Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="055a9bbc4e37ec59ba15c7ad47b0100274d2f2b65e6ae46082c60c57a382" Subject: I asked ChatGPT about bacon. It gave me jazz hands. From: "Elizabeth Laraki " To: Hidden Recipient Date: Fri, 14 Nov 2025 14:51:56 +0000 X-Hiring: We are hiring, reach out at header-hacker@emailshot.io X-EmailShot-Signature: D6bVe2Ml0rbGOOsAVrL2Bq4qvNgJhkIb2t25p2dyXdMDyJDNvY_FNIq-mYY8Q8sFRFnXJHVGoGvGo6NDe0LbNg== --055a9bbc4e37ec59ba15c7ad47b0100274d2f2b65e6ae46082c60c57a382 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable View this post on the web at https://elizlaraki.substack.com/p/i-asked-chat= gpt-about-bacon-it-gave Yesterday, I asked ChatGPT a simple question. But it decided to jazz up its= answer. I simply asked: What=E2=80=99s the best temperature to cook bacon on the st= ove? I expected a straightforward answer. I intentionally went to ChatGPT becaus= e I didn=E2=80=99t want the classic food-blog experience where you scroll t= hrough ten pages of someone=E2=80=99s childhood memories describing how the= ir grandmother=E2=80=99s house smelled like bacon on Sunday mornings =E2=80= =94 all while dodging ads that take over 90% of the screen. I just wanted t= he temperature. But ChatGPT had other plans. It responded: =E2=80=9CThink of a shy actor stepping onto a warm stage. If the lights are= too hot, it panics and burns. If they=E2=80=99re too cool, it sulks in its= own fat.=E2=80=9D WTF?=20 I stared at the screen, confused, and double-checked I had actually opened = ChatGPT. I had expected a straightforward, factual answer, like normal.=20 Instead, it felt like I requested a Waymo, but got a chatty guy in a yellow= VW Bug blaring psychedelic surf-rock. What=E2=80=99s actually going on Later, I saw an announcement from Fidji Simo (CEO of Applications at OpenAI= ) explaining that ChatGPT now offers personality modes [ https://substack.c= om/redirect/28606205-1d10-4dba-a481-51c6c3754d2e?j=3DeyJ1IjoiM2dmeXZtIn0.xu= 76uFObqArDfP822j-jnN48_jCfgM3m0rbAsF0l24U ], like professional, quirky, ner= dy, friendly, etc. Voice assistants like Siri or Alexa, in-car navigation systems, and even Ch= atGPT voice mode have all offered alternate voices for years. But those wer= e explicit, opt-in choices. Users could pick an alternate voice, but knew e= xactly what they were getting.=20 What felt so jarring here is that ChatGPT has had the same personality for = years. While it=E2=80=99s voice has maybe gotten a bit warmer over time, it= was still clearly machine-like and neutral. But then overnight, it became = a theater kid.=20 If your spouse woke up one morning with a totally different personality, yo= u=E2=80=99d have questions=E2=80=A6 How products change personality =20 But this raises an interesting question: How can you shift a product=E2=80= =99s personality without unsettling the people who use it? Back in 2015, Facebook encountered a version of this. Brand sentiment was d= ropping and many users didn=E2=80=99t like or trust the product. Facebook h= ad spent years shaping itself to be a neutral platform. It had no editorial= voice or emotional tone anywhere. But this neutrality came at a cost =E2= =80=94 the product felt inhuman. So it was easier to blame, distrust, and d= islike. I joined a product team tasked with making Facebook more human and more lik= eable. So we gave Facebook a warmer voice and brought this voice forward to= celebrate happy moments like birthdays, holidays, and memories. We acknowl= edged holidays and joined in celebrations with simple messages like:=20 =E2=80=9CWishing you a happy birthday from all of us at Facebook.=E2=80=9D =E2=80=9CHere=E2=80=99s a memory we hope you enjoy.=E2=80=9D =E2=80=9CLeena and Nabil have birthdays today. Wish them the best!=E2=80=9D= =20 Now, while we experimented with volume, Facebook=E2=80=99s newfound voice d= idn=E2=80=99t show up everywhere. The adjustment wasn=E2=80=99t perfect, bu= t there was no personality setting or dial for users to adjust. The product= =E2=80=99s voice just gently shifted in moments where a human-sounding mess= age felt natural. That subtlety mattered. The product didn=E2=80=99t suddenly sprout a person= ality, but it did start to feel a little less stark and a little more human= =2E=20 Modeling humans Humans shift tone intuitively. We don=E2=80=99t speak the same way when jok= ing with a friend about a wild night out in college vs sharing a cancer dia= gnosis. We each have our own stable personality and we modulate our tone ba= sed on context. =20 Choosing a personality from a drop down is a reasonable first step. But Cha= tGPT=E2=80=99s new personality modes operate like a fixed character, showin= g up the same way across all contexts without understanding social queues.= =20 The real opportunity is to make AI products that feel natural by modeling h= umans =E2=80=94=C2=A0and dynamically shift tone based on the context of the= question and conversation. In this world:=20 Quick factual questions get crisp answers. Emotional conversations get warmth. Creative prompts get big energy. Not because you toggled a setting, but because the AI understood you and th= e moment. For more musings on AI design, design decisions that shaped the products we= use today, and observations of designs around us, subscribe to my free new= sletter. Just for kicks, here are a few snippets of conversations from the day ChatG= PT was all about theatrical answers: Unsubscribe https://substack.com/redirect/2/eyJlIjoiaHR0cHM6Ly9lbGl6bGFyYWt= pLnN1YnN0YWNrLmNvbS9hY3Rpb24vZGlzYWJsZV9lbWFpbD90b2tlbj1leUoxYzJWeVgybGtJam= 95TURrd01UYzBNallzSW5CdmMzUmZhV1FpT2pFM09EZ3hPVFEzTVN3aWFXRjBJam94TnpZek1UT= XhPVEl6TENKbGVIQWlPakUzT1RRMk5qYzVNak1zSW1semN5STZJbkIxWWkweU1EY3lNRGszSWl3= aWMzVmlJam9pWkdsellXSnNaVjlsYldGcGJDSjkuOVhNY21HQnd5MGkwZ1U1elh6Z3U5dnd3bUt= TWjdsYmVfQ3lOeG1NdFNQWSIsInAiOjE3ODgxOTQ3MSwicyI6MjA3MjA5NywiZiI6dHJ1ZSwidS= I6MjA5MDE3NDI2LCJpYXQiOjE3NjMxMzE5MjMsImV4cCI6MjA3ODcwNzkyMywiaXNzIjoicHViL= TAiLCJzdWIiOiJsaW5rLXJlZGlyZWN0In0.VCJc3Gpn5veOibBu8sU5YJzOsBIp132PAWxTXMaz= ffU? --055a9bbc4e37ec59ba15c7ad47b0100274d2f2b65e6ae46082c60c57a382 Content-Type: text/html; charset="utf-8" Content-Transfer-Encoding: quoted-printable I asked ChatGPT about ba= con. It gave me jazz hands.3D""
ChatGPT responded to a simple question with a theatrical answer, ra= ising some interesting questions about how AI products can shift their pers= onality.
͏     ­͏     ­͏ &n= bsp;   ­͏     ­͏     ­= ͏     ­͏     ­͏   ̳= 9; ­͏     ­͏     ­͏ &nb= sp;   ­͏     ­͏     ­&= #847;     ­͏     ­͏    = ; ­͏     ­͏     ­͏ &nbs= p;   ­͏     ­͏     ­&#= 847;     ­͏     ­͏    = ­͏     ­͏     ­͏  = ;   ­͏     ­͏     ­= 47;     ­͏     ­͏     = ­͏     ­͏     ­͏  =   ­͏     ­͏     ­T= 7;     ­͏     ­͏     &= #173;͏     ­͏     ­͏   =   ­͏     ­͏     ­͏= ;     ­͏     ­͏     &#= 173;͏     ­͏     ­͏   &= #8199; ­͏     ­͏     ­͏=     ­͏     ­͏     = 73;͏     ­͏     ­͏   &#= 8199; ­͏     ­͏     ­͏ =     ­͏     ­͏     = 3;͏     ­͏     ­͏   = 199; ­͏     ­͏     ­͏ &= nbsp;   ­͏     ­͏     ­= ;͏     ­͏     ­͏   Q= 99; ­͏     ­͏     ­͏ &n= bsp;   ­͏     ­͏     ­= ͏     ­͏     ­͏   ̳= 9; ­͏     ­͏     ­͏ &nb= sp;   ­͏     ­͏     ­&= #847;     ­͏     ­͏    = ; ­͏     ­͏     ­͏ &nbs= p;   ­͏     ­͏     ­&#= 847;     ­͏     ­͏    = ­͏     ­͏     ­͏  = ;   ­͏     ­͏     ­= 47;     ­͏     ­͏     = ­͏     ­͏     ­͏  =   ­͏     ­͏     ­T= 7;     ­͏     ­͏     &= #173;͏     ­͏     ­͏   =   ­͏     ­͏     ­͏= ;     ­͏     ­͏     &#= 173;͏     ­͏     ­͏   &= #8199; ­͏     ­͏     ­͏=     ­͏     ­͏     = 73;͏     ­͏     ­͏   &#= 8199; ­͏     ­͏     ­͏ =     ­͏     ­͏     = 3;͏     ­͏     ­͏   = 199; ­͏     ­͏     ­͏ &= nbsp;   ­͏     ­͏     ­= ;͏     ­͏     ­͏   Q= 99; ­͏     ­͏     ­͏ &n= bsp;   ­͏     ­͏     ­= ͏     ­͏     ­͏   ̳= 9; ­͏     ­͏     ­͏ &nb= sp;   ­͏     ­͏     ­&= #847;     ­͏     ­͏    = ; ­͏     ­͏     ­͏ &nbs= p;   ­͏     ­͏     ­&#= 847;     ­͏     ­͏    = ­͏     ­͏     ­͏  = ;   ­͏     ­͏     ­= 47;     ­͏     ­͏     = ­͏     ­͏     ­͏  =   ­͏     ­͏     ­T= 7;     ­͏     ­͏     &= #173;͏     ­͏     ­͏   =   ­͏     ­͏     ­͏= ;     ­͏     ­͏     &#= 173;͏     ­͏     ­
<= table role=3D"presentation" width=3D"100%" border=3D"0" cellspacing=3D"0" c= ellpadding=3D"0">
Forwarded this email? Subscribe here for more

I asked ChatGPT about bacon. I= t gave me jazz hands.

ChatGPT responded to a simple question with a the= atrical answer, raising some interesting questions about how AI products ca= n shift their personality.

Elizabeth Laraki
=
=
=
 = =
3D""
=3D""
<= tr>
=3D""
READ IN APP
 <= /td>

Yesterday, I asked ChatGPT a simple quest= ion. But it decided to jazz up its answer.

I simply a= sked: What’s the best temperature to cook bacon on the sto= ve?

I expected a straightforward answer. I intentional= ly went to ChatGPT because I didn’t want the classic food-blog experi= ence where you scroll through ten pages of someone’s childhood memori= es describing how their grandmother’s house smelled like bacon on Sun= day mornings — all while dodging ads that take over 90% of the screen= =2E I just wanted the temperature.

=

But ChatGPT had other plan= s. It responded:

“Think of= a shy actor stepping onto a warm stage. If the lights are too hot, it pani= cs and burns. If they’re too cool, it sulks in its own fat.”

WTF?

I stared at the scree= n, confused, and double-checked I had actually opened ChatGPT. I had expect= ed a straightforward, factual answer, like normal.

Instead= , it felt like I requested a Waymo, but got a chatty guy in a yellow VW Bug= blaring psychedelic surf-rock.

3D""

What’s actually going on

Later, I saw an announcement from Fidji Simo (CEO of Applications at OpenA= I) explaining that ChatGPT now offers personality modes, like profes= sional, quirky, nerdy, friendly, etc.

Voice assistan= ts like Siri or Alexa, in-car navigation systems, and even ChatGPT voice mo= de have all offered alternate voices for years. But those were explicit, op= t-in choices. Users could pick an alternate voice, but knew exactly what th= ey were getting.

What felt so jarring here is that ChatGPT= has had the same personality for years. While it’s voice has maybe g= otten a bit warmer over time, it was still clearly machine-like and neutral= =2E But then overnight, it became a th= eater kid.

If your spo= use woke up one morning with a totally different personality, you’d h= ave questions…

How products change personality

But this raises an interesting question: How can you shift a product’s personality without unsettling the pe= ople who use it?

Back in 2015, Facebook encountere= d a version of this. Brand sentiment was dropping and many users didn’= ;t like or trust the product. Facebook had spent years shaping itself to be= a neutral platform. It had no editorial voice or emotional tone anywhere. = But this neutrality came at a cost — the product felt inhuman. So it = was easier to blame, distrust, and dislike.

I joined a prod= uct team tasked with making Facebook more human and more likeable. So we ga= ve Facebook a warmer voice and brought this voice forward to celebrate happ= y moments like birthdays, holidays, and memories. We acknowledged holidays = and joined in celebrations with simple messages like:

“Wishing you a happy birthday from= all of us at Facebook.”

“Here’s= a memory we hope you enjoy.”

&= #8220;Leena and Nabil have birthdays today. Wish them the best!”=

3D""

= Now, while we experimented with volume, Facebook’s newfound voice did= n’t show up everywhere. The adjustment wasn’t perfect, but ther= e was no personality setting or dial for users to adjust. The product’= ;s voice just gently shifted in moments where a human-sounding message felt= natural.

That subtlety mattered. The product didn’t = suddenly sprout a personality, but it did start to feel a little less stark= and a little more human.

Modeling humans

Humans shift tone intuitively. We don’t speak the same way when = joking with a friend about a wild night out in college vs sharing a cancer = diagnosis. We each have our own stable personality and we modulate our tone= based on context.

Choosing a personality from a drop dow= n is a reasonable first step. But ChatGPT’s new personality modes ope= rate like a fixed character, showing up the same way across all contexts wi= thout understanding social queues.

The real opportunity is= to make AI products that feel natural by modeling humans — and = dynamically shift tone based on the context of the question and conversatio= n. In this world:

  • Quick factual questions get crisp answ= ers.

  • Emotional co= nversations get warmth.

  • Creative prompts get big energy.

Not becau= se you toggled a setting, but because the AI understood you and the moment.=

For more musings on AI design, design decisions= that shaped the products we use today, and observations of designs around = us, subscribe to my free newsletter.

Upgrade to paid

Just for k= icks, here are a few snippets of conversations from the day ChatGPT was all= about theatrical answers:

=

You're currently a free subscr= iber to Elizabeth L= araki. For the full experience, upgrade your= subscription.

Upgrade to paid=

 
=
3D""Like
3D""Comment
<= table role=3D"presentation" width=3D"auto" border=3D"0" cellspacing=3D"0" c= ellpadding=3D"0">
3D""Restack
<= /td>
=
 
© 2025 Elizabeth Laraki
Unsubscribe=

3D"Get3D"Start

3D""3D"" --055a9bbc4e37ec59ba15c7ad47b0100274d2f2b65e6ae46082c60c57a382--