3 operators covering the hottest tech news Top 50 Technology Pod on @spotify Hosted by @anikamirzaa @priya_kamdar @mayamshahh
San Francisco, CA
Joined June 2024
- Tweets313
- Following202
- Followers510
- Likes383
Anthropic just filed its S-1 this morning. A week ago the podcast wasn’t sure if IPO numbers were at risk:
@mayamshahh : What we’re seeing is that enterprises are not willing to pay for the frontier model over the open source models as much as we thought a quarter ago.”
@priya_kamdar : I think they’ll be fine, but I do think that their IPO numbers are at risk.”
@anikamirzaa : 5 to 7 percent [lost from distillation revenue] is billions of dollars. And apparently, a lot of it has been through wrapper companies, which is pretty interesting because it makes it harder to track.”
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO.
Read more: anthropic.com/news/s1-sec
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
Phones get blamed for the loneliness epidemic. @nadyaokamoto thinks they can be part of the cure.
Nadya: "I don't actually think it's realistic to ask this generation to put their phones away... there's a lot of magic that can happen online"
"Pie is like matcha. Coffee and Red Bull, it's all caffeine, slightly addicting and probably not the best for you. But matcha is a healthier version"
“How do you make an app that is matcha? You know, like, that there’s some benefits to it, that are beyond it and you feel those benefits in other parts of your life”
In SF, not having a full-time job is a red flag. In NY and LA it's a flex
@nadyaokamoto: "In small talk with other creators, they ask like, 'Oh, do you do this full time?' And it's more of something to be proud of that you don't have a full-time job"
@anikamirzaa says in SF the view on creators is different:
"In San Francisco, saying you don't have a full-time job is like, 'Oh, what are you doing?'"
Nadya:
"That question is coded as... are you so successful as a content creator that you're doing that?"
"It's cringy until, as a content creator, you make someone else's full-time year salary on one post"
"It's kind of like a fuck you amount to be like, 'I'm living my life and I'm getting paid to live my best life and make content about it'"
When you’re reselling compute this is the loop that hits companies the hardest
@anikamirzaa
"There's also a way that CoreWeave gets doubly screwed here."
"Their old bonds were at that fixed 10%. Their new bonds are at the 16%... since they're just reselling compute, that increase on their debt directly cuts into their profit margin."
"And the cut into their profit margin makes them a riskier business to loan to... it's an endless loop."
On credit default swaps:
"You don't actually have to own the debt to get insurance on it."
"Let's say I got insurance on a fire happening in your house... I think I have alpha, so I'm gonna buy it too."
"Hedge funds are going out and buying insurance on their debt... a really bear signal for them."
CoreWeave’s DEFAULT RISK is exploding higher:
CoreWeave, $CRWV, 5-year credit default swap (CDS) spread surged as much as +74.5 basis points on Thursday, to a record 1,194.5 basis points, implying a more than 60% probability of default within 5 years.
Meanwhile, its 9% bonds due 2031 fell ~0.5 cents on the dollar to 82.75 cents, pushing yields to 14.5%, while its 2032 bonds were quoted at ~78 cents on the dollar.
At the same time, $CRWV shares bounced +5.9% after Nvidia, $NVDA, CEO Jensen Huang pushed back against concerns over CoreWeave’s financial position, saying he does not expect the AI data-center company to need a bailout from Nvidia, temporarily easing concerns over its growing reliance on external financing.
Meanwhile, $CRWV shares are falling another -4% on Friday, putting them on track for their worst week since mid-August.
Credit fears intensified after co-founder said CoreWeave aims to secure investment-grade credit ratings to unlock access to “unlimited” funding from insurance companies for its AI expansion.
The pressure in credit markets also intensified as September US producer prices (PPI) rose at their fastest pace in more than 4 years, raising concerns over further Fed rate hikes.
Investors are increasingly questioning the cost of financing the AI buildout.
The internet recently tried to cancel @nadyaokamoto for being in the office at 1 AM, but she doesn't regret posting her real life:
"In my line of work, whether it be content or entrepreneurship, you have to fucking hustle... people aren't investing in you if they don't think you're gonna hustle"
"What you also didn't see is that then I slept until 9 AM. I still got eight hours of f*cking sleep"
"I did try posting, like, a meditation... and it f*cking flopped. Nobody wanted to see it"
@anikamirzaa:
"What people get mad about is actually also what they wanna see"
Should candidates get to use AI in interviews? @rvivek answers this using an analogy from aviation
@mayamshahh asks:
"The belief is that a smart person, once exposed to AI enough, will be able to use AI in the best way possible. And so why test them with AI versus not test them with AI?"
Vivek:
"If you think about flying, 90% of your flight is flown by AI. But you are not gonna get on a plane with a vibe piloter."
"The first round is evaluating you on your fundamentals of aeronautical engineering, which is a set of 100 multiple choice questions. There's no AI. The second is a flight simulator."
"I still think most companies will retain one round that is without AI."
"Maybe it's just reviewing code… like a code review question gets very popular."
How do you pick the next Stripe?
@cjc joined at employee 28 after meeting @patrickc at a barbecue
"I knew nothing about payments. I didn't know what an API was"
"if you meet with Patrick Collison, he'll use words that you perhaps have to look up in the dictionary"
"a lot of great founders have a very strong vision for what they want to achieve. and that vision has never wavered"
"he was most interested in building a great company, not necessarily just a great product"
"often it's not the first version of your app that makes you into a really fantastic business. it's what you can evolve into"
Linear's COO @cjc on why Linear is NOT 9-9-6
"our founders are late thirties to late forties. they all have kids and have families"
"we don't require that people come into the office and show their face and sleep in a nap room, or on a mattress under their desk, to prove that they're doing good work"
"they've done the hyper-growth thing. they worked at Uber and Airbnb and Coinbase, and realized you can only do that for so long before you start churning and burning through people"
"the idea was to build a culture that is more of a marathon and not a sprint"
Linear clearly has a unique culture, I respect the intentionality with which the company operates
More founders should study this example but not as an alternative to the 996. Quiet the opposite. It's to understand that copy cat culture doesn't work
Leaders need to figure out what team culture uniquely aligns with the product strategy, customer, category
It might be 996 for some very competitive spaces where velocity is king. It might be in person for real world operating environments. Remote first for hyperlocal, etc
Intentional culture >> copy cat culture FTW
> be @nadyaokamoto
> use instagram to organize political movements
> start a nonprofit at 16 and scale to 800 chapters
> run for office at 19
> don't get elected but realize the power of social media
> become a published author at 20
> make forbes 30u30 at 21
> start a d2c company August, cross $1M in sales in year one
> reconnect with your sisters after 8 years of not talking
> grow to 6M+ followers
> start using the app Pie to organize your run club Zoomies
> get approached by @dunn (prev. exited Bonobos) to become Pie's co-founder
> your wedding is in the NYT
> work on Pie during your honeymoon because you want people to build community irl
Full episode out now!
.@nikitabier's Intro page lists $7,500 for a 1:1 session. Would you pay to talk to his AI twin instead?
One pricing model @sam_kececi has explored for Sentience: people pay to talk to your digital twin, Sentience takes a cut, and you get paid without being on the call.
"Sentience takes some cut, and then you actually can profit."
Sam gets a lot of inbound from early founders asking for advice himself:
"I would love to just take an hour with every single person, but it would just add up like crazy."
Get the Check retweeted
Anybody can create an output with AI right now. So evaluation can't be about the output anymore. It's about the process: how you got there, the quality of your prompts, how you actually think.
From my chat with @mayamshahh on @Getthecheckpod
Get the Check retweeted
Thanks for inviting me to the pod. This was fun!
[EPISODE] The first time @rvivek flew internationally, he asked the flight attendant, “This is going to San Francisco, right?” He was on his way to a 10-minute YC interview. He got in, built HackerRank, and helped define how the world hires engineers. Now he’s rebuilding it for the age of AI.
Vivek is the co-founder of HackerRank, and he and @mayamshahh chatted about:
>How tech interviews evolved from Microsoft’s golf ball questions to AI agent orchestration
>Why HackerRank is tearing down its old assessment model
>How to tell a vibe coder from a real production engineer
00:00 Meet Vivek, Founder of HackerRank
02:15 From counting Golf Balls in a 747 to LeetCode
05:55 Refounding HackerRank With Chakra
08:56 Process vs. Output and why partial credit matters
12:31 Who Adopted Online Assessments First
14:20 Why Hiring Is a Lagging Indicator
17:38 What Candidates Like About Chakra
18:20 How to Remove the Incentive to Cheat
22:44 The Three Ways People Cheat
26:59 Why Hiring Is a Hard Problem and Build-Your-Own Is Overrated
28:11 First Indian YC Company and a First International Flight
Remember when every YC startup was AI for dentists?
@sam_kececi doesn't think that was a coincidence:
Sam: "AI does have taste already in some sense, but it has taste across one dimension. It is the taste of Claude."
"If you ask Claude, 'How should I run my business?' First of all, just like NGMI if you're doing that."
"But if you ask Claude for startup ideas, notoriously, it gives everyone the same startup ideas."
"Billions of people are using these models, of course it's gonna make its way into... what types of companies get started."
here’s a concrete example of how to make money with this new Claude drop
build and sell AI agents for dentists
a dentist’s office has the same 6 problems every single month:
1. patients not booking
2. no one answering calls at night
3. unpaid bills piling up
4. bad reviews (or no good reviews)
5. chasing people who haven’t been back in a year
6. new patient forms done manually
here’s what your agent will do to solve these problems:
1. books appointments. patient fills out a form, it checks the calendar, sends the confirmation, reminds them 2 days before and the morning of. no receptionist needed.
2. answers every question at 3am. “do you take delta dental?” “how much is a crown?” it’s trained on the dental office’s info and answers instantly.
3. chases unpaid bills. sends polite follow ups, escalates if ignored. no awkward phone calls.
4. collects google reviews automatically (my favorite part). after every appointment it texts the patient. happy ones are asked to leave a Google review. unhappy ones get flagged privately before they post anything.
5. handles new patient intake. sends forms before the appointment, collects insurance info, follows up if anything is missing.
6. reactivates old patients. “hi sarah, it’s been 8 months since your last cleaning, want to book?” sent automatically to every single lapsed patient on the list.
just build one AI agent that handles all of that and charge dental offices a $1,500 set up fee and $500 per month to maintain it.
20 clients = $10k per month
the same playbook works for gyms, law firms, real estate agents, accountants.
never has it been so easy to make money online
[EPISODE] The first time @rvivek flew internationally, he asked the flight attendant, “This is going to San Francisco, right?” He was on his way to a 10-minute YC interview. He got in, built HackerRank, and helped define how the world hires engineers. Now he’s rebuilding it for the age of AI.
Vivek is the co-founder of HackerRank, and he and @mayamshahh chatted about:
>How tech interviews evolved from Microsoft’s golf ball questions to AI agent orchestration
>Why HackerRank is tearing down its old assessment model
>How to tell a vibe coder from a real production engineer
00:00 Meet Vivek, Founder of HackerRank
02:15 From counting Golf Balls in a 747 to LeetCode
05:55 Refounding HackerRank With Chakra
08:56 Process vs. Output and why partial credit matters
12:31 Who Adopted Online Assessments First
14:20 Why Hiring Is a Lagging Indicator
17:38 What Candidates Like About Chakra
18:20 How to Remove the Incentive to Cheat
22:44 The Three Ways People Cheat
26:59 Why Hiring Is a Hard Problem and Build-Your-Own Is Overrated
28:11 First Indian YC Company and a First International Flight
Will the AI bubble burst?
One case study, to prove or disprove this point, is asking if Fable 5 will be able to generate enough revenue to exceed the model development costs.
@priya_kamdar, @anikamirzaa and @mayamshahh explain the fundamental economics of how it needs to work
Maya: "Do you guys think they collapse, or do you think they find new revenue to capture?"
Anika: "I definitely don't think they'll collapse at all. Anthropic's revenue growth rate right now is insane."
Maya: "My concern is not that their revenue collapses. It's how much they're paying back their older generations of models."
"They spent X amount of money to train Fable 5, but they're not gonna get that amount of money back in revenue. And then that model will be outdated."
"At some point these models have to be priced to how much it takes to train them."
Priya: "Coding will decrease from where it is right now, for sure. They definitely will not have the same growth rate."
Maya: "Seems a little Jevons paradox-y."
🚨 Anthropic IPO S-1 prospectus LEAKED:
Valuation target: $2 TRILLION
>$4.6 billion in revenue in 2025
>$8.06 billion in operating loss
>$42 billion in net loss
The prospectus warns that nearly 25% of revenue came from just TWO customers, and that many top enterprise clients aren't locked into long-term contracts and could cut spending or walk away anytime
Anthropic plans to spend $518 BILLION on compute commitments ‘in the coming year’
This data center company found free energy.
@mayamshahh : "What they found in Abilene, Texas was wind turbines producing so much energy, 'cause it was so windy there, that the energy was negative. It was priced negatively."
"They literally had to turn off the wind turbines 'cause they didn't know what to do with it. No one wanted that power."
"You can't move power, it's too expensive. You have to use it there. But no one lives in Abilene, Texas, right? So what's gonna use the power? A big f**king GPU data center."
@priya_kamdar : "Abilene is where the first Stargate facility is gonna be. It's a 1.2 gigawatt campus. That's what put Crusoe on the map."
[EPISODE] An investor talked to @sam_kececi's digital twin for days and it convinced them to ask for a SAFE
Sam is the founder of The Sentience Company and is building digital twins for operators that think and write like them. @anikamirzaa and him chatted about:
>How we can stop AI from converging on the mean
>What post-AGI high school should look like
>Why you may pay to access someone else’s digital twin
03:25 Wildest Sentience use case: it made someone invest in his company
06:54 Is taste actually a moat, or will AI catch up
08:36 The reason there are so many AI dentistry companies
17:31 Why is Anthropic bad at creative writing
18:24 Will Sentience be a dumb product if you're dumb
25:15 Why Sam doesn't believe in engineering standups
26:30 How to monetize consumer AI agents
30:16 Are we all going to have our own models, or will agent harnesses be durable
36:00 What should a post-AI high school look like
Anthropic and OpenAI are slashing prices and one graph explains why.
@priya_kamdar "OpenAI released GPT-6 Sol and Luna at 50% less token prices, which clearly shows they're being pressured by the open source models, and by usage, because ever since Fable came out we've basically seen that no one's using it."
"A majority of the spend on those tokens comes from the top 1% of users. It's not even companies. There's a few power users at the very top, and everyone else is way below."
"Fable 5 lost two-thirds of its spend share in one month. It went from 13% to 4%. Opus 5 tripled to 22%."
"And the open source models went from 7% in December to 56% now."
"The same pattern is seen on the consumer side of AI spend too."
Looks like today may be a record day for token volume % of open models on Vercel AI Gateway:
🟦 Open 78.4% 🟨 Closed 21.6%
While spend 💲 usually tells a different story, #3 and #4 today are Moonshot AI & DeepSeek. Adding Z.ai, their combined spend surpasses OpenAI (#2).
(Do note that's the spend for inference of the model across providers (mostly in the US), not revenue going directly to the open weight labs.)
