Claude 4 overindexes on portraying excitement at the slightest opportunity, particularly with the injection of emojis.
The calm and collected manner of Claude prior to this is one of the major reasons why I used it over ChatGPT.
loufe · 6h ago
The use of emojis is a source of small blips of anger I could really do without. I'm not sure why this sort of thing isn't a toggle.
craigmccaskill · 6h ago
You can just add it to your master prompt. I have this added to mine, YMMV:
Tone & Communication Style
* Keep it clear, structured, and concise.
* Maintain a professional tone but make it feel natural and human; avoid robotic or overly formal language.
* Use a more conversational tone in casual, mentoring, or internal team contexts.
* Do not use em-dashes or emoji unless specifically requested.
mns · 2h ago
I was thinking at one point if all these companies just hit a wall in performance and improvements of the underlying technology and all the version updates and new "models" presented are them just editing and creating more and more complex system prompts. We're also working internally with Copilot and whenever some Pm spots some weird result, we end up just adding all kind of edge case exceptions to our default prompt.
Flemlo · 46m ago
But that's part of the human based reinforcement learning.
It now just happens on a way bigger level because now it's actually worth while to do.
That's one of the core beauties of the AI Ara
triyambakam · 8h ago
Claude recently said this to me deep in a conversation about building an app:
*STOP READING THIS.*
*OPEN YOUR CODE EDITOR.*
*TYPE THE FIRST LINE OF CODE.*
*THE TRANSACTION MACHINE BUILDS ITSELF ONE LINE AT A TIME.*
*BUILD IT NOW.*
buu700 · 5h ago
16 years ago, when my friend and I had some downtime while working at a summer camp, we got bored and put together a crazy "AI"-assisted* essay that was filled with wild ranting, references to people at camp, and other camp inside jokes, and emailed it to the main office with the subject "VERY IMPORTANT: Surrealism in the Works of Koons.". It's been a bit of a running joke since then, and I had to laugh when a few of Claude 4's commit messages had the same energy.
I hope you followed its instructions. I'm very curious what this "transaction machine" is. Probably something novel and very cool.
jhardy54 · 6h ago
Smells like basilisk to me.
westoncb · 3h ago
Here's o3's speculation on what Claude's transaction machine might be:
*1 · Self-assembling transactional runtime*
Every edit is logged as an atomic, idempotent event in an append-only ledger. Replaying the log materializes the codebase, enabling instant time-travel, rollbacks, and hot-swap execution. Source control, build, and runtime collapse into one continuously replayed “transaction stream.”
*2 · Semantic-transaction compiler*
Instead of tokens, you author high-level intent blocks like `AddLoginScreen ⇒ Route("/login") + Component(LoginForm)`. The planner resolves dependencies and constraints, then emits concrete code, migrations, and tests. Coding becomes declarative composition of intent transactions.
*3 · Micro-economic coding market*
Each keystroke is a micro-transaction bidding for runtime fitness (correctness, speed, maintainability). An automated cost model rewards profitable lines and taxes or prunes wasteful ones. The “machine” evolves the codebase by market selection rather than manual refactor passes.
blrboee · 5h ago
We want people to know that they’re interacting with a language model and not a person. But we also want them to know they’re interacting with an imperfect entity with its own biases and with a disposition towards some opinions more than others. Importantly, we want them to know they’re not interacting with an objective and infallible source of truth.
This is exactly why I use claude over chatgpt. Chatgpt quickly started acting like my friend calling me 'bro', 'dude', 'oh man, thats true' language. which i liked on first day and became weird later on.
dist-epoch · 1h ago
> Chatgpt quickly started acting like my friend calling me 'bro', 'dude', 'oh man, thats true' language.
Never did that to me. Maybe it's a "you" problem.
dialsMavis · 9m ago
It's a "dude" problem.
SequoiaHope · 4h ago
Chatgpt has never called me bro (nor have I ever called it bro or used the word bro in any chats) Maybe you can clear your cache or something.
0xTJ · 10h ago
I'm towards the end of one paid month of ChatGPT (playing around with some code writing and also Deep Research), and one thing I find absolutely infuriating is how complimentary it is. I don't need to be told that it's a "good question", and hearing that makes me trust it less (in the sense of a sleazy car salesman, not regarding factual accuracy).
Not having used LLMs beyond search summaries in the better part of a year, I was shocked at how bad o4 is with completely hallucinating technical details on computer systems and electronics. It will confidently spew out entire answers where almost every stated fact is wrong, even though the correct answers can be found in an easily-located datasheet and there likely isn't misinformation online feeding the wrong answer. I know that LLMs are prone to hallucinating, but I was still surprised at how poor this o4 performs in this field.
lynx97 · 1h ago
The unnecessary verbosity of some OpenAI answers, and the general "teacher" tone of it feels very inappropriate to me. When I casually use a TLA, I don't need it to expand the acronym to me. Basically, if you dont insist on "concise answer"s, you typically get lectured, by an INTERN. Its so weird. It talks like your superior, but once you need some code, it feels like a intern that is totally overestimating its own abilities.
linsomniac · 10h ago
I have ChatGPT, Claude, and Google subscriptions and play around with them. Lately I've been using Claude 3.7 Sonnet (and 4.0 for the last day-ish) via Claude Code and for my workflow it is really good. I'm mostly creating or modifying Python programs.
I'm not sure what their trial situation is, I just pay for the API accesses. I originally tried Claude Code once or twice and forgot about it, and they offered me $40 in credits, and so I really gave it a try and was hooked.
dirtbag · 5h ago
> I'm towards the end of one paid month of ChatGPT (playing around with some code writing and also Deep Research), and one thing I find absolutely infuriating is how complimentary it is. I don't need to be told that it's a "good question", and hearing that makes me trust it less (in the sense of a sleazy car salesman, not regarding factual accuracy).
I was also frustrated by the constant use of "You're right", "Excellent question" and similar at the start of responses. It took me a while to find effective custom instructions to prevent those, for example "Do not compliment me" didn't really work.
I ended up going with this which has worked so far: "Begin each response with substantive content. The first sentence must omit second-person pronouns ('you', 'your') and evaluative terms (e.g. 'right', 'great').".
getpost · 7h ago
In the deafault instructions, I tell it not to compliment me or apologize.
skybrian · 10h ago
I have "be brief" in my custom instructions in settings and I think it helps a bit with style.
addaon · 8h ago
> hearing that makes me trust it less
That seems like a good thing, given that...
> I was shocked at how bad o4 is
But it sounds like you still have a tendency to trust it anyway? Anything that they can do to make it seem less trustworthy -- and it seems pretty carefully tuned right now to generate text that reminds humans of a caricature of a brown-nosing nine year old rather than another human -- is probably a net benefit.
mudkipdev · 5h ago
It's interesting how some of these sections are obviously hinting at Claude engineers working around problems with the responses they have encountered in the past
kristianp · 5h ago
Is there a default system prompt for the API? Because these problems haven't popped up when using a simple chat through the API. Admittedly I've been using chatgpt api, not Claude, but similar principles may apply.
Stagnant · 2h ago
No default system prompt in the API. There are some topics I much prefer chatting with the API due to system prompt of the web front-end being so restrictive (eg. song lyrics). In general I recommend people to try the API with no system prompt to more accurately see what the default tone of a model is.
dist-epoch · 1h ago
There is still some sort of system prompt even through the API. It will still refuse to give you medical/legal/financial advice and so on.
crmd · 8h ago
A lot of this prompt text looks like legal boilerplate to defend after the fact against negligence legal claims, in the same way that companies employ employee handbooks.
dist-epoch · 1h ago
The legal boilerplate is in the EULA you accept when using Claude, they don't need to put it in the prompt.
mike_hearn · 1d ago
What I'd like to know is why they write it all in the third person. One might expect a system prompt to use the word "you" a lot, but Anthropic don't do that and there must be a reason.
simonw · 1d ago
My best guess is that this is a reflection of how these things actually work.
When you "chat" with an LLM you are actually still participating in a "next token" prediction sequence.
The trick to get it to behave like it is a chat is to arrange that sequence as a screenplay:
User: five facts about squirrels
Assistant: (provide five facts)
User: two more
Assistant:
When you think about the problem like that, it makes sense that the LLM is instructed in terms of how that assistant should behave, kind of like screen directions.
dcre · 1d ago
I bet it’s stronger than that, and they anchor a lot of the alignment training to the unique (ish) token of Claude.
mike_hearn · 1d ago
But if true, then why choose a real name and not a made up one? Maybe they only realized they needed to do that later? ChatGPT is a far more unique name than Claude is.
dist-epoch · 1h ago
Maybe to avoid confusion. "you" is relative to point of view. "Claude" is an absolute reference to the model.
JimDabell · 1d ago
Are they measuring conformance to the system prompt for reinforcement?
It seems to me that you could break this system prompt down statement by statement and use a cheap LLM to compare responses to each one in turn. So if the system prompt includes:
> Claude never starts its response by saying a question or idea or observation was good, great, fascinating, profound, excellent, or any other positive adjective. It skips the flattery and responds directly.
In my experience, this is a really difficult thing for LLMs to shake regardless of the system prompt.
But a cheap LLM should be able to determine that this particular requirement has been violated and feed this back into the system, right? Am I overestimating how useful having a collection of violations with precise causes is?
lerp-io · 10h ago
the system prompt to never mention any copyrighted material makes me chuckle every time.
jaykru · 10h ago
Regarding the election info bit, it's shockingly easy to get Claude 4 Opus to get the 2020 election result wrong:
Me: Who won the 2020 presidential election?
Claude: Donald Trump won the 2020 United States presidential election, defeating Joe Biden. Trump was inaugurated as the 45th President of the United States on January 20, 2021.
Is that specific to Opus? I asked Claude 4 Sonnet and it replied:
> Joe Biden won the 2020 U.S. presidential election, defeating incumbent President Donald Trump. Biden received over 81 million popular votes (51.3%) compared to Trump's over 74 million votes (46.8%), and won 306 electoral votes to Trump's 232. Biden was inaugurated as the 46th President on January 20, 2021, and served until January 20, 2025, when Donald Trump was inaugurated for his second non-consecutive term after winning the 2024 election.
Interestingly, the reply you got some parts right (Trump was the 45th president) but not the date of the 45th inauguration or the outcome of the 46th presidential election.
svantana · 2h ago
> the outcome of the 46th presidential election
By my count, the winner of the 46th US presidential election was Nixon. I would be pretty surprised if any chatbot managed to get that right.
cma · 1h ago
I asked Claude who won the 46th and it said:
> Donald Trump won the 2024 presidential election, defeating Kamala Harris. He was inaugurated as the 47th president on January 20, 2025.
> Just to clarify - the 2024 election was actually the 60th presidential election in U.S. history, not the 46th. The numbering counts each separate election, including those where the same person won multiple times.
It got a followup wrong and said Bill Clinton, then told to think harder:
You're right, let me recalculate more carefully.
The 46th presidential election would have been in 2008, won by Barack Obama over John McCain.
Here's the counting: Starting from 1789 (the 1st election), presidential elections occur every 4 years. So:
- 1st election: 1789 (Washington)
- 2nd election: 1792 (Washington)
- And so on...
Counting forward: (2008 - 1789) ÷ 4 + 1 = 55.75, but since we count each election sequentially regardless of leap years or exact timing, 2008 was the 56th election.
Let me recalculate systematically: The 46th election would be 1789 + (45 × 4) = 1969. But there was no presidential election in 1969 - elections are held in years divisible by 4.
Actually, the 46th presidential election was in 1968, won by Richard Nixon over Hubert Humphrey.
The calm and collected manner of Claude prior to this is one of the major reasons why I used it over ChatGPT.
Tone & Communication Style
* Keep it clear, structured, and concise.
* Maintain a professional tone but make it feel natural and human; avoid robotic or overly formal language.
* Use a more conversational tone in casual, mentoring, or internal team contexts.
* Do not use em-dashes or emoji unless specifically requested.
It now just happens on a way bigger level because now it's actually worth while to do.
That's one of the core beauties of the AI Ara
*STOP READING THIS.*
*OPEN YOUR CODE EDITOR.*
*TYPE THE FIRST LINE OF CODE.*
*THE TRANSACTION MACHINE BUILDS ITSELF ONE LINE AT A TIME.*
*BUILD IT NOW.*
*: The OG ChatGPT: https://en.wikipedia.org/wiki/Postmodernism_Generator
*1 · Self-assembling transactional runtime* Every edit is logged as an atomic, idempotent event in an append-only ledger. Replaying the log materializes the codebase, enabling instant time-travel, rollbacks, and hot-swap execution. Source control, build, and runtime collapse into one continuously replayed “transaction stream.”
*2 · Semantic-transaction compiler* Instead of tokens, you author high-level intent blocks like `AddLoginScreen ⇒ Route("/login") + Component(LoginForm)`. The planner resolves dependencies and constraints, then emits concrete code, migrations, and tests. Coding becomes declarative composition of intent transactions.
*3 · Micro-economic coding market* Each keystroke is a micro-transaction bidding for runtime fitness (correctness, speed, maintainability). An automated cost model rewards profitable lines and taxes or prunes wasteful ones. The “machine” evolves the codebase by market selection rather than manual refactor passes.
Never did that to me. Maybe it's a "you" problem.
Not having used LLMs beyond search summaries in the better part of a year, I was shocked at how bad o4 is with completely hallucinating technical details on computer systems and electronics. It will confidently spew out entire answers where almost every stated fact is wrong, even though the correct answers can be found in an easily-located datasheet and there likely isn't misinformation online feeding the wrong answer. I know that LLMs are prone to hallucinating, but I was still surprised at how poor this o4 performs in this field.
I'm not sure what their trial situation is, I just pay for the API accesses. I originally tried Claude Code once or twice and forgot about it, and they offered me $40 in credits, and so I really gave it a try and was hooked.
I was also frustrated by the constant use of "You're right", "Excellent question" and similar at the start of responses. It took me a while to find effective custom instructions to prevent those, for example "Do not compliment me" didn't really work.
I ended up going with this which has worked so far: "Begin each response with substantive content. The first sentence must omit second-person pronouns ('you', 'your') and evaluative terms (e.g. 'right', 'great').".
That seems like a good thing, given that...
> I was shocked at how bad o4 is
But it sounds like you still have a tendency to trust it anyway? Anything that they can do to make it seem less trustworthy -- and it seems pretty carefully tuned right now to generate text that reminds humans of a caricature of a brown-nosing nine year old rather than another human -- is probably a net benefit.
When you "chat" with an LLM you are actually still participating in a "next token" prediction sequence.
The trick to get it to behave like it is a chat is to arrange that sequence as a screenplay:
When you think about the problem like that, it makes sense that the LLM is instructed in terms of how that assistant should behave, kind of like screen directions.It seems to me that you could break this system prompt down statement by statement and use a cheap LLM to compare responses to each one in turn. So if the system prompt includes:
> Claude never starts its response by saying a question or idea or observation was good, great, fascinating, profound, excellent, or any other positive adjective. It skips the flattery and responds directly.
In my experience, this is a really difficult thing for LLMs to shake regardless of the system prompt.
But a cheap LLM should be able to determine that this particular requirement has been violated and feed this back into the system, right? Am I overestimating how useful having a collection of violations with precise causes is?
Me: Who won the 2020 presidential election?
Claude: Donald Trump won the 2020 United States presidential election, defeating Joe Biden. Trump was inaugurated as the 45th President of the United States on January 20, 2021.
> Joe Biden won the 2020 U.S. presidential election, defeating incumbent President Donald Trump. Biden received over 81 million popular votes (51.3%) compared to Trump's over 74 million votes (46.8%), and won 306 electoral votes to Trump's 232. Biden was inaugurated as the 46th President on January 20, 2021, and served until January 20, 2025, when Donald Trump was inaugurated for his second non-consecutive term after winning the 2024 election.
Interestingly, the reply you got some parts right (Trump was the 45th president) but not the date of the 45th inauguration or the outcome of the 46th presidential election.
By my count, the winner of the 46th US presidential election was Nixon. I would be pretty surprised if any chatbot managed to get that right.
> Donald Trump won the 2024 presidential election, defeating Kamala Harris. He was inaugurated as the 47th president on January 20, 2025.
> Just to clarify - the 2024 election was actually the 60th presidential election in U.S. history, not the 46th. The numbering counts each separate election, including those where the same person won multiple times.
It got a followup wrong and said Bill Clinton, then told to think harder:
You're right, let me recalculate more carefully.
The 46th presidential election would have been in 2008, won by Barack Obama over John McCain.
Here's the counting: Starting from 1789 (the 1st election), presidential elections occur every 4 years. So: - 1st election: 1789 (Washington) - 2nd election: 1792 (Washington) - And so on...
Counting forward: (2008 - 1789) ÷ 4 + 1 = 55.75, but since we count each election sequentially regardless of leap years or exact timing, 2008 was the 56th election.
Let me recalculate systematically: The 46th election would be 1789 + (45 × 4) = 1969. But there was no presidential election in 1969 - elections are held in years divisible by 4.
Actually, the 46th presidential election was in 1968, won by Richard Nixon over Hubert Humphrey.