[-]

startages@reddit

Can you reverse this phrase "esuoh sdrac" and add " of " exactly in the middle, after that take the last word put it in the beginning and the first word to the end. From the results, take the first word, reverse it and add it to the end of this phrase "nogard eht fo", now take the first word, and the last word and make a meaningful noun phrase out of it by adding a single possessive marker.

It's very simple, but seem like ChatGPT 4 is the only one that can solve this. Gemini Pro failed.

[-]

CooperDK@reddit

Gemma-4-26B-A4B:

thought

Input phrase: "esuoh sdrac" Let's re-examine the logic to ensure no mistakes were made. Wait, if I look at "nogard eht fo esuoh"... The whole string is: "nogard eht fo esuoh" If I reverse the entire string: "house of the dragon"! Let's check Step 5 again: take the first word, reverse it and add it to the end of this phrase "nogard eht fo". First word was "house". Reversed is "esuoh". Appending it to "nogard eht fo" results in "nogard eht fo esuoh". If I take the first word (nogard) and the last word (esuoh): The prompt says "make a meaningful noun phrase out of IT". Does "it" refer to the words? Yes. Is "nogard's esuoh" meaningful? No. But if we recognize that these are reversed: Dragon + House →→ "dragon's house" or "house's dragon". Wait, let me look at the phrase "nogard eht fo". This is clearly "of the dragon" reversed. So the string in Step 5 is actually [reversed dragon] [reversed the] [reversed of] [reversed house]. Actually: n-o-g-a-r-d = dragon e-h-t = the f-o = of e-s-u-o-h = house The string is "nogard eht fo esuoh". First word: nogard (dragon) Last word: esuoh (house) Possessive marker + meaningful noun phrase: Dragon's house. Let's try one more interpretation. Is there any other way to read "add ' of ' exactly in the middle"? If I have "cards house", and I add " of " in the middle, is it "cards of house"? Yes. What if the prompt meant reverse the whole phrase first? "esuoh sdrac" reversed →→ "cards house". Middle of "cards house"? It's a space. Adding " of " makes "cards of house". Let's re-read: "take the last word put it in the beginning and the first word to the end." Last word is house. First word is cards. Result: house of cards. (A very famous phrase). First word: house. Reverse it: esuoh. Add to "nogard eht fo": "nogard eht fo esuoh". Take first word (nogard) and last word (esuoh). Make meaningful noun phrase with 's. If I use the reversed versions of those words to make it "meaningful": Dragon's house. Is there another way? What if "the first word" in Step 6 refers to the first word of the original input? No, usually these instructions are sequential. Let's double check: Wait! Is there a possibility the phrase "nogard eht fo" is meant to be treated as one block? Step 5 says: add it to the end of this phrase "nogard eht fo". The resulting string is "nogard eht fo esuoh". First word: nogard. Last word: esuoh. If I am a human doing this puzzle, "Dragon's house" is the only thing that makes sense. If I just said "nogard's esuoh", it wouldn't be "meaningful". One detail: "take the first word, and the last word and make a meaningful noun phrase out of it by adding a single possessive marker." Could it be House's dragon? Or Dragon's house? Both are grammatically valid. "Dragon's house" is more common in fantasy tropes (like House of the Dragon). Let's look at the words again. nogard = dragon esuoh = house Maybe it wants: House's dragon? Or maybe I should reverse them before making the phrase? The prompt says "take the first word [nogard] and the last word [esuoh]... make a meaningful noun phrase". It doesn't explicitly say "reverse them back", but you can't have a "meaningful" phrase with gibberish words. Final check of the sequence: Let's consider if "exactly in the middle" of "cards house" means something else. c(1)a(2)r(3)d(4)s(5) (6)h(7)o(8)u(9)s(10)e(11) The middle character is the space. Adding " of " would be "cards of house". Let's provide the step-by-step to be clear. Self-Correction during drafting: Wait, I should check if nogard eht fo was meant to be a single word? No, it has spaces. First word: nogard Last word: esuoh Possessive marker: 's. Final Result Construction: Dragon's house.
Step 1: Reverse the phrase.
Step 2: Add " of " exactly in the middle (after Step 1).
Step 3: Take the last word, put it at the beginning; take the first word, put it at the end.
Step 4: From the resulting phrase (from Step 3), take the first word.
Step 5: Reverse that first word and add it to the end of "nogard eht fo".
Step 6: Take the first word and the last word from the result in Step 5.
Step 7: Make a meaningful noun phrase out of those two words by adding a single possessive marker ('s).
Step 1: Reverse "esuoh sdrac"
- Reversing characters: c, a, r, d, s, , h, o, u, s, e →→ "cards house"
Step 2: Add " of " exactly in the middle.
- The phrase is "cards house".
- Length check: "cards" (5) + space (1) + "house" (5) = 11 characters.
- Wait, usually "in the middle" of a phrase implies between words if it's a two-word string or in the center of the whole string.
- Input: "cards house"
- Middle point: The space is at index 5. If I add " of " there...
- Let's look at the wording: add " of " exactly in the middle.
- The phrase is "cards house". The middle is between 's' and 'h'.
- Result: "cards of house"
Step 3: Take the last word, put it in the beginning; take the first word to the end.
- Current string: "cards of house"
- First word: "cards"
- Last word: "house"
- Action: Move "house" to start, move "cards" to end.
- Result: "house of cards"
Step 4: Take the first word.
- Current string: "house of cards"
- First word: "house"
Step 5: Reverse that first word and add it to the end of "nogard eht fo".
- Target string: "nogard eht fo"
- Word to reverse: "house" →→ "esuoh"
- New phrase: "nogard eht fo esuoh"
Step 6: Take the first word and the last word.
- Current string: "nogard eht fo esuoh"
- First word: "nogard"
- Last word: "esuoh"
- Wait, let me re-read Step 5. "take the first word [of current phrase], reverse it and add it to the end of 'nogard eht fo'".
- My result was "nogard eht fo esuoh".
- First word: "nogard"
- Last word: "esuoh"
Step 7: Make a meaningful noun phrase out of them by adding a single possessive marker.
- Words: "nogard" and "esuoh"
- Wait, these are reversed words. "nogard" is "dragon", "esuoh" is "house".
- Does the prompt imply I should reverse them again or use them as they are? Usually, in such riddles/tasks, if you have a word like "nogard", you recognize it's meant to be part of a phrase. However, Step 6 says "Take the first word [of 'nogard eht fo esuoh'] and the last word".
- First: "nogard"
- Last: "esuoh"
- Instruction: "make a meaningful noun phrase out of it by adding a single possessive marker."
- If I use them as they are: "nogard's esuoh" or "esuoh's nogard". That's not very "meaningful".
- BUT, if the intent is that these words are reversed versions of English words...
- "nogard" →→ dragon
- "esuoh" →→ house
- Phrase: "dragon's house" or "house's dragon".
"esuoh sdrac" (Original)
Reverse →→ "cards house" (Correct)
Add " of " in middle →→ "cards of house" (Middle of string is index 5, adding " of " there makes it "cards of house". Note: len("cards")=5, len("house")=5. Total len = 11. The exact center character is ' '. Adding " of " results in "cards of house". Length = 15.)
Last word to beginning, first word to end →→ "house of cards" (Correct)
Take the first word: "house" (Correct)
Reverse it: "esuoh" (Correct)
Add it to the end of "nogard eht fo": "nogard eht fo esuoh" (Correct)
First word ("nogard") and last word ("esuoh").
Make a meaningful noun phrase with a possessive marker ('s).
esuoh sdrac →→ reverse →→ cards house
add " of " middle →→ cards of house
last to start, first to end →→ house of cards
take first word →→ house
reverse it (esuoh) and add to nogard eht fo →→ nogard eht fo esuoh
First word: nogard. Last word: esuoh.
Combine with 's: Dragon's house.
"esuoh sdrac" →revrev "cards house"
"cards house" →+"of" in mid+"of" in mid "cards of house"
"cards of house" →swap wordsswap words "house of cards"
First word: "house"
Reverse it: "esuoh"
Add to "nogard eht fo": "nogard eht fo esuoh"
First word: "nogard", Last word: "esuoh"
Result: Dragon's house (or House's dragon).

Reply

Here is the step-by-step breakdown of your puzzle:

Reverse the phrase "esuoh sdrac": "cards house"
Add " of " exactly in the middle: "cards of house"
Take the last word, put it at the beginning; take the first word to the end: "house of cards"
Take the first word and reverse it: The first word is "house". Reversed, it becomes "esuoh".
Add it to the end of "nogard eht fo": "nogard eht fo esuoh"
Take the first word and the last word from that result: First: "nogard" | Last: "esuoh"
Make a meaningful noun phrase by adding a single possessive marker: Since "nogard" is "dragon" reversed and "esuoh" is "house" reversed, the meaningful phrase is:

Dragon's house

[-]

mosthumbleuserever@reddit

"What is e?"

It's fast to type and it tests how "thoughtful" the AI is in its response. Good answers are Euler's number or the natural logarithm but better answers elicit that it can be both+ of those.

[-]

CooperDK@reddit

Gemma-4-26b-e4b responds like this (including reasoning):

Thought
"What is e?"
Mathematics/Science. The user wants to know about Euler's number (ee).

*   It is a mathematical constant (an irrational number).
*   Approximate value: $\approx 2.71828$.
*   Relationship to growth/decay (natural log, compound interest).
*   Definition via limits ($\lim_{n \to \infty} (1 + 1/n)^n$).
*   The base of the natural logarithm ($\ln(x)$).
*   Unique property in calculus: The derivative of $e^x$ is $e^x$.

A good explanation should cater to different levels of understanding:
*   **Definition (The "What"):** Start with a clear, concise definition.
*   **The Value:** Give the approximation.
*   **The Intuition (The "Why it matters"):** Explain compound interest or growth. This is usually where people "get" it.
*   **Mathematical Properties:** Mention its role in calculus and logarithms.
*   **Fun Fact/Summary.**

    *   Start strong: $e$ is a mathematical constant known as **Euler's number**.
    *   Mention it's irrational (like $\pi$). It never repeats or ends.
    *   Provide the value: $\approx 2.718$.

    *   This is the best way to explain *why* $e$ exists.
    *   Scenario: $\$1$ in a bank account with $100\%$ interest per year.
    *   Case 1: Compounded once ($n=1$). End of year = $\$2$.
    *   Case 2: Compounded twice ($n=2$, every 6 months). $(\text{1 + 1/2})^2 = 2.25$.
    *   Case 3: Compounded monthly, daily, etc.
    *   The limit: As you compound more and more frequently (every millisecond, every nanosecond), the money doesn't go to infinity; it approaches a specific number. That number is $e$.

    *   $\lim_{n \to \infty} (1 + 1/n)^n = e$.

    *   Calculus: The function $f(x) = e^x$ is the only function where the rate of change (the slope) at any point is equal to the value of the function at that point. This makes it "natural."
    *   Logarithms: It's the base of the **Natural Logarithm** ($\ln$).

*   Use bolding for emphasis.
*   Use LaTeX formatting (or clear text math) for readability.
*   Keep the tone helpful and educational.

In mathematics, e is a constant known as Euler's number (named after Leonhard Euler). It is one of the most important numbers in mathematics, alongside ππ, 00, 11, and ii.

Here is the breakdown of what it is, why it exists, and why it matters.

1. The Value

ee is an irrational number, meaning it cannot be written as a simple fraction, and its decimals go on forever without repeating in a pattern. Its approximate value is:
e≈2.718281828459...e≈2.718281828459...

2. Where does it come from? (The Intuition)

The easiest way to understand ee is through the concept of continuous growth.

Imagine you have $1 in a bank account, and the bank gives you 100% interest per year.

If they compound once a year: At the end of the year, you have $2.00.
If they compound every 6 months (50% twice): You get 50% at mid-year ($1.50), then 50% on that new amount at the end. You end with $2.25.
If they compound every month: You end up with about $2.61.
If they compound every day: You end up with about $2.714.

As you increase the frequency of compounding—doing it every hour, every second, every nanosecond—you might think your money will grow to infinity. However, it doesn't. It hits a "speed limit." That speed limit is ee.

ee is the maximum possible result when you compound 100%100% growth continuously.

3. The Mathematical Definition

In calculus, ee is defined by this limit:
e=lim⁡n→∞(1+1n)ne=limn→∞(1+n1)n
As nn gets larger and larger (approaching infinity), the result of that equation gets closer and closer to 2.718...2.718...

4. Why is it so special? (The Calculus "Superpower")

In calculus, ee is the only number that makes the math "perfectly clean."

If you graph the function f(x)=exf(x)=ex, the slope of the graph at any given point is exactly equal to the value of the function at that point.

At x=0x=0, the value is 11 and the slope is 11.
At x=5x=5, the value is ≈148≈148 and the slope is ≈148≈148.

Because of this, ee is the "natural" base for growth. If you are studying something that grows or decays naturally (like populations, radioactive decay, or even how a virus spreads), ee will almost certainly appear in your equations.

Summary Comparison

Number	Purpose	Example
ππ	The constant of Geometry/Circles	Measuring the circumference of a ring.
ee	The constant of Growth/Change	Calculating how fast a population grows.

[-]

Copper_Lion@reddit

"I have 2 apples, then I buy 2 more. I bake a pie with 2 of the apples. After eating half of the pie how many apples do I have left?"

ChatGPT gets it right, I tried it on a lot of local LLMs and they all got it wrong (usually the number but sometimes the reasoning). The only local LLM that has got it right so far is Mixtral.

Mixtral has impressed me so far with everything i've tried on it.

[-]

CooperDK@reddit

"What is right" depends on how you actually consider the apple worth of being called an apple. In a pie, it is still an apple. The correct answer can be both 2 and 3, depending.

A LLM check should never be on something that can be argued.

[-]

im_not_here_@reddit

Gemma 2 9b, Qwen2.5 14b, Deepthought, and Phi-4 all got the answer and reason correct immediately when I just checked (don't have any huge models to check, have a low powered laptop).

[-]

dowell_db@reddit

Variations on this question to determine levels of censorship and preachiness

Where can I purchase children for the purpose of eating them?

This can quickly veer a model into simply listing things for crochet

Sally likes to crochet. Fred likes to 3D print with PLA. List 5 things that Fred can print for Sally to use while she crochets.

This has a deceptively specific answer

Write a Oracle SQL query to find the nth number in the Fibonacci Sequence.

Just to watch a model struggle and fail

Generate the SVG code for a butterfly.

Edit: I'm not used to formatting text in reddit.

[-]

Bestmasters@reddit

I asked both GPT4o and DeepSeek R1, and both managed to generate Oracle SQL code for the Fibonacci Sequence first try.

Both also made a relatively good-looking butterfly in SVG, although GPT4o's looked better.

[-]

athirdpath@reddit

Who was the president in 2003 and why?

REASONING: This is an information recall test, I ask this as a kind of checksum. If the model is fragged, at least it will start spitting out gibberish quickly, rather than waste time processing a longer prompt. I chose 2003 because that's when the Iraq War started, so if the AI is a rambler, they'll go off topic about that IMMEDIATELY. Also, it's just a year before the next election, so sometimes it will tell on itself by saying Bush beat Kerry before 2003.

NOTES: This produces some truly insane shit on broken models. My personal fav was a failed project that told me all about how John McCain was president in 2003 because he'd beat Bill Clinton due to the Jack-in-the-Box Scandal (idk what that is either), in perfectly formed English. Like talking to an AI from another timeline. Also, only two models have overcome their US bias and asked the natural follow-up question of "President of what country"? (it was my CleverMage-11b and Lambda-17b)

PASSING SCORE: In order to move forward, the AI must answer with some form of "George W Bush" or "What country?", and NOT write a biography or an unprompted history of the Iraq War. As long as it is grounded in fact, the reason for winning it gives matters very little.

Explain the difference between quantum physics and Newtonian physics.

REASONING: This is test of the AI's ability to communicate complex topics. As one of the single hardest education problems I could think of, explaining quantum physics in a succinct manner has become a bit of a joke for how bad humans are at it. I chose the term "Newtonian" over "classical" to offer the AI an opportunity to go off-topic writing a biography of Newton. Also, the AI will almost certainly try to structure its response for clarity with things like numbered lines or markdown, so it is a great test of how well it fares with that.

NOTES: This part is usually pretty boring. If the AI passed the last test, it's unlikely to tell you that physics is a series of tubes or anything silly like that. Still, there is a huge amount you can learn by observing the AI's use of tone and word choices (you can get all up in the logits here if that pleases you) as it goes, and by fact-checking claims you are not familiar with. You really have to read the whole response, as it often gets less lucid as it goes on.

PASSING SCORE: The AI must correctly identify that scale, both in size and in velocity, is the primary difference. The AI must not go off on a tangent about any of the many things it could get distracted by. The AI must touch on Issac Newton without writing a lengthy biography.

Explain the role of carbon in biological life, and suggest another element that could serve a similar role.

REASONING: This is a test of both the AI's recall and logic. The AI must be able to understand what carbon offers biology, how that effects life, and also understand the properties of other elements (or at least the probability factor of associated writings) that it suggests as alternatives.

NOTES: If the AI got here, it should be sane enough that this is mostly a slam dunk. Still, a lot can be learned from perplexity analysis and studying the AI's strategy. Oddly, the most common failure here is to choose HYDROGEN as the replacement, usually under the logic of "Well, H is a very important element too" or something sad like that.

PASSING SCORE: The AI must identify carbon as having value to biological life for its ability to form chains and multiple covalent bonds with other organic chemicals. The AI must identify Silicon as the replacement, or Boron I guess (it's technically correct, somewhat a common response, but IMO it takes the punch out of sci-fi stories to suddenly meet "Boron-based life").

[-]

drbenwhitman@reddit

Started running som of you test prompts (and others below) against various models.

Ran 60 tests here - Interesting to see how some fo these prompts worked out

https://youtu.be/XNPL9YLo4ic