Cassie Kozyrkov
@decisionleader.bsky.social
CEO, AI Advisor, Keynote Speaker, fmr Chief Decision Scientist at Google
Newsletter: decision.substack.com
Newsletter: decision.substack.com
How many of these 9 #AI #headlines did you catch recently? 🧵
🔗 For the curious, my newsletter digs into each one: decision.substack.com
🗞️ 1 Nvidia becomes the world’s first $5 trillion company
🔗 For the curious, my newsletter digs into each one: decision.substack.com
🗞️ 1 Nvidia becomes the world’s first $5 trillion company
November 6, 2025 at 2:55 PM
How many of these 9 #AI #headlines did you catch recently? 🧵
🔗 For the curious, my newsletter digs into each one: decision.substack.com
🗞️ 1 Nvidia becomes the world’s first $5 trillion company
🔗 For the curious, my newsletter digs into each one: decision.substack.com
🗞️ 1 Nvidia becomes the world’s first $5 trillion company
1.2 MILLION of ChatGPT’s weekly users engage in chats that contain suicidal ideation. (Source: OpenAI)
OpenAI tells us:
GPT-5 shows 🟢safe🟢 behaviors in 91% of suicidal conversations.
Which also means:
GPT-5 shows 🔴unsafe🔴 behaviors in 9% of suicidal conversations.
👉 bit.ly/quaesita_therapist
OpenAI tells us:
GPT-5 shows 🟢safe🟢 behaviors in 91% of suicidal conversations.
Which also means:
GPT-5 shows 🔴unsafe🔴 behaviors in 9% of suicidal conversations.
👉 bit.ly/quaesita_therapist
November 5, 2025 at 10:57 PM
1.2 MILLION of ChatGPT’s weekly users engage in chats that contain suicidal ideation. (Source: OpenAI)
OpenAI tells us:
GPT-5 shows 🟢safe🟢 behaviors in 91% of suicidal conversations.
Which also means:
GPT-5 shows 🔴unsafe🔴 behaviors in 9% of suicidal conversations.
👉 bit.ly/quaesita_therapist
OpenAI tells us:
GPT-5 shows 🟢safe🟢 behaviors in 91% of suicidal conversations.
Which also means:
GPT-5 shows 🔴unsafe🔴 behaviors in 9% of suicidal conversations.
👉 bit.ly/quaesita_therapist
🗞️ 12 AWS outage in one region disrupts many major online platforms including Coinbase, Canva, Snapchat, Signal, Fortnite, Lyft, Spotify, Roblox, and even Amazon’s own site. H/T Linas Beliūnas for the image below.
October 23, 2025 at 2:02 PM
🗞️ 12 AWS outage in one region disrupts many major online platforms including Coinbase, Canva, Snapchat, Signal, Fortnite, Lyft, Spotify, Roblox, and even Amazon’s own site. H/T Linas Beliūnas for the image below.
How many of these 14 #AI #headlines did you catch recently? Tell me the one you find most interesting in the comments!
🔗 For the curious, my newsletter digs into each one: decision.substack.com
🗞️ 1 DeepMind’s new Gemma model uncovers potent cancer therapy pathway
🔗 For the curious, my newsletter digs into each one: decision.substack.com
🗞️ 1 DeepMind’s new Gemma model uncovers potent cancer therapy pathway
October 23, 2025 at 2:01 PM
How many of these 14 #AI #headlines did you catch recently? Tell me the one you find most interesting in the comments!
🔗 For the curious, my newsletter digs into each one: decision.substack.com
🗞️ 1 DeepMind’s new Gemma model uncovers potent cancer therapy pathway
🔗 For the curious, my newsletter digs into each one: decision.substack.com
🗞️ 1 DeepMind’s new Gemma model uncovers potent cancer therapy pathway
A great decision maker isn’t someone who works harder—it’s someone who maximizes their return on effort.
👉 Have you signed up for my new course Decision-Making with ChatGPT yet?
Enroll: bit.ly/decisiongptc...
Promo codes and more info in thread...
👉 Have you signed up for my new course Decision-Making with ChatGPT yet?
Enroll: bit.ly/decisiongptc...
Promo codes and more info in thread...
October 22, 2025 at 6:30 PM
A great decision maker isn’t someone who works harder—it’s someone who maximizes their return on effort.
👉 Have you signed up for my new course Decision-Making with ChatGPT yet?
Enroll: bit.ly/decisiongptc...
Promo codes and more info in thread...
👉 Have you signed up for my new course Decision-Making with ChatGPT yet?
Enroll: bit.ly/decisiongptc...
Promo codes and more info in thread...
💡 Leaders can't expect AI-first teams if they aren't AI-first themselves.
Announcing my new course: Decision-Making with ChatGPT!
👉 Enroll: bit.ly/decisiongptc...
Most people expect AI to make decisions for them.
Most leaders think AI is someone else's job.
Wrong and wrong... 🧵
Announcing my new course: Decision-Making with ChatGPT!
👉 Enroll: bit.ly/decisiongptc...
Most people expect AI to make decisions for them.
Most leaders think AI is someone else's job.
Wrong and wrong... 🧵
October 18, 2025 at 2:52 PM
💡 Leaders can't expect AI-first teams if they aren't AI-first themselves.
Announcing my new course: Decision-Making with ChatGPT!
👉 Enroll: bit.ly/decisiongptc...
Most people expect AI to make decisions for them.
Most leaders think AI is someone else's job.
Wrong and wrong... 🧵
Announcing my new course: Decision-Making with ChatGPT!
👉 Enroll: bit.ly/decisiongptc...
Most people expect AI to make decisions for them.
Most leaders think AI is someone else's job.
Wrong and wrong... 🧵
Introducing my new live course:
👉 Decision-Making with ChatGPT: bit.ly/decisiongptc...
Most people use ChatGPT for busywork. I’ll show you how to use it for what actually matters: career moves, negotiations, strategic planning, and high-stakes decisions.
👉 Decision-Making with ChatGPT: bit.ly/decisiongptc...
Most people use ChatGPT for busywork. I’ll show you how to use it for what actually matters: career moves, negotiations, strategic planning, and high-stakes decisions.
October 17, 2025 at 9:30 PM
Introducing my new live course:
👉 Decision-Making with ChatGPT: bit.ly/decisiongptc...
Most people use ChatGPT for busywork. I’ll show you how to use it for what actually matters: career moves, negotiations, strategic planning, and high-stakes decisions.
👉 Decision-Making with ChatGPT: bit.ly/decisiongptc...
Most people use ChatGPT for busywork. I’ll show you how to use it for what actually matters: career moves, negotiations, strategic planning, and high-stakes decisions.
Amid the rise of unsanctioned shadow AI, something new is forming: chimeric talent.
This isn’t the problem of #workslop: the employee who can’t use AI well, who doesn’t verify outputs, and floods systems with polished junk that creates work for everyone else.
This is the opposite problem... 🧵
This isn’t the problem of #workslop: the employee who can’t use AI well, who doesn’t verify outputs, and floods systems with polished junk that creates work for everyone else.
This is the opposite problem... 🧵
October 14, 2025 at 2:10 PM
Amid the rise of unsanctioned shadow AI, something new is forming: chimeric talent.
This isn’t the problem of #workslop: the employee who can’t use AI well, who doesn’t verify outputs, and floods systems with polished junk that creates work for everyone else.
This is the opposite problem... 🧵
This isn’t the problem of #workslop: the employee who can’t use AI well, who doesn’t verify outputs, and floods systems with polished junk that creates work for everyone else.
This is the opposite problem... 🧵
Are you ready for chimeric talent, the workforce AI is quietly creating?
Imagine:
A marketer who codes.
An analyst who writes like a strategist.
An admin who automates workflows.
All powered by generative tools.
These chimeric workers can reconfigure their skillsets on demand.
Imagine:
A marketer who codes.
An analyst who writes like a strategist.
An admin who automates workflows.
All powered by generative tools.
These chimeric workers can reconfigure their skillsets on demand.
October 8, 2025 at 2:34 PM
Are you ready for chimeric talent, the workforce AI is quietly creating?
Imagine:
A marketer who codes.
An analyst who writes like a strategist.
An admin who automates workflows.
All powered by generative tools.
These chimeric workers can reconfigure their skillsets on demand.
Imagine:
A marketer who codes.
An analyst who writes like a strategist.
An admin who automates workflows.
All powered by generative tools.
These chimeric workers can reconfigure their skillsets on demand.
🍏 🍊 Did you know: NVIDIA is worth more than all of Big Pharma. Over the next 10 years, which one will have contributed more (in your opinion) to enabling advances in human health: NVIDIA or Big Pharma?
Source: www.linkedin.com/feed/update/...
Source: www.linkedin.com/feed/update/...
October 6, 2025 at 3:02 PM
🍏 🍊 Did you know: NVIDIA is worth more than all of Big Pharma. Over the next 10 years, which one will have contributed more (in your opinion) to enabling advances in human health: NVIDIA or Big Pharma?
Source: www.linkedin.com/feed/update/...
Source: www.linkedin.com/feed/update/...
How many of these 7 #AI #headlines did you catch recently? Share your score below 👇
For the curious, my newsletter decison.substack.com digs into each one.
1. Claude Sonnet 4.5 is here to work for 30 hours straight on complex coding tasks
2. OpenAI debuts video generator Sora 2
For the curious, my newsletter decison.substack.com digs into each one.
1. Claude Sonnet 4.5 is here to work for 30 hours straight on complex coding tasks
2. OpenAI debuts video generator Sora 2
October 4, 2025 at 7:24 PM
How many of these 7 #AI #headlines did you catch recently? Share your score below 👇
For the curious, my newsletter decison.substack.com digs into each one.
1. Claude Sonnet 4.5 is here to work for 30 hours straight on complex coding tasks
2. OpenAI debuts video generator Sora 2
For the curious, my newsletter decison.substack.com digs into each one.
1. Claude Sonnet 4.5 is here to work for 30 hours straight on complex coding tasks
2. OpenAI debuts video generator Sora 2
I have a new favorite word: workslop!
Workslop is AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task.
In a survey of 1,150 US-based full-time workers, 40% reported having received workslop in the last month. Want more stats? 👇
Workslop is AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task.
In a survey of 1,150 US-based full-time workers, 40% reported having received workslop in the last month. Want more stats? 👇
October 1, 2025 at 12:23 PM
I have a new favorite word: workslop!
Workslop is AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task.
In a survey of 1,150 US-based full-time workers, 40% reported having received workslop in the last month. Want more stats? 👇
Workslop is AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task.
In a survey of 1,150 US-based full-time workers, 40% reported having received workslop in the last month. Want more stats? 👇
15. Google Gemini overtakes ChatGPT as top app after releasing viral Nano Banana image tool
Gemini surged to the # 1 spot on US and UK app charts after its 🍌 Nano Banana feature launch. Of all the Big Three LLMs, it’s by far the leader at generating images that actually look like you. (below)
Gemini surged to the # 1 spot on US and UK app charts after its 🍌 Nano Banana feature launch. Of all the Big Three LLMs, it’s by far the leader at generating images that actually look like you. (below)
September 25, 2025 at 7:17 PM
15. Google Gemini overtakes ChatGPT as top app after releasing viral Nano Banana image tool
Gemini surged to the # 1 spot on US and UK app charts after its 🍌 Nano Banana feature launch. Of all the Big Three LLMs, it’s by far the leader at generating images that actually look like you. (below)
Gemini surged to the # 1 spot on US and UK app charts after its 🍌 Nano Banana feature launch. Of all the Big Three LLMs, it’s by far the leader at generating images that actually look like you. (below)
How many of these 15 #AI #headlines did you catch recently?
12 of them are serious and 3 are goofy - read more about them in my newsletter at decision.substack.com. The illustration to this post was generated with # 15.
12 of them are serious and 3 are goofy - read more about them in my newsletter at decision.substack.com. The illustration to this post was generated with # 15.
September 25, 2025 at 7:17 PM
How many of these 15 #AI #headlines did you catch recently?
12 of them are serious and 3 are goofy - read more about them in my newsletter at decision.substack.com. The illustration to this post was generated with # 15.
12 of them are serious and 3 are goofy - read more about them in my newsletter at decision.substack.com. The illustration to this post was generated with # 15.
Which of the following US States has the highest per-capita ChatGPT use? Florida, California, DC, or Utah?
Hint: it's not California...
Answer in the link, along with a bunch of other info from two studies that analyzed over a million real ChatGPT and Claude conversations:
bit.ly/quaesita_cha...
Hint: it's not California...
Answer in the link, along with a bunch of other info from two studies that analyzed over a million real ChatGPT and Claude conversations:
bit.ly/quaesita_cha...
September 24, 2025 at 1:38 PM
Which of the following US States has the highest per-capita ChatGPT use? Florida, California, DC, or Utah?
Hint: it's not California...
Answer in the link, along with a bunch of other info from two studies that analyzed over a million real ChatGPT and Claude conversations:
bit.ly/quaesita_cha...
Hint: it's not California...
Answer in the link, along with a bunch of other info from two studies that analyzed over a million real ChatGPT and Claude conversations:
bit.ly/quaesita_cha...
Would you take a practical course on decision-making with ChatGPT with me?
I'm exploring interest and would love to know if it's something you might be into.
Let me know here (and get ahead of the waitlist if there's one):
bit.ly/3hdmchatgpt
Please share with anyone who might be interested!
I'm exploring interest and would love to know if it's something you might be into.
Let me know here (and get ahead of the waitlist if there's one):
bit.ly/3hdmchatgpt
Please share with anyone who might be interested!
September 20, 2025 at 3:35 PM
Would you take a practical course on decision-making with ChatGPT with me?
I'm exploring interest and would love to know if it's something you might be into.
Let me know here (and get ahead of the waitlist if there's one):
bit.ly/3hdmchatgpt
Please share with anyone who might be interested!
I'm exploring interest and would love to know if it's something you might be into.
Let me know here (and get ahead of the waitlist if there's one):
bit.ly/3hdmchatgpt
Please share with anyone who might be interested!
Today’s newsletter tackles #genai questions leaders should know:
🤔 What makes managing GenAI harder than managing human creatives?
🤔 How do you separate real value from polished inefficiency?
🤔 Why does alignment crumble when “good” is subjective?
My answers → bit.ly/quaesita_gen...
Hit ♻️
🤔 What makes managing GenAI harder than managing human creatives?
🤔 How do you separate real value from polished inefficiency?
🤔 Why does alignment crumble when “good” is subjective?
My answers → bit.ly/quaesita_gen...
Hit ♻️
September 10, 2025 at 12:44 PM
Today’s newsletter tackles #genai questions leaders should know:
🤔 What makes managing GenAI harder than managing human creatives?
🤔 How do you separate real value from polished inefficiency?
🤔 Why does alignment crumble when “good” is subjective?
My answers → bit.ly/quaesita_gen...
Hit ♻️
🤔 What makes managing GenAI harder than managing human creatives?
🤔 How do you separate real value from polished inefficiency?
🤔 Why does alignment crumble when “good” is subjective?
My answers → bit.ly/quaesita_gen...
Hit ♻️
If you understand the trick, say so.
Just because it feels real doesn’t mean it is.
A storm on a screen doesn’t make it rain in your computer.
Just because it feels real doesn’t mean it is.
A storm on a screen doesn’t make it rain in your computer.
September 6, 2025 at 9:34 PM
If you understand the trick, say so.
Just because it feels real doesn’t mean it is.
A storm on a screen doesn’t make it rain in your computer.
Just because it feels real doesn’t mean it is.
A storm on a screen doesn’t make it rain in your computer.
Examples of AI psychosis:
🚫 Falling in love with your AI
🚫 Believing it’s a god or a friend
🚫 Demanding “AI rights” or “model welfare”
🚫 Falling in love with your AI
🚫 Believing it’s a god or a friend
🚫 Demanding “AI rights” or “model welfare”
September 6, 2025 at 9:34 PM
Examples of AI psychosis:
🚫 Falling in love with your AI
🚫 Believing it’s a god or a friend
🚫 Demanding “AI rights” or “model welfare”
🚫 Falling in love with your AI
🚫 Believing it’s a god or a friend
🚫 Demanding “AI rights” or “model welfare”
What is Seemingly Conscious AI (SCAI)?
An AI system that mimics personhood so convincingly people believe it’s alive.
No new science required. Just today’s large models, prompting, and clever design.
An AI system that mimics personhood so convincingly people believe it’s alive.
No new science required. Just today’s large models, prompting, and clever design.
September 6, 2025 at 9:34 PM
What is Seemingly Conscious AI (SCAI)?
An AI system that mimics personhood so convincingly people believe it’s alive.
No new science required. Just today’s large models, prompting, and clever design.
An AI system that mimics personhood so convincingly people believe it’s alive.
No new science required. Just today’s large models, prompting, and clever design.
We don’t need another polarizing issue.
We don’t need another superstition dressed up as philosophy.
We don’t need people falling for SCAI.
We don’t need another superstition dressed up as philosophy.
We don’t need people falling for SCAI.
September 6, 2025 at 9:34 PM
We don’t need another polarizing issue.
We don’t need another superstition dressed up as philosophy.
We don’t need people falling for SCAI.
We don’t need another superstition dressed up as philosophy.
We don’t need people falling for SCAI.
95% of organizations report no measurable ROI from GenAI.
MIT Media Lab / NANDA’s new report (300+ initiatives, 50+ org interviews, 150+ leaders surveyed) finds:
🧵
MIT Media Lab / NANDA’s new report (300+ initiatives, 50+ org interviews, 150+ leaders surveyed) finds:
🧵
September 5, 2025 at 10:35 PM
95% of organizations report no measurable ROI from GenAI.
MIT Media Lab / NANDA’s new report (300+ initiatives, 50+ org interviews, 150+ leaders surveyed) finds:
🧵
MIT Media Lab / NANDA’s new report (300+ initiatives, 50+ org interviews, 150+ leaders surveyed) finds:
🧵
AI is getting cheaper, faster... yet your costs will still go up.
Per-call prices are falling. But in real deployments, multi-agent workflows plan, search, analyze, and verify -- often looping across tools. That means more calls, more context, more activity. 📈
bit.ly/quaesita_gpt...
Per-call prices are falling. But in real deployments, multi-agent workflows plan, search, analyze, and verify -- often looping across tools. That means more calls, more context, more activity. 📈
bit.ly/quaesita_gpt...
August 21, 2025 at 11:47 AM
AI is getting cheaper, faster... yet your costs will still go up.
Per-call prices are falling. But in real deployments, multi-agent workflows plan, search, analyze, and verify -- often looping across tools. That means more calls, more context, more activity. 📈
bit.ly/quaesita_gpt...
Per-call prices are falling. But in real deployments, multi-agent workflows plan, search, analyze, and verify -- often looping across tools. That means more calls, more context, more activity. 📈
bit.ly/quaesita_gpt...
AI prices are falling, but leaders shouldn't slash their budgets just yet! In this week's newsletter, I explain two big reasons why you'll pay more, not less, for AI: bit.ly/quaesita_gpt...
If you found this useful, a repost ♻️ makes my heart happy. Subscribe to my newsletter: decision.substack.com
If you found this useful, a repost ♻️ makes my heart happy. Subscribe to my newsletter: decision.substack.com
August 20, 2025 at 5:06 PM
AI prices are falling, but leaders shouldn't slash their budgets just yet! In this week's newsletter, I explain two big reasons why you'll pay more, not less, for AI: bit.ly/quaesita_gpt...
If you found this useful, a repost ♻️ makes my heart happy. Subscribe to my newsletter: decision.substack.com
If you found this useful, a repost ♻️ makes my heart happy. Subscribe to my newsletter: decision.substack.com
What's the big idea with GPT-5? It's money. Take a look at the cost table and see where GPT-5 (black rows) lands relative to competitors (lower is cheaper). bit.ly/quaesita_gpt5
August 15, 2025 at 2:56 PM
What's the big idea with GPT-5? It's money. Take a look at the cost table and see where GPT-5 (black rows) lands relative to competitors (lower is cheaper). bit.ly/quaesita_gpt5