and flowers for visitors to see) and functional ones (see the images above, for growing food and flowers).  

In some places, (e.g., Weifang, China) greenhouses spread over more than 820 square kilometers. 

Seeing all of these from the air sparked this week's Challenge with a few curious questions for you to ponder.  Can you find the answers?  If so, what did you do to discover the results?  

1. How long have greenhouses been around?  If greenhouses date to around Roman times (as I've heard), what were the greenhouses made of?  

I started with a simple: 

[ history of greenhouses] 

as a way to find reasonable resources and read from them directly.  The Wikipedia article on Greenhouses tells of an early origin (30 AD) when the Roman emperor Tiberius needed a "cucumber a day" to keep him in the best of health.  Clever people then made simple frames, "cucumber houses" glazed with either oiled cloth known as specularia or with sheets of selenite.  That's according to the historical description by Pliny the Elder. 

Oiled cloth I understand, but selenite?  It's a crystal that sometimes forms sheets (like mica), but I didn't think it would be very large.  But, a quick search for [selenite] and a couple of images of mineral sample showed me that they can form reasonable sizes rectangles that would be good  for making greenhouse window panes.  (You can even buy nice rectangles of selenite on Amazon.  Who knew?)  

Of course I read several articles to see if they all agreed--and they do--with the most authoritative voice coming from a paper in Horticultural Science (History of Controlled Environment Horticulture: Ancient Origins, by Jules Janick and Harry Paris) 

So... greenhouses have been around for at least 2000 years, with early Roman greenhouses covered in oiled-cloth or small windows of selenite.  

2. What is growing under all of those greenhouses? What's grown in Weifang that needs SO many greenhouses? 

Searched for:  

     [ vegetables grown in greenhouses in Weifang ] 

led me to lots of sources, including an interesting video on X where poster Teacher James walks through some of those greenhouses, pointing out what's growing there.  Answer: LOTS of veggies--I spotted sweet potatoes, pumpkins, tomatoes, eggplants, many kinds of greens, and flowers.  There was no one thing in particular--just lots of varieties.  

Answer: Everything, mostly veggies, of a huge variety and number.   

I also found a really beautiful illustration of side-by-side / before-and-after NASA Earth images at A Greenhouse Boom in China.

Try this yourself at the NASA website.

If you visit the page, you can pull the slider left and right to see how individual places have changed.  On the left, green area are/were regular fields open to the sky.  On the right, gray areas are greenhouses.  That's a huge change in 37 years!  

3. Those robotic greenhouses... how well are they doing?  Has there been a boom in robotic and/or vertical greenhouses in the past 10 years?  Is it a growth industry?  

It's pretty clear that robots are (and will) change the way agriculture works.  The bots aren't perfect, but getting better all the time.  

Let's split this Challenge into two parts: (a) How well are robots in greenhouses doing?  and (b) How well are vertical greenhouses doing?  

Robots: I put this question to Gemini 

[how well are robots working out in greenhouses? Show analyses of how effective robots are at working in greenhouses.  Show both upsides and downsides of robots in greenhouses.] 

This gave me a fairly generic answers, and (annoyingly) without citations.  I had to do an additional query to get the citations, only to find they were mostly old-ish. In a field that's moving as rapidly as this, references from 2020 are not super-relevant.  I had to ask for ONLY works from 2024-2025.  

By the 3rd query, I finally got a reasonable answer in the form of a tradeoff table (pluses and minuses). It still didn't give references, so I had to ask a 4th query. 

When I tried the same task with Perplexity, I got somewhat better results (with citations and links included), although many of the sources were very upbeat promotional sites for robo-agricutural companies.  That's useful data, but they don't show the downsides.  

But it DID lead me to a recent paper (late 2024) that's a survey of robots in greenhouses: "Robots in greenhouses: A scoping review" with an extensive set of citations covering all of the issues and tech needed to build successful working greenhouse robots.  

The paper concludes by pointing out the key problems in building greenhouse robots: they're slow, they're expensive, and they need constant maintenance.  

I also asked Gemini, ChatGPT, Claude, and Perplexity for their opinions (using the above prompts), and mostly got wide-ranging agreement.  The tech is cool and is attracting investment, but 

Bottom line:  There's a lot of promise in putting robots into greenhouses--but there are also a lot of difficult technical and economic problems to solve first.  

Some greenhouse-robot companies (like Iron Ox) go under after a few years of promising work.  The heavy capital expenditures, combined with the economics of operating in such a competitive and cost-sensitive domain, seems to make sustained growth untenable — a common challenge in high-tech farming circles.

Vertical farming test installation. P/C USDA

Vertical Greenhouses: I put this question to our favorite LLMs: 

 [how well are vertical greenhouses working out ? Show analyses of how well vertical greenhouses are performing economically.  Show both upsides and downsides of vertical greenhouses. Give citations.  Include work done in 2024-2025. ] 

Notice that I've learned from the previous queries: This time I added in a focus on economics and explicitly said that I want citations and recent work from 2024-2025.  

This change in the prompts gives much more useful results.  As Gemini says, 

"Vertical greenhouses (or vertical farms) are technically proficient but continue to face significant economic hurdles, primarily high operating costs. While the market is experiencing rapid expansion, the industry is currently undergoing a "shakeout" where many early ventures that scaled too aggressively are failing, while more efficient, specialized operators are beginning to achieve profitability."  

They're very efficient, but have large energy costs, and only work well for leafy greens and--significantly--not for the high value veggies that grow on vines or are big and difficult to handle (think cantaloups).  ChatGPT pointed out that: 

"The business is highly price-sensitive to electricity: with retail/industrial rates spiking the energy component alone can reach ~$6.75/kg lettuce—destroying margin unless prices normalize or power is hedged or renewable."

 

Bottom line:  It's a tough world out there for putting high-tech / new-tech into difficult environments, and vertical greenhouses, while promising, still have the old problem of making money.  Economics never goes away, no matter how shiny the tech.  

SearchResearch Lessons 

1. As you dig deeper into a topic, modify your query / prompts to get more of what you need.  Pay attention!  As you saw, between my first and second prompts I realized that I needed more of a research topic focus (economics) AND that I wanted citations AND only results from the past 2 years.  That all went into my modified prompts.  

2. LLMs (Gemini, ChatGPT, Claude, etc.) work quite well for topics that call for broad ranging synthesis.  Yes, there are still errors of the types we've mentioned in previous posts.  But as Mike Caulfield helpfully points out in his post When wrong answers get you to the right information, you can often refine your question, ask a follow-up question, or use the citations given by the AI to lead you to useful resources. 

3. Be skeptical.  Be VERY skeptical of self-serving pronouncements by industry websites, articles, or promo pieces.  It's easy to find articles telling you that things are going incredibly well and that the market size will be $X billion in 5 years.  Be dubious, be skeptical.  Look to see how things are going now, and see if you can plausibly find a path from current conditions to a realistic future.   

 

Keep searching.