Loading...
Loading...

The AI rumor mill is heating up again, with fresh speculation about when OpenAI’s next major model might land and whether it could reset the competitive narrative after a turbulent few months of launches. The episode puts those rumors in context, tracing how momentum has shifted between OpenAI, Google, and Anthropic, and what a real step-change model would need to accomplish. In the headlines: Microsoft moves to defuse political backlash over data center power costs, chip geopolitics tighten around NVIDIA and China, OpenAI makes a small but telling health-tech acquisition, and new model rumors swirl from DeepSeek, Google, and Anthropic.
Brought to you by:
KPMG – Discover how AI is transforming possibility into reality. Tune into the new KPMG 'You Can with AI' podcast and unlock insights that will inform smarter decisions inside your enterprise. Listen now and start shaping your future with every episode. https://www.kpmg.us/AIpodcasts
Zencoder - From vibe coding to AI-first engineering - http://zencoder.ai/zenflow
Optimizely Opal - The agent orchestration platform build for marketers - https://www.optimizely.com/theaidailybrief
AssemblyAI - The best way to build Voice AI apps - https://www.assemblyai.com/brief
LandfallIP - AI to Navigate the Patent Process - https://landfallip.com/
Robots & Pencils - Cloud-native AI solutions that power results https://robotsandpencils.com/
The Agent Readiness Audit from Superintelligent - Go to https://besuper.ai/ to request your company's agent readiness score.
The AI Daily Brief helps you understand the most important news and discussions in AI. Subscribe to the podcast version of The AI Daily Brief wherever you listen: https://pod.link/1680633614
Interested in sponsoring the show? [email protected]
Today on the AI Daily Brief, what the AI rumor mill says about when we might get the next big
model, and before that in the headlines, Microsoft coming in hot with some plans to make people
less mad about data centers and electricity. The AI Daily Brief is a daily podcast and video
about the most important news and discussions in AI. All right friends, quick announcement says
always before we dig in. Firstly, thanks to today's sponsors KPMG, robots and pencils, zen
coder and super intelligent to get an ad free version of the show go to patreon.com such
AI Daily Brief or you can subscribe on Apple podcasts to learn about sponsoring the show or really
anything else regarding the show or frankly anything else that I'm doing, go check out AI Daily Brief.
AI. You can find links to our new operators community, our new years project, and if you press that
AI DB intel button you can even get a little sneak peek of whatever the heck this cool looking
maturity map thing is. Again, AI Daily Brief. AI for everything for the show and beyond,
but for now let's dive into this episode. Welcome back to the AI Daily Brief headlines edition,
all the daily AI news you need in around five minutes. 2026 is an election year and it's been
clear for some time that AI was going to find its way into the political discourse. The odds
on bet for how it does make it there is less about AI itself, although there's plenty of issues
that people have, and more about the broader theme which is very clearly going to dominate this
election cycle which is affordability. In short, to the extent that data centers are perceived
to be a contributor to higher costs of living for Americans, those data centers and the larger
AI industry are going to have a not so fun time politically. Indeed as he goes after a number of
different affordability issues, Donald Trump has turned his attention and his truth social account
on this particular one as well. On Monday he wrote, I never want Americans to pay higher electricity
bills because of data centers. Therefore my administration is working with major American
technology companies to secure their commitment to the American people and we will have much to
announce in the coming weeks. First up is Microsoft who my team has been working with and which will
make major changes beginning this week to ensure that Americans don't pick up the tab for their
power consumption in the form of paying higher utility bills. We are the hottest country in the
world at number one in AI. Data centers are key to that boom and keeping Americans free and secure,
but the big technology companies who build them must pay their own way. Thank you and congratulations
to Microsoft more to come soon. Now it is way beyond the scope of this headlines episode to get
into the full complexity of why electricity costs are up and what percentage of it is actually from AI,
but frankly I think all of those are completely losing political arguments and all that matters
is basically exactly what President Trump is getting at here which is the perception of whether
the big companies are not only picking up the tab for themselves but perhaps even paying a little
bit more to try to make this viable for everyone else. Now people have been talking about this type
of policy for a while. Investor Chimath Palahapitiya started tweeting about it somewhere in the
middle of last year and kept it up throughout the fall. For example, in October writing,
the hyperscaler should take the electricity cost of local residents to zero and start buying good will.
Otherwise, I expect more local communities to push back on these data centers which will complicate
the AI buildout that needs to happen. So what did Microsoft actually announce? In a blog post on
Tuesday from Vice Chair and President Brad Smith, the company wrote about a five-part plan to build
with their calling community first AI infrastructure. They write that the plan commits them to
concrete steps needed to be a good neighbor in the communities where they build own and operate
their data centers. So what are the five parts of their plan? The first is that they'll pay their
own way to ensure their data centers don't increase other people's electricity prices. Basically,
they say they're going to pay utility rates that are high enough to cover their electricity costs
and make sure it doesn't get passed on to the communities in which they're operating. Pillar 2 is
they commit to trying to minimize their water use and replenish even more of the communities water
than they use. Pillar 3 is to create jobs for residents. Pillar 4 is to add to the tax base to
fund local hospital schools, parks and libraries. And Pillar 5 is to strengthen the community by
investing in local AI training and nonprofits. Now it's totally easy to be cynical about any
corporate initiative like this. But for my money, this is exactly the type of thing that needs to
happen from all of the big tech companies who are in the midst of this infrastructure buildout.
Frankly, I think it's a complete own goal that with something like this, where there is so much
opportunity for these data centers to actually be good for the communities that they're in,
that we have completely missed that vote until now. I'm glad to see Microsoft taking this on,
and frankly, I think they can go even farther. I think Shamaat is right. I think they should be going
way beyond just paying their own share, and frankly, just buying the goodwill of the community that
they're in. Ultimately, that is such a small fraction of the cost of these data centers.
The doing it to me just seems like an o-brainer. Still, this is good progress, and I want to
encourage Microsoft and everyone else in a similar space to double down on this type of initiative.
Now, moving over to a story that has been up and down and over and under and never quite
clear, on Tuesday, Reuters reported that Chinese customs officials have told Customs Agents
that in videos, H-200 chips are not permitted to enter the country. Their sources said that tech
companies were also summoned to meetings where they were explicitly told not to order chips
unless necessary. One of the Reuters sources commented,
the wording from the officials is so severe that it is basically a ban for now, though this might
change in the future should things evolve. Now, the information has a slightly different sourcing
on the story, who said that the directive from Beijing was, quote, deliberately vague.
They said that the imports were limited to special circumstances, which included university
research and R&D. Both reports used the word necessary to describe the limitations,
but the difference was in how each source interpreted the CCP directive.
Later that day, the US Commerce Department finalized their approval for H-200 exports,
but also with a few conditions. The chips will be inspected by a third-party testing lab to
confirm their AI capabilities before they can be shipped China. In video, it is also limited to
shipping 50% as many chips to China as they sell to US customers. On the Chinese side of the deal,
customers will need to demonstrate, quote, unquote, sufficient security procedures,
and cannot use the chips for military purposes. In a statement in video said that the approval
quote strikes a thoughtful balance that is great for America. And yet, while all that paperwork
is finalized, it's unclear if Nvidia can actually start shipping anytime soon due to the Beijing
Bands. Some China analysts do believe this is a power play in the lead up to trade negotiations
in April. Geopolitical strategist Reva Goshen writes, Beijing is pushing to see what bigger
concessions they can get to dismantle US-led tech controls. Chris McGuire, a senior fellow at the
Council on Foreign Relations commented, Beijing believes the US is desperate to sell AI chips to China,
so it believes China has the leverage to extract concessions from the US in exchange for license
approvals. Now, it's an open question whether the Trump administration is desperate to sell AI chips,
but the potential for an Nvidia lead stock market drawdown during an election year could be a motivating
factor. Staying on the chip train, chip-making startups to re-wrest is in talks to raise a billion
dollars at a $22 billion valuation. Bloomberg sources confirmed that fundraising efforts were
underway, but added no major details. The company was aiming to IPO last year, but scuttled plans
in October shortly after completing a fundraising round at an $8 billion valuation. Sources said
the company still plans to IPO, with rumors suggesting the aim is to go public in the second half
of this year. In M&A land, OpenAI has acquired a tiny health tech startup called Torch. The company
operates a platform to unify medical records, including lab results, prescriptions, and appointment
notes, while storing them in a format that's easily discoverable for AI. CoFounder Ilya Abbasov wrote,
we designed Torch to be a unified medical memory for AI, bringing every bit of data about you from
hospitals, labs, wearables, and consumer testing companies into one place. I can't imagine a
better next chapter than to now get to put our technology and ideas in the hands of the hundreds
of millions of people who already use chat GPT for health questions every week. Now, OpenAI
didn't announce the value of the acquisition, but sources speaking with the information said the
price tag was $100 million paid in OpenAI equity. Not bad for a four-person team.
Lots cooking is always in the world of AI, but for now, that is going to do it for the headlines.
Next up, the main episode.
Hello, friends. If you've been enjoying what we've been discussing on the show,
you'll want to check out another podcast that I've had the privilege to host, which is called
You Can With AI from KPMG. Season one was designed to be a set of real stories from real leaders
making AI work in their organizations, and now season two is coming and we're back with even
bigger conversations. This show is entirely focused on what it's like to actually drive AI change
inside your enterprise, and as case studies, expert panels, and a lot more practical goodness that
I hope will be extremely valuable for you as the listener. Search You Can With AI on Apple Spotify
or YouTube and subscribe today. Today's episode is brought to you by robots and pencils,
a company that is growing fast. Their work as a high-growth AWS and Databricks partner
means that they're looking for elite talent ready to create real impact at velocity.
Their teams are made up of AI native engineers, strategists, and designers who love solving hard
problems and pushing how AI shows up in real products. They move quickly using RoboWorks,
their agentic acceleration platform, so teams can deliver meaningful outcomes in weeks,
not months. They don't build big teams. They build high-impact nimble ones.
The people there are wicked smart with patents, published research, and work that's helped shape
entire categories. They work in velocity pods and studios that stay focused and move with intent.
If you're ready for career defining work with peers who challenge you and have your back,
robots and pencils is the place. Explore open roles at robotsandpensils.com slash careers.
That's robotsandpensils.com slash careers.
If you're using AI to code, ask yourself, are you building software or are you just playing
prompt roulette? We know that unstructured prompting works at first, but eventually it leads to AI
slop and technical debt. Enter Zenflow. Zenflow takes you from vibe coding to AI first engineering.
It's the first AI orchestration layer that brings discipline to the chaos. It transforms
freeform prompting into spec driven workflows and multi-agent verification, where agents
actually cross-check each other to prevent drift. You can even command a fleet of parallel
agents to implement features and fix bugs simultaneously. We've seen teams accelerate delivery to
X to 10X. Stop gambling with prompts. Start orchestrating your AI. Turn raw speed into reliable
production grade output at Zenflow.free. Today's episode is brought to you by Super Intelligent.
Super Intelligent is a platform that very simply put is all about helping your company figure out
how to use AI better. We deploy voice agents to interview people across your company,
combine that with proprietary intelligence about what's working for other companies,
and give you a set of recommendations around use cases, change management initiatives,
that add up to an AI roadmap that can help you get value out of AI for your company.
But now we want to empower the folks inside your team who are responsible for that transformation
with an even more direct platform. Our forthcoming AI strategy compass tool is ready to start to be
tested. This is a power tool for anyone who is responsible for AI adoption or AI transformation
inside their companies. It's going to allow you to do a lot of the things that we do at Super
Intelligent, but in a much more automated, self-managed way, and with a totally different
cost structure. If you are interested in checking it out, go to AIDailyBrief.ai-slash-compass,
fill out the form and we will be in touch soon.
Welcome back to the AIDailyBrief. Today we are talking about the latest leak-slash rumors about
the next ChatcheebT model, but I think it's important to put all of this in its proper context.
Let's do a quick hit of the last five months of OpenAI model releases.
Things started pretty inauspiciously in August with the release of GPT-5.
Now we've got over lots of times all of the problems with the GPT-5 release. One very big problem
was the deprecation of 40 alongside it, which had people anger at them for reasons that had
nothing to do with the model's performance and everything to do with other changes that were
being made at the same time. We've also discussed how if they had simply called their biggest
reasoning models like O3, GPT-5, the perception of the performance jump might have been very
different. Basically in some ways they were kind of a victim of their own making. Whatever the case,
the context that came into was not a great moment for the narrative around AI, and GPT-5 did
nothing to alleviate that. That's when we were getting these in retrospect very silly op-eds
in publications like The New Yorker. What if AI doesn't get much better than this? Fast forward
a couple months and the pressure was on for Google to deliver. There was a while there where I
wasn't even sure that Google was actually going to drop Gemini 3 in November because of the
amount of pressure they were under to get it right. But get it right they did, at least in the
court of public opinion. When Gemini 3 came out, people were extremely excited about it. They were
impressed with Gemini 3 Pro as a model for their intellectual and work tasks, and of course,
Nanobin and a Pro's ability to make infographics opened up all sorts of totally new possible
use cases. It turns out that OpenAI knew they were in for a rough patch. Back in October,
it turns out Sam Altman had warned some staff in a memo that he expected some rough vibes around
the launch of Google's new models. Rough vibes they got ultimately leading to Altman and the
team at OpenAI declaring a code red. Now what this code red meant, in short, was a cessation or at
least a slowdown of work on a lot of ancillary features and products to double triple quadruple
down on core chatchypt features including the models underneath powering it. That got us to GPT-5.2
as well as the new chatchypt images model, which is it should be noted a 1.5 model not a full jump
to image gen 2. Now 5.2 and the new chatchypt images are good models. 5.2 Pro in particular is very
much in my regular rotation, and when it comes to a lot of heavy intellectual work, there are many
folks who swear by it. Images, frankly, was better than I expected, given how much pressure they
had to put that out given Nanobin and a Pro. And so even though Gemini and Google had really
won a ton of momentum, I do think that the chatchypt releases in December maybe didn't fully
stem the bleeding, but for people who weren't interested in the horse race and just wanted high
performing models, you felt very lucky with all the options you had over the holiday season.
But then of course around all of this was Claude Opus 4.5. The opinion on this model has done
nothing but go up and up and up and up. So much so that in the last minute upset, I actually
said that I thought it might end up being the most important model release of 2025, and so far
at least I think that argument is holding up. Claude code Opus 4.5 and AGI are terms that are
very frequent co-inhabitants right now of tweets and posts on social networks. Since the beginning of
the year, these companies have not slowed down. In a major move to bring Claude code to everybody
else, Anthropic released co-work, tripling down in that way on their source of narrative momentum,
while Google and Apple announced the deal that was reported at the end of last year,
that forthcoming versions of Apple Intelligence will be powered in fact by Gemini models.
All of this has led to a sense of a lot of momentum around Anthropic and Google,
and kind of less so around OpenAI. In fact, last night I tweeted that OpenAI seemed to me to be
almost conspicuously quiet, which perhaps isn't totally fair given that they announced a major
product in chat GPT health, but still it feels to me both like something is percolating,
and also that perhaps the company decided to try to do a little bit less vague posting in this
new 2026 year. Yesterday the rumor mill kicked back up in a big way. Dan Mac tweeted,
GBT-5-3 codenamed garlic coming soon according to a source, a very reliable source,
batting a thousand, expected to be a doozy, likely with stronger pre-training and the IMO
gold-winning reasoning techniques. The AI leaker account I ruled the world responded saying that
they had also heard this month, and in a separate post shared something that they had told
subscribers back in December that the 5-2 model that we got was a quote rush-terally checkpoint
and that the full model is going to drop in January. Other speculation is that the model will be
multimodal generating both images and audio, although no one seems quite clear on the naming
conventions, whether it will be GBT-5-5 or even something like GBT-5-3. And of course while
these are all just rumors, although rumors with sourcing, there is also starting to be some evidence
that some newness might be percolating and poking through. Andrew Current tweeted,
My chat is acting quite differently as of last night. I assumed it was part of a new personality
test group. This has happened many times over the last couple of months. OpenAI did say the next
big update would make 5 more personable. It's possible it's about to arrive. Now maybe I am wrong
about the vague tweeting, given that VB from OpenAI's developer experience team,
posted the eyes emoji getting everyone talking as part of this conversation as well.
Obviously, any time we get a new model, it's a very exciting moment. And certainly with the stakes
as high as they are, I would love to see a big powerful effort drop that shakes the race up once
again. Now believe it or not, that wasn't the only information in the AI rumor mill. A Chinese
consumer electronics blogger has a new leak from their contacts and the supply chain. They wrote,
hearing fresh detail on OpenAI 2Go hardware project from last report, now confirmed it's a
special audio product to replace AirPods. Internal code name is SweetPie. A manufacturing Foxconn
has been told to prepare for a total of five devices by Q42028. All are not known, but a
home-style device and pen are still considered. However, many sources repeated the same thing.
SweetPie is now front of the line due to the priority of the Johnny Ive team.
The release has been told to be near September, volume projection 40-50 million in the first year.
Only some details currently known. Hardware design is said to be unique, unseen before,
and the main device is to be metal in resembling the shape of an eggstone. Inside the eggstone,
there are two pills who are removed and rest behind the ear. A custom chip is being developed to
allow the device to replace iPhone actions by commanding Siri. And overall, Foxconn leaders are still
embarrassed by losing all AirPods programs to Lux. Now they see this as a golden chance to win back
the category. We had a show last year about how sneakily AirPods could be the most obvious AI
device form factor that not enough people were thinking about in that way. And so it's interesting
to see the OpenAI hardware team seemingly exploring some similar space, although of course
something behind the ear is something totally different once again. Now moving away from OpenAI,
but staying in the new model pool, another report of a likely model to drop very soon is the
next flagship DeepSeek model. The information reports that DeepSeek V4 will be released in mid-February
and will have a heavy focus on coding performance. Rides the information, the new model V4 is the
successor to the V3 model DeepSeek released in December 2024. Initial tests done by DeepSeek
employees based on the company's internal benchmarks showed that it outperformed existing model,
such as Anthropics Cloud and OpenAI's GPT series, encoding the sources said. The report also
states that V4 will showcase DeepSeek's advances in handling extremely long context windows,
which of course is critical for large coding tasks. Now obviously it would be quite the shake-up
to have a state-of-the-art open source coding model, but of course we don't actually know how it
will perform until we see it. Summer convinced though that a model codenamed Beluga on LM Arena is our
first look at the whale's next release. DeepSeek fan tier taxes is skeptical of the sourcing but
still believes the hype posting. I dunk on breathless insider leaks about V4, but nobody is more
confident than me in what DeepSeek is about to achieve. Personally in my opinion it'll be the
end of the road for Dario-style fantasies. Developer Vasio posted, feels like we're about to get
another overnight jump that was clearly years in the making. DeepSeek was also in the western
press recently, for a different reason, with the founder of DeepSeek's quantitative hedge fund
generating returns of 57% last year. Reds Bloomberg's Joe Weisenthal,
man this dude's had a good couple of years. Now if those are the forthcoming model rumors,
we also got a small but real update in VO31 ingredients to video. The feature allows users
to upload reference images for characters, props, and background to guide the video generation.
Google said the upgrade makes videos more expressive and creative even with simple prompts.
They also promised better visual consistency across multiple scenes to ensure clips can be
easily stitched together to tell a coherent story. VO can also now use ingredients to video
when generating vertical videos for mobile which wasn't previously possible. And in a last
bit of lab news which isn't a model right now but could be an interesting product in the future,
Anthropic has announced the expansion of their labs team into a full blown internal incubator.
Anthropic labs was started in mid-2024 with just two members and helped develop
Cloud Code, MCP, and more recently co-work. Labs will now become a more substantial part of the
company, with Anthropic gaming to double the lab's headcount within the next six months.
The expanded team will be co-led by Chief Product Officer Mike Krieger and Product
Engineering Lead Ben Mann. The team will report to Anthropic President Daniela Amode.
Now if you want to hear more about how Mike thinks about building AI products,
I did an interview with him as part of our end of year episodes which you can find on YouTube
or on this podcast feed. Ultimately it sounds like a big part of the move is about restructuring
the team so that Anthropic's product velocity can match the rapid pace of the industry.
Wrote Daniela, the speed of advancement in AI demands a different approach in how we build,
how we organize, and where we focus. Labs gives us room to break the mold and explore.
Look, if it leads to more products like Cloud Code, I think most folks in the industry will say
count us in. For now, that's the latest on what's cooking around the AI rumor mill.
Hopefully we get some real models in the next couple of weeks to explore.
For now, though, I appreciate you listening or watching as always and until next time, peace!
The AI Daily Brief: Artificial Intelligence News and Analysis
