davideisinger.com/static/archive/stopsloppypasta-ai-lzdmgr.txt

[1]Skip to main content

Stop Sloppypasta

slop·py·pas·ta  n.  Verbatim LLM output copy-pasted at someone, unread,
unrefined, and unrequested. From slop (low-quality AI-generated content) +
copypasta (text copied and pasted, often as a meme, without critical thought).
It is considered rude because it asks the recipient to do work the sender did
not bother to do themselves.
[2]A few examples [3]Why it's rude [4]Guidelines to do better

You just got an unread message notification.

Maybe it's on Slack (or Teams), a text message, or an email. Maybe you were
tagged in Notion or an Office doc.

You open it to find several paragraphs of text - or perhaps a listicle - [5]
with all the hallmarks of AI-generated writing: headings, heavy formatting, and
"it's not X it's Y", with em-dashes sprinkled gratuitously throughout.

The person who sent it probably spent about ten seconds on it. They asked a
chatbot and forwarded its response to you verbatim without validation or
critical review.

But now you're expected to read, vet, and act on this wall of text. When
someone forwards text they themselves have not considered, they are asking you
to do work they chose not to do. The asymmetric effort makes it rude.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

A few examples

❌ Example
#strategy

The team is mid-discussion about whether to expand into the German market.

👤:

I asked Claude about this! Here's what it said:

Expanding into the German market presents both significant opportunities and
notable challenges. Germany is the largest economy in the EU and offers a
robust consumer base. However, businesses should consider the following key
factors:

1. Regulatory Environment

Germany has strict regulations including GDPR compliance, local labour laws...

[6 more sections follow. None mention your company, your product, your budget,
or reference the discussion already underway.]

buries the live discussion · unrequested · generic

The Eager Beaver

A conversation participant wants to contribute to the topic at hand, so they
ask a chatbot and share whatever comes back. The intention is good - they
genuinely want to help - but the wall of generic AI text they contributed
blocks the discussion already underway. Now other participants have to scroll
past it to continue, or stop to read and validate it.

It feels helpful to send. It creates work to receive.

❌ Example
you:

Does anyone know why our email open rates have been dropping? We changed the
subject line format last month.

👤:

ChatGPT says:

Email open rate declines can be attributed to several factors. These include
changes in subject line strategy, sender reputation issues, list hygiene
problems, and deliverability concerns. Here are the most common causes:

1. Subject Line Fatigue

If subject lines have become too similar or predictable, subscribers may stop
engaging…

[Provides 5 more sections of generic email open diagnostics. Does not mention
your subject line change, your audience, or your platform.]

irrelevant to the specific question · generic

The OrAIcle

Someone asks a specific question. Another person puts it into a chatbot and
pastes the response as the answer.

"ChatGPT says" is the enshittified LLM-era equivalent of [6]LMGTFY (Let Me
Google That For You). Shared as a link or a GIF, LMGTFY was easy to ignore, and
clear about what it was (sarcastic commentary). Sloppypasta is neither.
Recipients are left to figure out whether it's AI generated, whether it's
correct, and which part actually answers the question (if it's actually
relevant at all). If you ask a person a question, you're looking for their
perspective and expertise. In this sense, both LMGTFY and sloppypasta are
etiquette failures where sender disregarded the recipient the dignity of the
basic human reply.

❌ Example
👤:

Hey team - I did some research on our competitors this week. Here's a summary:

Competitive Landscape Overview

The market is highly competitive, with several established players and emerging
challengers. Key competitors offer distinct value propositions across pricing
tiers…

[It's a 5-page essay with handwavy assertions and no concrete details. No
dates. No sources. No live pricing.]

presented as personal work · no one knows to check · hallucinated details
possible

The Ghostwriter

The sender shares AI output as their own work, with no indication a chatbot
wrote it. Recipients have no reason to question it, and may act on information
that is out of date, incomplete, or simply wrong.

Using AI as a ghostwriter borrows the sender's credibility. If the content
turns out to be wrong, that credibility is what gets spent.

Why it's rude

             As a Recipient              As a Sender          Feedback loop
       Previously, effort to read  Writing requires        Sender's skipped
       was balanced by the effort  effort, which           effort becomes
       to write. Now LLMs make     contributes to          recipient's added
Effort writing "free" and increase comprehension. LLMs     effort, increasing
       the effort to read due to   increase cognitive debt frustration as
       additional verification     by reducing struggle.   incidence increases.
       burden.
       LLM propensity for
       hallucination and           What you share directly
       capability to bullshit      influences your         Eroding trust from
Trust  convincingly mean that      reputation. Sharing raw LLM sloppypasta is
       "trust but verify" is       LLM output - especially the modern 'Boy Who
       broken. All correspondence  unvetted - burns your   Cried Wolf.'
       must be untrusted by        credibility.
       default.

Sharing raw AI output is like eating junk food: it's easy and may feel good,
but it's not in your best interest. You'll negatively influence your
relationship with the recipient, and do yourself a disservice by reducing your
own comprehension.

    "For the longest time, writing was more expensive than reading. If you
    encountered a body of written text, you could be sure that at the very
    least, a human spent some time writing it down. The text used to have an
    innate proof-of-thought, a basic token of humanity."

    — Alex Martsinovich, [7]It's rude to show AI output to people

Before LLMs, writing took effort. Authors spent time and effort considering and
selecting their words with intention; time and effort that was balanced by that
spent by the audience as they read. This balance is broken with LLMs; the
effort to produce text is effectively free, but the effort required to read the
text hasn't changed. [8]The increasing verbosity of LLMs further increases the
effort asymmetry. In some circumstances (like pasting raw LLM output into a
chat thread), the sloppypasta effectively becomes a filibuster, crowding out
the existing conversation and blocking the viewport.

    "Cognitive effort — and even getting painfully stuck — is likely important
    for fostering mastery."

    — Anthropic, [9]How AI assistance impacts the formation of coding skills

Writing is thinking. The writing process forces the author to work through
their thoughts, building their comprehension and retention. [10]Multiple [11]
studies have found that delegating tasks to LLMs creates cognitive debt.
Shortcutting thinking with LLMs ultimately reduces comprehension of and recall
about the delegated subject.

    "A polished AI response feels dismissive even if the content is correct"

    — Blake Stockton, [12]AI Writing Etiquette Manifesto

Before LLMs, trust was the default. Authors wrote from their personal expertise
and perspective, and readers could judge an author's understanding of the
subject based on the coherence of their writing. LLMs generate the most
probable next token given an overarching goal to be helpful, which explains
their propensity for hallucination ([13]confabulation) and why many people feel
that [14]LLMs are bullshit generators. Modern LLMs are typically provided tools
to help them look up grounding information that reduces (but does not
eradicate) their likelihood to outright make up facts during their responses.
But that still doesn't solve the trust problem; the reader still has no way to
know what the sender checked and what they didn't. LLM responses, therefore,
cannot be trusted by default and compound the effort asymmetry on the reader by
adding a verification tax.

Beyond accuracy, LLMs write authoritatively with the tone and confidence of an
expert. This adds further uncertainty to the reader's burden; they have no way
to gauge the sender's actual level of expertise with the subject matter. The
result is a further erosion of trust, because the AI's voice removes signal
that recipients previously used to distinguish expertise from
plausible-sounding slop.

    "I think it's rude to publish text that you haven't even read yourself. I
    won't publish anything that will take someone longer to read than it took
    me to write."

    — Simon Willison, [15]Personal AI Ethics

Formerly, "Trust but verify" ruled. Readers would trust until that trust was
broken; the author was trustworthy or they weren't. However, shared LLM output
obfuscates the chain of trust. Did the prompter do the appropriate due
diligence to validate the LLM response? If problems or errors are discovered,
who is to blame, the prompter or the AI? Was it an oversight, a missed
verification step, or was verification skipped altogether? The uncertainty
means the recipient doesn't know what they can trust, what has or has not been
verified; they must treat everything as untrusted. Just like the Boy Who Cried
Wolf, once the trust is broken, the uncertainty spreads to all future messages
from the sender.

Assumptions of balanced effort and presumed trust are no longer guaranteed in a
post-LLM world. Sloppypasta creates a compounding negative feedback loop where
the sender forfeits learning and credibility while the recipient burns effort
and loses trust. Receiving raw AI output feels bad due to the cognitive
dissonance of having these assumptions violated.

Read the full essay

Simple guidelines to do better

Read.

Read the output before you share it. If you haven't read it, you don't know
whether it's correct, relevant, or current.

Delegating work to AI creates cognitive debt. Working with the results helps
run damage control for your own understanding.

Verify.

Check the facts before you forward them. Anything you forward carries your
implicit endorsement -- your reputation depends on managing the quality of what
you share.

LLMs are trained to "be helpful", and will produce outdated facts, wrong
figures, and plausible nonsense to provide a response to your requests.
Further, an LLM is inherently out-of-date; their knowledge cutoffs contain at
best information on the state of the world when their training started (months
ago).

Distill.

Cut the response down to what matters. Distilling the generated response to the
useful essence is your job.

LLMs are incentivized to use many words when few would do: API-priced models
have a per-token incentive to train chatty LLMs that use many tokens, and [17]
research shows that longer, highly formatted posts are often preferred as more
engaging.

Disclose.

Share how AI helped.

If you've read, verified, and edited it, send it as yours -- preferably with a
note that you worked with AI assistance. If you're sharing raw output, say so
explicitly. In both cases, it may be useful to share your prompt and how you
worked with the AI to get the final output.

Disclosure restores the trust signals that sloppypasta destroys and tells the
recipient what you checked and what they may be on the hook for.

Share only when requested.

Never share unsolicited AI output into a conversation.

Remember that AI generations create effort asymmetry and be respectful of those
you share with. Sloppypasta delegates the full burden of reading, verifying,
and distilling to a recipient who didn't ask for it and may not realize the
effort required of them.

Share as a link.

Share AI output as a link or attached document rather than dropping the full
text inline.

In messaging environments, a large paste takes over the viewport and crowds out
the existing conversation. A link lets the recipient choose when - and whether
- to engage, rather than having that choice imposed on them.

AI capabilities keep increasing, and using it to draft, brainstorm or
accelerate you will be increasingly useful. However, using AI should not make
your productivity someone else's burden. New tools require new manners.

Use AI to accelerate your work or improve what you send.
Don't use it to replace thinking about what you're sending.

Further reading

  • [18]It's Rude to Show AI Output to People
  • [19]Personal AI Ethics by Simon Willison
  • [20]AI Manifesto
  • [21]Using AI Responsibly in Development & Collaboration
  • [22]AI Writing Etiquette Manifesto

inspired by [23]nohello.net · [24]dontasktoask.com [25]open source

References:

[1] https://stopsloppypasta.ai/en/#main-content
[2] https://stopsloppypasta.ai/en/#types
[3] https://stopsloppypasta.ai/en/#why
[4] https://stopsloppypasta.ai/en/#rules
[5] https://tropes.fyi/directory
[6] https://lmgtfy.app/?q=what+is+lmgtfy
[7] https://distantprovince.by/posts/its-rude-to-show-ai-output-to-people/
[8] https://epoch.ai/data-insights/output-length
[9] https://www.anthropic.com/research/AI-assistance-coding-skills
[10] https://www.media.mit.edu/publications/your-brain-on-chatgpt/
[11] https://www.anthropic.com/research/AI-assistance-coding-skills
[12] https://www.blakestockton.com/ai-writing-etiquette-manifesto/
[13] https://pmc.ncbi.nlm.nih.gov/articles/PMC10619792/
[14] https://machine-bullshit.github.io/
[15] https://simonwillison.net/2023/Aug/27/wordcamp-llms/#personal-ai-ethics
[17] https://arxiv.org/abs/2310.10076
[18] https://distantprovince.by/posts/its-rude-to-show-ai-output-to-people/
[19] https://simonwillison.net/2023/Aug/27/wordcamp-llms/#personal-ai-ethics
[20] https://noellevandijk.com/ai-manifesto/
[21] https://ai-manifesto.dev/
[22] https://www.blakestockton.com/ai-writing-etiquette-manifesto/
[23] https://nohello.net/
[24] https://dontasktoask.com/
[25] https://github.com/ahgraber/stopsloppypasta