August 18, 2025 • 12 mins
In this episode, we start by welcoming listeners to The OpenAI Daily Brief. Sam Altman addresses criticisms surrounding the GPT-5 launch and the challenges of its adoption. We then discuss Oracle's integration of GPT-5 into its product suite and its partnerships with OpenAI and Google Cloud, followed by analyst insights on Oracle's strategic maneuvers. The episode provides an overview of the capabilities, performance, and safety measures of the ChatGPT agent. We also explore the financial success of the ChatGPT mobile app. Finally, we examine the implications of OpenAI employees selling their stock, wrapping up with a summary of these key topics.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Imagine launching a groundbreaking AI modelonly to face a wave of criticism.
(00:05):
That is the reality OpenAI is dealing withright now.
Welcome to The OpenAI Daily Brief, your go-tofor the latest AI updates.
Today is Monday, August 18, 2025.
Here’s what you need to know about the recentdevelopments surrounding GPT-5's rollout.
(00:26):
Let’s dive in.
Sam Altman, the Chief Executive Officer ofOpenAI, has admitted to some missteps in the
rollout of their latest AI model, GPT-5.
After dealing with multiple delays due tosafety testing and compute limitations, OpenAI
is now addressing user feedback that the modelfelt too formal.
(00:49):
To make interactions more natural, they haveintroduced updates that include subtle
behavioral tweaks, such as adding phrases like"Good question" or "Great start" into
conversations.
These changes are not meant to flatter users,and internal tests show no increase in
sycophancy compared to earlier versions ofGPT-5.
(01:10):
This admission of error comes amid growingcriticism of GPT-5’s launch.
The rollout, which finally happened on August7, was seen by many as offering only
incremental improvements, focusing mainly oncost and speed rather than groundbreaking
capabilities.
Despite this, Sam Altman points out thatadoption has surged.
(01:33):
OpenAI's API traffic doubled in just 48 hoursafter the launch, and the company is currently
out of graphics processing units.
ChatGPT has reached a new high in daily users,with reports indicating that its weekly user
base is nearing 700 million.
Altman has acknowledged that they "totallyscrewed up some things on the rollout," but he
(01:56):
remains optimistic.
He notes that the company has learned valuablelessons about upgrading a product used by
hundreds of millions of people simultaneously.
It is a testament to the challenges andcomplexities of deploying cutting-edge
technology on such a massive scale.
Oracle is shaking things up by deployingOpenAI's GPT-5 across its vast array of
(02:18):
applications and database offerings.
This move not only expands choices for Oracle'scustomers but also provides OpenAI with
additional distribution channels.
Oracle has integrated GPT-5 into its OracleFusion Cloud Applications, Oracle NetSuite, and
industry-specific solutions like Oracle Health.
(02:39):
They're also combining GPT-5 with OracleDatabase 23ai and Oracle AI Vector, making it a
versatile addition to their tech stack.
Now, here’s where it gets interesting (02:49):
OpenAI has been tapping into Oracle Cloud
Infrastructure for its workloads, alongsideGoogle Cloud, as it seeks to diversify beyond
its primary compute base at Microsoft Azure.
For Oracle's Software as a Service andapplication customers, GPT-5 will be available
in three different sizes, adding flexibility tohow businesses can leverage this technology.
(03:14):
Plus, customers have the option to use ChatGPTEnterprise for accessing GPT-5, broadening the
utility of the model.
In a related move, Oracle recently announcedthat Google Cloud's Gemini models and Vertex AI
are also being made available to its customers.
This is part of a broader strategy to offer awide array of AI tools across its platforms,
(03:37):
making it a one-stop-shop for enterpriseslooking to harness AI.
Holger Mueller, an analyst at ConstellationResearch, sees Oracle's addition of GPT-5 as a
logical step, especially since OpenAI isincreasingly utilizing Oracle Cloud
Infrastructure.
Mueller points out that if training occurs onOracle Cloud Infrastructure, OpenAI could
(04:00):
potentially deploy its models on Oracle fasterthan before.
He highlights three key ways Oracle is creatingvalue with OpenAI: training models on Oracle
Cloud Infrastructure, enabling enterprises touse ChatGPT-5 for Oracle Database data via
Vector search and Model Context Protocol, andallowing those same customers to integrate
(04:22):
custom ChatGPT-5 results with their OracleFusion applications.
In another announcement, Oracle is addingOpenAI's open weight models—gpt-oss-120b and
gpt-oss-20b—to its Oracle Cloud InfrastructureData Science's No Code Interface.
This is part of Oracle's ongoing effort to makeAI tools more accessible and user-friendly for
(04:46):
businesses.
OpenAI's latest creation, the ChatGPT agent, ismaking waves by doing more than just chatting.
Imagine having a digital assistant that canactually control your computer to perform tasks
on your behalf.
That is exactly what this new upgrade toOpenAI’s flagship artificial intelligence model
(05:07):
can do.
It is equipped with a virtual computer and anintegrated toolkit, allowing it to handle
complex, multi-step tasks that previousversions of ChatGPT simply could not manage.
Picture this (05:20):
you have a packed schedule, and instead of sifting through your calendar, you
just ask ChatGPT to brief you on your upcomingevents and reminders.
Or perhaps you need a summary of a hefty dataset, and the agent not only analyzes it but
also presents it in a clear synopsis or evencreates a slide deck for you.
This is the kind of functionality that theChatGPT agent brings to the table.
(05:44):
Now, while this all sounds incredibly useful,it is important to note that the ChatGPT agent
is still reliant on human input andsupervision.
It is not fully autonomous.
For example, its spatial reasoning is not quitethere yet, so planning physical routes is a
challenge.
Plus, it does not have true persistent memory,meaning it processes information in the moment
(06:09):
without being able to recall previousinteractions beyond the immediate context.
Despite these limitations, the ChatGPT agentshows significant improvements in OpenAI's
benchmarking.
It more than doubled its accuracy in respondingto expert-level questions on Humanity’s Last
Exam, a benchmark that evaluates a model'sability to answer complex questions across
(06:32):
various disciplines.
And in FrontierMath, the world’s toughest mathbenchmark, it performed much better than
previous models.
So, how does it pull all this off?
The ChatGPT agent is built on three pillars (06:44):
a virtual browser called ‘Operator’ that can
search the web, a deep research capability forsynthesizing large data sets, and the
conversational fluency of previous ChatGPTversions.
It can autonomously browse the web, generatecode, and create files, all under human
(07:05):
supervision.
Professor Kofi Nyarko from Morgan StateUniversity emphasizes that while it is not
autonomous, built-in safeguards like permissionprompts are crucial to minimizing errors.
However, with great power comes greatresponsibility.
OpenAI acknowledges that the increased autonomyof the ChatGPT agent introduces new risks.
(07:29):
It has high biological and chemicalcapabilities, which could potentially assist in
creating chemical or biological weapons ifmisused.
This is a big concern for biosecurity expertswho see this as a "capability escalation
pathway."
To address these risks, OpenAI has implementedseveral safeguards, including threat modeling
(07:49):
and dual-use refusal training, which teachesthe model to decline harmful requests.
However, a risk management assessment bySaferAI scored OpenAI's efforts as weak, at
just 33 out of 100, and the Future of LifeInstitute gave them a C grade on their AI
Safety Index.
Clearly, there's room for improvement, butOpenAI is taking steps to address these
(08:12):
concerns.
The ChatGPT mobile app from OpenAI has hit aremarkable milestone, generating a whopping
$1.35 billion in revenue through July thisyear.
That is a sixfold increase compared to the sameperiod last year when the app brought in $174
million.
This surge in revenue is a testament to thegrowing appetite for AI-driven tools and the
(08:38):
app's expanding global reach.
Now, here's something to think about (08:41):
since its launch in May 2023, the ChatGPT app has seen
explosive growth, with global consumer spendingon iOS and Android platforms reaching $2
billion.
These figures are even more impressive when youconsider they only account for purchases made
through mobile app stores, leaving out websubscriptions and other income sources.
(09:05):
The app's popularity is not just aboutrevenue—downloads have skyrocketed too.
In 2025 alone, it reached 318 milliondownloads, a significant jump from the previous
year.
India leads the way, accounting for 13.7% oftotal downloads, followed closely by the United
(09:26):
States at 10.3%.
Here is where it gets interesting (09:29):
while India dominates in terms of downloads, American users
are the big spenders.
They contribute nearly 40% of the app’s totalrevenue, with an average spend per download of
$10.
Germany is next in line, accounting for 5.3% ofthe revenue.
(09:50):
ChatGPT is not just leading in downloads; italso tops the charts for revenue per
installation among AI chatbot apps.
Competitors like Claude, Grok, and Copilot aretrailing behind, highlighting ChatGPT's
dominant position in the market.
Now, let us talk about a significant financialmove at OpenAI.
(10:10):
Current and former employees are reportedlyplanning to sell six billion dollars worth of
the company's stock.
This is a huge number, and it is catching theattention of big players in the investment
world.
Thrive Capital, SoftBank Group Corporation, andDragoneer Investment Group are among the
investor group that is looking to buy theseshares.
(10:30):
While the talks are still in the early stages,and the size of the share sale could change, it
is already making waves in the businesscommunity.
So, why does this matter?
First, it is a testament to OpenAI's valuationand the confidence investors have in its
potential.
Selling such a large chunk of stock suggeststhat there is significant interest in the
(10:52):
company’s future prospects.
It also reflects the liquidity event foremployees, offering them a chance to cash in on
their hard work and dedication to advancing AItechnology.
The involvement of major investment firms likeThrive Capital and SoftBank Group Corporation
indicates that OpenAI is seen as a valuableasset in the rapidly evolving tech landscape.
(11:15):
These investors are known for backinginnovative companies with strong growth
potential, and their interest in OpenAI stockreinforces the company's standing in the
industry.
While the sale is still being negotiated, thismove could have broader implications for the
technology sector.
It highlights the increasing value placed oncompanies that are at the forefront of AI
(11:36):
innovation.
As OpenAI continues to push boundaries, thefinancial backing it receives through this
stock sale could further fuel its advancementsand initiatives.
That is it for today’s OpenAI Weekly Brief.
OpenAI’s staffers planning to sell six billiondollars in stock underscores the confidence
investors have in the company’s future.
(11:59):
Thanks for tuning in—stay updated for moredevelopments.
This is Bob, signing off.
Until next time.
Imagine launching a groundbreaking AI modelonly to face a wave of criticism.
(00:05):
That is the reality OpenAI is dealing withright now.
Welcome to The OpenAI Daily Brief, your go-tofor the latest AI updates.
Today is Monday, August 18, 2025.
Here’s what you need to know about the recentdevelopments surrounding GPT-5's rollout.
(00:26):
Let’s dive in.
Sam Altman, the Chief Executive Officer ofOpenAI, has admitted to some missteps in the
rollout of their latest AI model, GPT-5.
After dealing with multiple delays due tosafety testing and compute limitations, OpenAI
is now addressing user feedback that the modelfelt too formal.
(00:49):
To make interactions more natural, they haveintroduced updates that include subtle
behavioral tweaks, such as adding phrases like"Good question" or "Great start" into
conversations.
These changes are not meant to flatter users,and internal tests show no increase in
sycophancy compared to earlier versions ofGPT-5.
(01:10):
This admission of error comes amid growingcriticism of GPT-5’s launch.
The rollout, which finally happened on August7, was seen by many as offering only
incremental improvements, focusing mainly oncost and speed rather than groundbreaking
capabilities.
Despite this, Sam Altman points out thatadoption has surged.
(01:33):
OpenAI's API traffic doubled in just 48 hoursafter the launch, and the company is currently
out of graphics processing units.
ChatGPT has reached a new high in daily users,with reports indicating that its weekly user
base is nearing 700 million.
Altman has acknowledged that they "totallyscrewed up some things on the rollout," but he
(01:56):
remains optimistic.
He notes that the company has learned valuablelessons about upgrading a product used by
hundreds of millions of people simultaneously.
It is a testament to the challenges andcomplexities of deploying cutting-edge
technology on such a massive scale.
Oracle is shaking things up by deployingOpenAI's GPT-5 across its vast array of
(02:18):
applications and database offerings.
This move not only expands choices for Oracle'scustomers but also provides OpenAI with
additional distribution channels.
Oracle has integrated GPT-5 into its OracleFusion Cloud Applications, Oracle NetSuite, and
industry-specific solutions like Oracle Health.
(02:39):
They're also combining GPT-5 with OracleDatabase 23ai and Oracle AI Vector, making it a
versatile addition to their tech stack.
Now, here’s where it gets interesting (02:49):
OpenAI has been tapping into Oracle Cloud
Infrastructure for its workloads, alongsideGoogle Cloud, as it seeks to diversify beyond
its primary compute base at Microsoft Azure.
For Oracle's Software as a Service andapplication customers, GPT-5 will be available
in three different sizes, adding flexibility tohow businesses can leverage this technology.
(03:14):
Plus, customers have the option to use ChatGPTEnterprise for accessing GPT-5, broadening the
utility of the model.
In a related move, Oracle recently announcedthat Google Cloud's Gemini models and Vertex AI
are also being made available to its customers.
This is part of a broader strategy to offer awide array of AI tools across its platforms,
(03:37):
making it a one-stop-shop for enterpriseslooking to harness AI.
Holger Mueller, an analyst at ConstellationResearch, sees Oracle's addition of GPT-5 as a
logical step, especially since OpenAI isincreasingly utilizing Oracle Cloud
Infrastructure.
Mueller points out that if training occurs onOracle Cloud Infrastructure, OpenAI could
(04:00):
potentially deploy its models on Oracle fasterthan before.
He highlights three key ways Oracle is creatingvalue with OpenAI: training models on Oracle
Cloud Infrastructure, enabling enterprises touse ChatGPT-5 for Oracle Database data via
Vector search and Model Context Protocol, andallowing those same customers to integrate
(04:22):
custom ChatGPT-5 results with their OracleFusion applications.
In another announcement, Oracle is addingOpenAI's open weight models—gpt-oss-120b and
gpt-oss-20b—to its Oracle Cloud InfrastructureData Science's No Code Interface.
This is part of Oracle's ongoing effort to makeAI tools more accessible and user-friendly for
(04:46):
businesses.
OpenAI's latest creation, the ChatGPT agent, ismaking waves by doing more than just chatting.
Imagine having a digital assistant that canactually control your computer to perform tasks
on your behalf.
That is exactly what this new upgrade toOpenAI’s flagship artificial intelligence model
(05:07):
can do.
It is equipped with a virtual computer and anintegrated toolkit, allowing it to handle
complex, multi-step tasks that previousversions of ChatGPT simply could not manage.
Picture this (05:20):
you have a packed schedule, and instead of sifting through your calendar, you
just ask ChatGPT to brief you on your upcomingevents and reminders.
Or perhaps you need a summary of a hefty dataset, and the agent not only analyzes it but
also presents it in a clear synopsis or evencreates a slide deck for you.
This is the kind of functionality that theChatGPT agent brings to the table.
(05:44):
Now, while this all sounds incredibly useful,it is important to note that the ChatGPT agent
is still reliant on human input andsupervision.
It is not fully autonomous.
For example, its spatial reasoning is not quitethere yet, so planning physical routes is a
challenge.
Plus, it does not have true persistent memory,meaning it processes information in the moment
(06:09):
without being able to recall previousinteractions beyond the immediate context.
Despite these limitations, the ChatGPT agentshows significant improvements in OpenAI's
benchmarking.
It more than doubled its accuracy in respondingto expert-level questions on Humanity’s Last
Exam, a benchmark that evaluates a model'sability to answer complex questions across
(06:32):
various disciplines.
And in FrontierMath, the world’s toughest mathbenchmark, it performed much better than
previous models.
So, how does it pull all this off?
The ChatGPT agent is built on three pillars (06:44):
a virtual browser called ‘Operator’ that can
search the web, a deep research capability forsynthesizing large data sets, and the
conversational fluency of previous ChatGPTversions.
It can autonomously browse the web, generatecode, and create files, all under human
(07:05):
supervision.
Professor Kofi Nyarko from Morgan StateUniversity emphasizes that while it is not
autonomous, built-in safeguards like permissionprompts are crucial to minimizing errors.
However, with great power comes greatresponsibility.
OpenAI acknowledges that the increased autonomyof the ChatGPT agent introduces new risks.
(07:29):
It has high biological and chemicalcapabilities, which could potentially assist in
creating chemical or biological weapons ifmisused.
This is a big concern for biosecurity expertswho see this as a "capability escalation
pathway."
To address these risks, OpenAI has implementedseveral safeguards, including threat modeling
(07:49):
and dual-use refusal training, which teachesthe model to decline harmful requests.
However, a risk management assessment bySaferAI scored OpenAI's efforts as weak, at
just 33 out of 100, and the Future of LifeInstitute gave them a C grade on their AI
Safety Index.
Clearly, there's room for improvement, butOpenAI is taking steps to address these
(08:12):
concerns.
The ChatGPT mobile app from OpenAI has hit aremarkable milestone, generating a whopping
$1.35 billion in revenue through July thisyear.
That is a sixfold increase compared to the sameperiod last year when the app brought in $174
million.
This surge in revenue is a testament to thegrowing appetite for AI-driven tools and the
(08:38):
app's expanding global reach.
Now, here's something to think about (08:41):
since its launch in May 2023, the ChatGPT app has seen
explosive growth, with global consumer spendingon iOS and Android platforms reaching $2
billion.
These figures are even more impressive when youconsider they only account for purchases made
through mobile app stores, leaving out websubscriptions and other income sources.
(09:05):
The app's popularity is not just aboutrevenue—downloads have skyrocketed too.
In 2025 alone, it reached 318 milliondownloads, a significant jump from the previous
year.
India leads the way, accounting for 13.7% oftotal downloads, followed closely by the United
(09:26):
States at 10.3%.
Here is where it gets interesting (09:29):
while India dominates in terms of downloads, American users
are the big spenders.
They contribute nearly 40% of the app’s totalrevenue, with an average spend per download of
$10.
Germany is next in line, accounting for 5.3% ofthe revenue.
(09:50):
ChatGPT is not just leading in downloads; italso tops the charts for revenue per
installation among AI chatbot apps.
Competitors like Claude, Grok, and Copilot aretrailing behind, highlighting ChatGPT's
dominant position in the market.
Now, let us talk about a significant financialmove at OpenAI.
(10:10):
Current and former employees are reportedlyplanning to sell six billion dollars worth of
the company's stock.
This is a huge number, and it is catching theattention of big players in the investment
world.
Thrive Capital, SoftBank Group Corporation, andDragoneer Investment Group are among the
investor group that is looking to buy theseshares.
(10:30):
While the talks are still in the early stages,and the size of the share sale could change, it
is already making waves in the businesscommunity.
So, why does this matter?
First, it is a testament to OpenAI's valuationand the confidence investors have in its
potential.
Selling such a large chunk of stock suggeststhat there is significant interest in the
(10:52):
company’s future prospects.
It also reflects the liquidity event foremployees, offering them a chance to cash in on
their hard work and dedication to advancing AItechnology.
The involvement of major investment firms likeThrive Capital and SoftBank Group Corporation
indicates that OpenAI is seen as a valuableasset in the rapidly evolving tech landscape.
(11:15):
These investors are known for backinginnovative companies with strong growth
potential, and their interest in OpenAI stockreinforces the company's standing in the
industry.
While the sale is still being negotiated, thismove could have broader implications for the
technology sector.
It highlights the increasing value placed oncompanies that are at the forefront of AI
(11:36):
innovation.
As OpenAI continues to push boundaries, thefinancial backing it receives through this
stock sale could further fuel its advancementsand initiatives.
That is it for today’s OpenAI Weekly Brief.
OpenAI’s staffers planning to sell six billiondollars in stock underscores the confidence
investors have in the company’s future.
(11:59):
Thanks for tuning in—stay updated for moredevelopments.
This is Bob, signing off.
Until next time.
Popular Podcasts
NFL Daily with Gregg Rosenthal
Gregg Rosenthal and a rotating crew of elite NFL Media co-hosts, including Patrick Claybon, Colleen Wolfe, Steve Wyche, Nick Shook and Jourdan Rodrigue of The Athletic get you caught up daily on all the NFL news and analysis you need to be smarter and funnier than your friends.
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com