3 operators covering the hottest tech news Top 50 Technology Pod on @spotify Hosted by @anikamirzaa @priya_kamdar @mayamshahh
San Francisco, CA
- Tweets 284
- Following 199
- Followers 462
- Likes 359
Is a decision-only model any good for trading? We disagreed, then tested it on our own recruiting inbox.
@priya_kamdar: "Trading bots can operate so quickly. If you're computing all the probabilities, it's a great area. Tons of inputs, a clear output, do you buy or sell."
@anikamirzaa: "Buy or sell is not a deterministic answer. The vast majority of people can't figure it out with any amount of information. That's like early stage VCs using Jev to figure out if they should invest in a startup."
Priya: "It depends on your trading strategy. A lot of people literally never look at anything. They have 100 different indicators, and when they all line up, do something."
Anika: "If it's effectively a long if-then statement, then yeah, I do think it would be really good."
Anika: "We get a lot of applications into our recruiting form. Priya used Jev to create a classifier of our candidates, and it was pretty good. We already have a golden dataset, because when people come in, we decide if we're gonna text them or not."
Jev is INSANE.
I built this in an evening and morning. A real-time trading bot ingesting onchain+offchain data to make rapid decisions about trades. Fully autonomous.
So far it has lost me $31,680.
1
297
[LIVE RECORDING] Sat down with @cjc, @linear COO and @stripe's 28th employee to talk about:
>Why Linear did a $2.5b tender instead of going public
>Hot takes on 996 and DEI
>How @patrickc and @collision were the og forward-deployed engineers
> Where she would apply if she could join any startup right now!
Thank you @SutardjaCenter for co-hosting <3
00:00 Cristina Cordova, COO of Linear
02:54 Joining Stripe before knowing what an API call is
05:34 The Collisons were the OG FDEs
08:57 Leading DEI at Stripe
12:52 Angel investing in 80+ companies
18:52 Why Linear did a $2.5B tender offer
22:25 Cristina's hot take on 996
24:27 How Linear stays profitable
27:15 Where she'd work if she were 22
29:09 Linear is hiring!!
1
2
7
826
Jev's launch video did 36 million views and barely anyone outside the company can explain how it works. @mayamshahh explains it plainly:
"Normal LLMs are not great classifiers because they're just slow. They have to go through all their chain of thought reasoning before they even get to their final decision."
"Say you have a patient note. It'll run three questions, appending 'has diabetes,' 'has cancer,' 'has other,' and then it looks at the distribution between true and false for each of those. That's how it's picking the decision."
"It's taking a normal transformer architecture and not doing the last step, which is actually predicting the token, but taking all the probabilities you get right before you predict the token."
"They called it a System 1 model because they want it to be your immediate reaction, your immediate gut."
"No one knows how they were able to get the model to be so accurate. I have no idea how they made this universally great classifier with low thinking."
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
1
3
701
All three of us switched to Codex this month. @priya_kamdar has a pod prediction:
"This marks the return of @OpenAI. Everyone's been switching from Claude Code to Codex."
"People counted OpenAI out for a while, and they were all in on Anthropic. They're back."
@mayamshahh: "I use Codex now. It's just really, really good at executing fast and accurately."
"If there's something I really, really don't know how to solve, I still use Fable. But if I'm 90% sure, I will just use Codex."
Priya: "Claude just takes so much time."
Maya: "The Claude spinner thing. I hate it."
Just met a Claude Code user who has been the top daily user numerous times.
He says Codex is better now.
1
4
1,269
Get the Check retweeted
New @typesafeai Jev + @ExaAILabs use case
We manually triage every candidate that comes through @Getthecheckpod's recruiting pipeline, which some days is 100+
To test Jev's decision-making, we had Exa enrich each LinkedIn profile, then Jev scored whether to schedule an intro call + rated candidate strength 0-2
Within seconds it triaged 400 candidates for $1.80 ($1.75 Exa, $0.05 Jev)
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
6
4
27
2,888
Get the Check retweeted
People always ask me about the Chief of Staff role:
Being an assistant is very underrated.
It sounds low-status but if you're someone's really important right-hand person, you get visibility into crazy things and your scope is pretty much infinite that person will always have more things to do than time.
They'll help you with whatever is next. Link to the podcast's recruiting network in the replies. We place 100s of engineers and ops people.
2
2
25
6,314
Moderna ripped as much as 177% in a day on its melanoma vaccine data. The FDA's former vaccine chief @VPrasadMDMPH says the details that matter are the ones not in the press release.
"Melanoma is a very unique cancer. I think this has not been in the news enough."
"There's a lot of failure in this field of cancer therapeutic vaccines."
"They didn't announce overall survival. They didn't give any numbers, but apparently Merck never gives numbers in their press release. So we'll see."
"Merck is very smart because now whoever gets the shot's gonna have to get the Merck product. They're gonna get Coke and not Pepsi."
Moderna and @Merck today announced positive topline results from the Phase 3 INTerpath-001 trial evaluating adjuvant treatment with intismeran autogene, a novel investigational mRNA-based individualized neoantigen therapy jointly developed by Merck and Moderna, in combination with KEYTRUDA® (pembrolizumab), Merck's anti-PD-1 therapy, in patients with completely resected Stage IIB-IV melanoma.
Read more:
4
20
10,536
What happens inside the FDA when it rejects a drug? @VPrasadMDMPH Prasad says half of Congress shows up in your inbox.
"The FDA is under constant political bombardment to lower [the bar for approving a drug]."
"When you work at the FDA, my email, phone is blowing up with half of Congress."
"I've been summoned to senator's office. The senator says, it sounds like you're not approving this product. They came in here earlier today and they say you guys are being very difficult."
"And then we say, well, look, Senator, here's things you may not know. And they start saying, oh my goodness, I had no idea. They didn't mention all those things. Oh, well, carry on kids."
2
5
16
1,313
Why does Instinct feel different? Its memory allegedly forgets 90% of you, but that’s actually the point. @anikamirzaa breaks down @ashwingop's thesis:
"Everyone's question is, how do you solve memory? Most AI assistants have a database of memory: all your Granola notes, transcripts, everything."
"What Instinct seems to have figured out is identifying what is and isn't important, and forgetting the unimportant stuff. That's the difference between human memory and an agent's memory built on a database."
"Humans forget 90% of the things that happen to us. Just ask Maya and me. We're still good at our jobs because we remember the 10% that matters. When agents try to remember everything, they remember nothing."
Anika’s question around the idea that this architecture is more secure:
"I replied and said, I don't really get that. If it still remembers everything most important about me, my Social Security number for example, that's still a decent amount of information."
"And he said: in a breach, it makes the radius much more confined, because it's not retaining your full email inbox. It’s meaningfully more secure."
4
46
13,075
This week: our take on if the @DarioAmodei, @sama, @elonmusk doomerism narrative was organic or a coordinated psyop to combat open-source catching up
Also, GPT-6 Astra scoring 98.6 on ARC-AGI-3 and getting much better at browser work means a lot of seed stage companies focused on the “back office” are going to zero
Finally, @Apple dropping a $2k foldable phone 7 years late into a market the size of one month of normal sales is classic Apple, but could actually work. Tune in to hear if @mayamshahh's buying it
00:00 Debriefing NYC
07:29 The AI researcher tweet that hit 171M views
09:46 Does @AnthropicAI have an incentive to over blow AI doomerism
13:05 Dario’s essay on pacing the frontier
25:13 The nuclear weapons parallel
26:15 Our take on how serious AI safety risks are
28:59 @OpenAI's GPT-6 Astra
29:22 Is it AGI
31:28 Why AI’s computer use was so bad for so long
35:11 Astra on the benchmarks
39:25 Loop transformers are more efficient but a safety risk
41:02 ChatGPT tries to looksmaxx Anika
41:51 Apple’s September event
43:20 Maya wants a Duo (or does she??)
48:01 Coding from your phone is here
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
1
7
2,315
The FDA sat on proof that biotech and pharma companies were misleading investors about their drug rejections. Vinay Prasad told the FDA commissioner to publish everything.
"FDA has all this data where we can't go and say, this company that's badmouthing us, they actually have all these problems in the dataset"
"We started issuing CRL letters, which is our rejection letter. We made that public."
"Every lawyer there says, don't do it. They told Marty, the commissioner of the FDA, you personally will face litigation if you do this."
"I told Marty one day, just do it. What are they gonna do to you? Nothing."
"He did it one day and we dumped 300 letters. All these investors downloaded them."
"The industry does not want it. They have lobbyists trying to fight this. The investors do want it."
.@US FDA is pausing CRL releases for drug apps as it seeks to create legal foundation for “proactive disclosure of complete response letters.” Publication of proposed rule an attempt to clean up mess created by 'radical transparency' approach of Marty Makary, says @steveusdin1
1
4
543
Get the Check retweeted
have a super cool tech week event planned for Saturday 10/10 and still finalizing a venue if any office spaces are free!
great way for a company to get exposure to early career talent, would need a good space to host a panel :)
can give more info in dms
1
1
11
1,272
Should Instinct charge a subscription?
Anika compares it to Raya:
"There's an interesting psychological thing where when you get something for free, you don't value it. That's why a lot of platforms charge you something small."
"When you get on Raya, you feel like you've been let into something really exclusive, and your willingness to pay skyrockets. Everyone on Instinct right now would pay for it."
Maya puts her ex-Stripe hat on:
"I actually think it's a credit card play."
"If you can wire all your purchases through Instinct because it's easier, they charge you 1%, or they charge the merchant 1%, which is traditionally how it's done."
"I think they stay free for a very long time and don't worry about any cost barrier."
Instinct is not free. They may not charge you directly, but I’ve spent $1000+ I wouldn’t have otherwise
2
28
13,194
Get the Check retweeted
Anika and Maya’s AGI is never having to login to something again
Anika: "I spend a shocking amount of time trying to log into things. I never know my password. It's somehow never saved, even though I click save every single time."
Maya: "Any password manager that asks to save your password, you're like, 'Yes. Save it, because I will never remember it.'"
Anika: "I say yes, save it, and then it saved the wrong password. And it probably had the right password saved before."
Maya: "I've always forgotten my password, and then I just don't do it. 50% of my tasks, if it's behind a login, it's gone."
Anika: "It's always like, 'No worries, you can recover this password. Text sent to Priya.' And it's 11:30 PM, so she's asleep for the next eight hours. By the time it's the next morning, my whole day has started over. At that point it's not gonna happen for another six months."
Instinct + 1Password
32% of Instinct users already securely share their credentials with Instinct through our Vault. Today, we’re excited to announce our partnership with @1Password . Together, we are working with 1Password to roll out a product integration that enables users to seamlessly and securely share their credentials with Instinct.
Our users already use Instinct’s Vault to:
- Log in to loyalty accounts to book flights and hotels with points
- Manage streaming, software, and other recurring subscription accounts
- Book or cancel restaurant, fitness, and event reservations
- Sign in to healthcare portals to schedule and cancel appointments
- Easy access to tax, payroll, benefits, and expense management services
Together with 1Password, we’re working toward a future where people can share their credentials without compromising on security or control.
We’ll be rolling this out across our early access group.
If you’re not already in the platform, join at app.instinct.com/login.
More details about 1Password here:
1password.com/
1
6
1,809
Get the Check retweeted
The SCET Alumni Association is officially live! 🎉
Thanks to everyone who joined us in SF for our launch mixer featuring a live podcast with @linear COO Cristina Cordova 🎙️
SCET alum? Reconnect and join today: na2.hubs.ly/H07C46r0
#ucberkeley @AshurstPerkins @Getthecheckpod
2
4
137
