Intro
The Limitless hosts run their weekly AI roundup, working through Nvidia's record quarter and its Hugging Face acquisition, the mystery model that turned out to be Z.ai's GLM 5.3 Flash, Waymo's new on-board chip and Tesla's CyberCab launch. The back half covers the SEC probe into Leopold Aschenbrenner, a Financial Times chart on Anthropic model usage, a robotics model that learns from a single video, the September 9 Apple event and Sam Altman's custom Swiss watches.
Hosts: Ejaaz Ahamadeen and Josh Kale
Published: 28 August 2026 on Limitless: An AI Podcast
Episode page | 41 min
Key Takeaways
Nvidia's quarter is bigger than most of the S&P 500 is in a year
"Nvidia's quarterly revenue now exceeds the annual revenue of roughly 480 of the S&P 500." — Josh Kale
Guidance moved to 70% year-over-year growth in 2028, against expectations of 44%
Jensen Huang's answer to custom silicon is that he owns the general-purpose moat
Nvidia still controls 70% to 85% of the market on Ejaaz Ahamadeen's numbers
OpenAI's Jalapeno is an ASIC that has not been seen running at scale yet
The $13 billion Hugging Face deal buys the layer every open model already sits on
Open weights are where the token volume is going, because they are cheaper and good enough
Hugging Face's inference service could become the cloud business Nvidia never had
AI cloud runs at 40% to 50% margins against roughly 80% for ordinary cloud, because Nvidia takes the rest
A Chinese lab served 100 trillion free tokens entirely on Chinese silicon
Z.ai's GLM 5.3 Flash benchmarks at Opus-tier agentic level at 15 cents per million tokens
Chinese chip companies were two and a half years behind six months ago
Waymo's new chip is 20 times more powerful and 5 times cheaper, and it all fits in the trunk
The sensor and compute suite drops from six figures to five, with the car no longer depending on a cloud link
Tesla's CyberCab goes into service on September 3 across six cities
Eight smartphone-grade cameras instead of a spinning LiDAR stack, at a lower cost per mile
The SEC has subpoenaed banks that facilitated Situational Awareness Fund's trades
Josh Kale's read: opening an investigation takes a complaint, and this one is probably nothing
A robotics model learned tasks in context from a single 30-second video
It corrects the mistake in the demonstration rather than copying it, and improvises when the tool is swapped for a banana
Apple's hardware event is September 9, and it is the first one Tim Cook is not running
John Ternus takes the stage; Siri ships after the hardware, not at it
Local inference is getting more expensive, and memory pricing is the reason
Nvidia raised GPU prices 15% because memory costs soared
Nvidia's Quarter Is Bigger Than Most of the S&P 500 Manages in a Year
Ejaaz Ahamadeen opened the episode with the scale of the quarter: almost $100 billion earned in three months, with the company growing 100% year over year and worth over $5 trillion.
Josh Kale's framing was the comparison: "Nvidia's quarterly revenue now exceeds the annual revenue of roughly 480 of the S&P 500."
On his arithmetic, Nvidia is printing just over a billion dollars in cash every day it operates, and the number is going up.
The guidance was the part that moved the stock: 70% year-over-year growth in 2028, against expectations of 44%.
The stock was down after hours and then up 7% in early trading.
Every bank on the hosts' screen raised its price target, the highest at $420 per share.
Jensen Huang was on CNBC the day before, asked about OpenAI's Jalapeno chip, and did not sound worried.
"And Jensen was like, dude, we're the kings here. We've been doing this for 33 years. You've been doing this for three months. We'll talk to you when you've actually produced a chip, when you've put it into data centers and when you've scaled this thing." — Josh Kale
Nvidia is sold out for as much as it can make, operating at full capacity, and is now working with five of the largest banks in the world to fund more data center build-out.
Jalapeno, ASICs, and the Moat Jensen Says He Still Owns
Ejaaz Ahamadeen gave two reasons the revenue keeps climbing while everyone calls it a bubble.
The first is pricing: Nvidia raised the price of all its GPUs by 15%, largely because the cost of memory has soared.
He was careful that this may not correlate with the profit side, since Nvidia pays the most for memory of anyone building GPUs.
Memory stocks have been very volatile on the back of it.
The second is Jalapeno, OpenAI's own custom chip, which some read as a threat the way Google's TPUs are read as a threat.
On Ejaaz Ahamadeen's account Jensen Huang's reply is a distinction between general and specialized silicon: "we own the moat for generalized GPUs, for generalized AI chips."
Any lab can run anything on them: a neolab, a hyperscaler, Anthropic, OpenAI, and not only LLMs but financial models, fine-tuned models and models aimed at scientific breakthroughs.
Jalapeno is an ASIC, which has to be proven at scale, and nobody has seen the things in operation yet. They are expected by the end of the year.
Nvidia's own line, as Ejaaz Ahamadeen relayed it, is that it still controls 70% to 85% of the entire market.
The $13 Billion Bet on Open Source
Nvidia acquired Hugging Face for $13 billion, which Ejaaz Ahamadeen called the biggest of Jensen Huang's open source purchases so far.
He explained the platform for listeners who have not used it: anyone releasing an open model uploads it there, and anyone can upload datasets to fine-tune those models into their own versions.
His worked example was Zhipu's GLM family: around 10 official models, but roughly 155,000 GLM variants on the platform, tuned for financial trading, research and other niches.
The price puzzled him out loud: "Now, I ask myself, why on earth is Jensen paying 50X the amount of money that Hugging Face earns to acquire this open source model?"
His answer is that Jensen Huang does not believe the world ends up running on two frontier models.
Hundreds of niche and broad models, used collaboratively, with agents running everything — and all of it needing his GPUs.
The deal follows two other open source commitments: "He invested in Poolside, I believe, last week for $9 billion to acquire 100 of their employees so that he can build new open models." — Ejaaz Ahamadeen
Tens of billions more are going into Nemotron, Nvidia's own open source model, over the next couple of years.
Josh Kale's shorthand: "Hugging Face is essentially the GitHub for open source AI."
He noted OpenRouter had gone the same way, and Hugging Face makes two.
Three Reasons the Volume Is Moving to Open Weights
Josh Kale laid out why Nvidia has been so adamant about open source, in three parts.
First, it is defensive: "open source is a hedge, in a way, against its own best customers."
OpenAI, Google and Amazon are all building alternatives to Nvidia GPUs. Open source finds more demand for tokens if some of that share goes away.
Second, the trend is ossifying around open weights, because they are cheaper and mostly effective enough.
Frontier intelligence carries much higher margins and will be worth a tremendous amount to the people who need it, but the set of use cases that require it is shrinking as open model quality rises.
A GPU provider paid per token generated wants to be where the tokens are.
Third, the moat is moving up the stack. Nvidia optimizes models for CUDA, which works by default with its GPUs, and adds one-click deployment onto the Nvidia cloud.
Cloud businesses have insane profit margins and print cash, and this puts Nvidia in the middle of that.
Ejaaz's Tinfoil Hat: Hugging Face as the Cloud Business Nvidia Never Had
Ejaaz Ahamadeen asked to put his tinfoil hat on, and the argument starts from a gap: Nvidia does not have a good cloud business, which is remarkable for the company that sells the GPUs to Amazon and Google.
Its DGX service did not make much money at all.
The margin picture is the tell. Ordinary AWS or Google Cloud runs around 80% margins; AI cloud runs at only 40% to 50%, because Jensen Huang takes the bulk of the profit.
That is why the hyperscalers want their own chips and their own independence.
Hugging Face's inference service lets a user click a button and run any of the models on the platform — three million of them, on his count — and those calls get routed out to neoclouds.
"So my whole thing is Jensen is not only providing the chips, but he's buying a company where he can effectively churn into his own cloud business that can compete with AWS and GCP." — Ejaaz Ahamadeen
He tied it back to the $500 billion Nvidia raised to back purchases that end up coming back to itself, which the show had covered the week before.
The open question then was what happens if Jensen Huang cannot sell the chips. Owning Hugging Face means he can sell them to Hugging Face to serve that same cloud business.
He called it a circular virtuous loop that still makes business sense, and said he likes seeing Nvidia expand past making bleeding-edge GPUs.
The Mystery Model Was Z.ai, and It Ran on Chinese Silicon
The show had promised a follow-up on the anonymous model handing out 100 trillion tokens free to the public for a full week. It is Z.ai, the people behind GLM, and the model is 5.3 Flash.
"But they were generated entirely on a chip cluster that runs on Chinese silicon, which is a first of its kind." — Josh Kale
He had never seen an instance of Chinese silicon producing that volume of tokens and that much intelligence.
On the benchmarks it lands at roughly Opus-tier agentic level: "This isn't frontier, but it's 15 cents per million tokens." — Josh Kale
It understands video, audio and text.
Coming days after the show's Robot Olympics episode, Josh Kale read the two together as China making moves — dominant robots, and now silicon working at scale with quick inference and a lot of it.
Ejaaz Ahamadeen had used the model earlier, while it was still running anonymously as Ox Alpha, and it returned responses quickly, cheaply, and functioned very similarly to an Nvidia inference stack.
Export Bans, Huawei, and the Half of the Market Where the Money Is
Ejaaz Ahamadeen's context: America's export bans mean Nvidia cannot sell chips to China, and the record quarter was earned without selling a single chip there.
He called that impressive, since China had been a major revenue holder.
China's response has been to build its own. It has not yet built a chip good enough to train frontier models, so it uses techniques like distillation to work around that.
The chips it can build are inference chips, and Ejaaz Ahamadeen argues that is where the money in AI will be made, by a vast order of magnitude over pre-training.
"So can someone please explain to me how six months later, they're now here running the top stack of one of the top open source models, which happens to compete very well with U.S. frontier models purely on Chinese silicon." — Ejaaz Ahamadeen, on chip companies that were two and a half years behind six months ago
The chipmaker is Huawei. He would not attempt the chip's name from memory; the cluster was announced about a month ago and is now at full-scale production.
The Chinese state has mandated that labs stop relying on Nvidia chips for anything, and the labs replied that they cannot do it for training but can for inference.
"Like, what happens when they own the chip infrastructure layer and the robotics hardware manufacturing layer?" — Ejaaz Ahamadeen
Josh Kale's read is that the ban runs both ways: Huawei devices are banned in the United States, and a country without access to chips has to go and build its own.
Waymo Moves the Whole Driving Stack Into the Trunk
Waymo introduced a new vehicle, the Ojai, and a new chip architecture inside it.
"But the chip that handles it is now 20 times more powerful than it was over the last eight years." — Josh Kale
He was clear this is not a cost story for the ride, and put the Jaguar Waymo at around 260,000 per vehicle.
Ejaaz Ahamadeen explained why the chip matters. Earlier Waymos carry 15 to 20 sensors — cameras, proximity sensors — and all that data has to become a steering, braking, get-there-safely decision.
Waymo did not have a chip powerful enough, so it ran a hybrid of on-board silicon and a cloud service.
"Now, that's dangerous, because what if the cloud service breaks connection?" — Ejaaz Ahamadeen
The new design runs the entire system locally, on the car: "it brings all the compute on board the actual device—the device in this case being a car." — Ejaaz Ahamadeen
The promotional images show the compute sitting under the trunk, which both hosts thought was the best visual in the announcement.
The chip is 20X more powerful and 5X cheaper on Ejaaz Ahamadeen's reading of the blog post, which he called very impressive for any successive generation.
Aesthetics got a mention: he respects Google a lot and thinks the older cars were among the ugliest autonomous vehicles out there. Josh Kale found the new one fun and different, like Disney World, and wondered why a fully autonomous car still has two forward-facing seats and a steering wheel.
The Correction: That Number Is the Hardware, Not the Car
Ejaaz Ahamadeen read a breaking-news figure off screen and put the new cost at $25,000 a car, down from $100,000 to $125,000. Josh Kale stopped him.
"No, no, no. So this is for the hardware. This is not for the vehicle." — Josh Kale
The comparison is like for like once corrected: roughly $125,000 of sensor suite on top of a $100,000 vehicle before, and $25,000 of hardware on top of an assumed lower-cost vehicle now.
Ejaaz Ahamadeen accepted the distinction and restated the claim as a hardware one: it runs locally, so it is cheaper to run, quicker, more reliable and more efficient, and Waymo's cost comes down by three to four times.
On the show's own terms this is the part that matters: "You know, we like hardware on this show." — Ejaaz Ahamadeen
The CyberCab Starts Rolling on September 3
Josh Kale's counter-number: "The entire CyberCab from Tesla, the whole thing, including the vehicle." — for what the Waymo sensor suite alone costs.
The CyberCab uses about eight cameras, very similar to the ones in a smartphone, rather than a spinning LiDAR stack and a complicated sensor suite.
It is being deployed starting September 3rd at scale, in Austin, Dallas, Houston, Miami, Orlando and Tampa.
Josh Kale gave Waymo its due: much more accessible today, many more miles per day, and in Los Angeles or San Francisco a part of life people no longer think twice about.
The CyberCab's opening is cost per mile, which he expects to come in significantly below Waymo's.
The Waymo vehicle has big bumpers and is not efficient; the CyberCab is, and efficiency is what drives the per-mile number down.
Geofencing is the other difference. Waymos work in specific locations; he has used Tesla full self-driving for years with no oversight, and says it works anywhere.
"So as soon as regulation comes online, we may see a world in which, like, 12 months from now, CyberCabs are just roaming around everywhere." — Josh Kale
The end state he described is not owning a car at all, because cost per mile falls below the cost of buying the physical vehicle.
He flagged the Tesla event happening the following week.
New York, and a Producer Who Has Not Ridden One
Ejaaz Ahamadeen's one place he is not watching for any of this is New York City. Josh Kale's explanation was that the city likes to ban all innovation.
Both hosts agreed the city is cool and brutal at once, and asked for the new tech to arrive there early.
The segment ended on a call-out to producer Luke, who lives in Austin, one of the launch cities.
Ejaaz Ahamadeen put his robotaxi ride count at zero and did not believe the photo: "I don't trust it. That image was AI generated. I need to see you, Luke."
Josh Kale allowed he may have done one, but not in a CyberCab. His sign-off on the exchange was that they will get a fact checker in.
Leopold Aschenbrenner, the SEC, and a Fund That Is Still Up
Josh Kale's one-sentence history, offered as the answer to "who is our favorite investor in the world": "Oh, you mean the 25-year-old who was fired by OpenAI, raised a billion dollars and then had 45 and then lost it all and then Citadel bought it all out and now he's being investigated by the SEC?"
He then had to ask for the name. Ejaaz Ahamadeen supplied it: Leopold Aschenbrenner.
The specific action is that the SEC has subpoenaed a number of banks that facilitated trades for Situational Awareness Fund, Aschenbrenner's fund, on suspicion of insider trading.
The show has referenced the surrounding facts before: Avital Balwit, now officially Aschenbrenner's wife, is chief of staff at Anthropic.
The question Ejaaz Ahamadeen framed is whether Aschenbrenner traded on information he should have disclosed publicly. His answer was that nobody knows.
Aschenbrenner has been getting back into the game, with a $400 million private investment in another company, and technically the fund is still up.
Josh Kale's view is that the process is being read as a verdict: "It could be nothing, it could be something. Chances are it's probably nothing."
A complaint gets filed, someone acknowledges it, an investigation is open, and then the investigation itself becomes the story.
"And I think Leopold's just down, people want him to be down even more, and chances are he's going to be just fine." — Josh Kale
He said he is looking forward to the new 13F to see what the comeback portfolio holds after the public one was wiped out.
The FT Chart on Anthropic's Model Usage
A Financial Times chart of usage across Anthropic's models showed Fable accelerating from mid-June to the end of July and then flattening out.
Ejaaz Ahamadeen's first response was to question the source: "I don't know where they've got these numbers from."
He read the flattening as substitution rather than decline. Companies pick other Claude models for cost, efficiency, preference or the personality of the model.
The chart pairs the Anthropic line against rising OpenAI usage, particularly on Codex, which he called a fantastic product.
His caveat is a base-rate one: OpenAI has a smaller user base than Anthropic, so any rise shows up as a larger percentage increase.
The cost logic he described is a split rather than a switch: "maybe if I can get 80% of my work done using a cheap model, I'll use Fable for the other 20%."
Fable is notably more expensive but also operating at the frontier better than any other model, which is why the hesitancy makes sense to him.
The new open source models, GLM 5.3 Flash among them, are what makes the cheap 80% possible.
He called the piece a hit piece and a nothing burger, on the grounds that the pie is growing bigger and bigger.
Josh Kale agreed the chart shows the pie growing: total business spending is going up and to the right, consistent with heavy demand for lower-cost tokens on top of rising overall demand.
He could not work out what the chart's "other" category was, given Sonnet, Opus and Fable are all named.
Zero Data Retention Is the Enterprise Blocker
Ejaaz Ahamadeen offered a second explanation for the enterprise numbers, and it is a compliance one rather than a product one.
His starting point is that price is not the obstacle it looks like: "when you're an enterprise, you don't care about spending more money if it means you're going to make even more money." Enterprises want the latest frontier models.
Zero data retention is the sticking point: whether the vendor holds a customer's proprietary information and can use it to build better models.
He believes Anthropic is in the process of evolving its position, though he was clear this is rumor.
If that switch flips, his expectation is that the same chart spikes.
He thinks OpenAI is not enforcing zero data retention right now, which may be part of why it has accelerated over the last couple of months.
Josh Kale's summary of both segments was that the chart may be a nothing burger, but token use is real, and tokens are generated mostly by agents.
One Video Is Enough: In-Context Learning for Robots
Josh Kale set the week up as a big one for robots: the Robot Olympics, then full self-driving vehicles, then robotic learning.
He described a parallel announcement where people are paid to perform household tasks on camera to generate training data.
"There was another one that I thought was super cool where you can actually go and get paid like $15 an hour to perform tasks in your house." — Josh Kale, on being paid as though you were a full-time employee
Earlier examples he recalled: DoorDash paying Dashers to strap cameras to their heads to capture what delivering an item looks like, and Niantic, the Pokemon Go company, doing something similar.
What he liked about this particular demo is that it is not a humanoid — it is a narrow-purpose robot, and the physical capabilities emerge from throwing a large amount of data at the model rather than being programmed.
He met the obvious objection head on: the question is not whether it is bad to offload chores to robots, but how many people actually love to clean, cook, take out the trash or deliver food. The other side of automating those is much lower costs for everyone using the services.
Ejaaz Ahamadeen explained the technical breakthrough, which is in-context learning.
"You can feed this robot one video of you making pancakes and flipping the pancake, and it'll be able to figure it out from that 30-second video and do it itself." — Ejaaz Ahamadeen
The part he found remarkable is that it does not copy the demonstrator's failure: "It knows when you've made a mistake and it can take it further and start making multiple pancakes at the same time and then stacking them on top of each other."
The demo in the video swapped a dustpan brush's handle for a banana. The robot recognized it was not a brush and swept the block onto the pan anyway.
His conclusion is that it has an intuitive understanding of the task and will use any tool to do it, which gets a useful robot into a home much faster than hard-coding every action.
The Apple Event Is September 9, and Tim Cook Is Not Running It
"Here we have a date for the Apple event. It's coming. It is September 9th at 1 p.m. Eastern time." — Josh Kale
Siri is not launching at it. He drew the distinction between the developer conference, which unveiled Siri, and the hardware event, with the Siri release coming shortly after and paired with the new devices.
Expected hardware: a new iPhone 18 Pro and a foldable device, which he is incredibly excited to get his hands on.
The bigger change is who is on stage: "Tim Cook isn't going to be on stage here. It's John Ternus who's running it." A new chief executive, and a hardware person.
The hardware is built specifically for Apple Intelligence, which he pointed out they have been waiting roughly three years to get inside these phones.
New Macs, and What Local Inference Now Costs
Ejaaz Ahamadeen brought up the M5 Pro and M6 announcements alongside a new Mac Mini and a new Mac Studio, and thought the price points were up there.
"So it's like, I think it's like 900 bucks for the M6, and then 1,500 for the M5." — Ejaaz Ahamadeen
Josh Kale asked him to guess the high end of the Studio. Ejaaz Ahamadeen guessed three thousand dollars and was told he would be wrong.
"No. I think it's like eighteen thousand five hundred dollars for them." — Josh Kale, on a maxed-out Mac Studio with the M6 Ultra and all the RAM available
Josh Kale's conclusion for anyone eyeing a home rig is to let the big providers serve the inference, because doing it locally is getting expensive. His alternative was paying for a subscription or an API key.
Ejaaz Ahamadeen's alternative was buying an Nvidia GPU instead.
Ejaaz Ahamadeen put the blame in one place: "Wait, blame the memory manufacturers for this. That's why the price is hiking across everything right now."
Josh Kale added that the memory makers have great margins and are driving up all the prices.
Sam Altman's Seven Swiss Watches
The last item, which Ejaaz Ahamadeen billed as arguably the biggest topic of the episode: Sam Altman has ordered seven ultra-rare custom Swiss watches from Vanguart, with an OpenAI logo on the dial.
He asked listeners wearing an Apple Watch to take it off for a second and appreciate the piece.
Josh Kale, holding an Apple Watch, observed that its band looked oddly similar to the band on the Vanguart, and asked how many tokens per second the watch generates.
Ejaaz Ahamadeen's answer was not enough, and that the strap is not the point — the complex cognitive function inside the case is.
His reading of the gesture: "So the point is, I think this is Sam giving a peace offering before potential rumored IPO that I think is coming very soon, and he wants to give it to his ranking order of people that haven't left OpenAI over the last two weeks."
He apologized for the shade, and said he has seen a lot of executives leave recently.
On resale, he thinks the crowd is guessing low: "people are guesstimating this would be in the order of 500K to a million dollars. I personally think it would be higher because there's only seven of them." One of the seven goes on Altman's own wrist.
Josh Kale's analogy was a push present, the gift given when a partner has a child: "It's like when your executives produce AGI."
He tied the timing to something he had read: "and they fully expect to have AGI internally by the end of this year." — from a Time piece on Sam Altman
His note on that: the episode is airing as September begins, which is close to the end of this year.
Josh Kale's read on the week is that it belonged to the robots and to China — a Robot Olympics, self-driving vehicles at scale, a continual-learning robotics model, and Chinese silicon serving frontier-adjacent intelligence at a fraction of the price — with Nvidia and Jensen Huang taking the other half of the win.
Products, Companies & Tools Mentioned
Nvidia (Almost $100 billion in a quarter, growing 100% year over year, worth over $5 trillion, guiding to 70% growth in 2028 against 44% expected; raised GPU prices 15% on soaring memory costs)
Hugging Face (Acquired for $13 billion; the GitHub for open source AI, and on Ejaaz Ahamadeen's theory the seed of an Nvidia cloud business via its inference service)
OpenAI and its Jalapeno chip (A custom ASIC that has to be proven at scale; the claim that it is more efficient than Nvidia's chips is what Jensen Huang was asked about on CNBC)
Z.ai and GLM 5.3 Flash (The anonymous model handing out 100 trillion free tokens, run entirely on Chinese silicon at 15 cents per million tokens)
Huawei (Built the inference cluster behind the Z.ai run; announced about a month ago and now at full-scale production)
Waymo (New vehicle and a new chip that is 20 times more powerful and 5 times cheaper, moving all compute on board and out of the cloud)
Tesla CyberCab and Tesla full self-driving (Eight smartphone-grade cameras, deploying September 3 in six cities, with a lower cost per mile and no geofence)
Anthropic, Claude, Fable, Opus and Sonnet (The subject of the Financial Times usage chart; Fable more expensive but operating at the frontier better than any other model)
OpenAI Codex (Called a fantastic product, and the driver of the OpenAI usage line on the same chart)
Poolside (Nvidia's acquisition of its team to build new open models)
Nemotron (Nvidia's own open source model, getting tens of billions over the next couple of years)
OpenRouter (The other recent acquisition in the same direction, which the hosts had covered previously)
AWS, Google Cloud and Nvidia DGX (Roughly 80% margins on ordinary cloud against 40% to 50% on AI cloud; DGX never made much money)
Situational Awareness Fund and Citadel (Aschenbrenner's fund, the banks that facilitated its trades now subpoenaed, and the buyer that took it out)
Apple, the iPhone 18 Pro, Mac Mini, Mac Studio, the M5 and M6 chips and Apple Watch (September 9 hardware event under John Ternus, with Siri arriving after the devices)
DoorDash and Niantic (Paid humans wearing cameras as a way of collecting robotics training data)
Vanguart (The Swiss maker of Sam Altman's seven custom OpenAI watches)
Securities and Exchange Commission (Subpoenaed banks over suspected insider trading around Situational Awareness Fund)
Books & Resources Mentioned
The Financial Times chart on Anthropic model usage (The chart the hosts spent a segment on, and which Ejaaz Ahamadeen called a hit piece)
Time's piece on Sam Altman (Josh Kale's source for the claim that OpenAI expects internal AGI by the end of this year)
Situational Awareness Fund's next 13F (What Josh Kale is waiting on to see the comeback portfolio)
Get the latest market chatter and takes as they happen:
X | Threads | Instagram | YouTube | TikTok | Facebook