patrick wingo
577 posts
patrick wingo
@patrick_wingo
ex-palantir. atlantan. product @ https://t.co/1eGfYUKnLt
Присоединился Mart 2021
247 Подписки367 Подписчики
Perplexity Computer now runs on your portfolio.
Connect your brokerage accounts securely through @Plaid, then ask Computer to make a personal terminal that's always on.
English
@burkov I love your writing - couldn't put down the 100-page books, even when I was supposed to be "out" on my honeymoon. Out of curiosity, what are you building?
English
I've been coding almost 24/7 with LLMs for the last 4 months (my codebase just reached 1M tokens, so Gemini no longer accepts the full dump of the app's code), and so far I haven't encountered a problem that, despite my efforts to make the LLMs fix it, none of them succeeded.
Twice, I was close to this, but the solution was to just reimplement the buggy part of the application from scratch. It was annoying to accept, but it wasn't a showstopper.
I'm not saying that this is how you are supposed to build software. I'm saying that *anyone* seems to be able to build *any* app this way if they are persistent and creative enough.
English
> 85% user satisfaction became 67% more revenue (no citation)
> it made up implementation timelines and a project plan for a product when there was no supporting information for it
👉 Great draft aid, terrible final authority. Always verify.
English
Tried o3 for “please summarize this 30-page report to explain this product.” It hallucinated numbers and arguments with completely made up case studies for the product. 😬
Nabeel S. Qureshi@nabeelqu
If you’re using o3 for work I strongly recommend double-checking every number it gives you
English
Studying anon accounts I admire
No bio, no motives—classic projection halo
I fill the gaps and think “wow, genius”
Even more respect for real-name posters who hit the same bar—skin-in-the-game brilliance.
English
@marisbest2 You’re right that procurement and integration is broken and point solutions don’t solve the real problem. Alt. approach might be to fix the procurement process to make it both rigorous and fast, and let that efficient matching force better product. Moving towards building that OS
English
Agree with much of this, but Palantir-esque services+product vendors will avoid the pitfalls of point solutions and enable systems / payors to build what they need cheaper and faster than procuring vendors and then needing to integrate those anyway
Dhruv Vasishtha@dvasishtha
Fundamentally payers, health systems, and EHRs don't have the talent to handle agent management to enable the quality control they will want. Maybe that problem will be solved by their cloud provider or Palantir but it'll be more expensive and slower vs procuring vendors.
English
@yishan @wiesbauerMD CT contrast rarely ruins sleep long-term, but one possible mechanism: iodine surge → iodine-induced hyperthyroidism (Jod-Basedow) → jitters + insomnia. To test, ask for TSH/free T4 labs; treat the thyroid and sleep usually rebounds. Not medical advice.
English
@wiesbauerMD If you can spend some time thinking about this, even if it is at the level of hypothesizing in your spare moments about potential mechanisms or tests that can be performed to rule out certain possibilities, there are many people who will be grateful for your engagement on this.
English
Just so everyone knows, there is a side effect of CT scans with contrast where you stop being able to sleep for YEARS.
The doctors just ignore you because it’s “not supposed to happen.”
Céline@healingfromlc
How the actual FUCK is it possible that dozens of people I am talking to completely stop being able to sleep without like 3-5 heavy meds after a CT scan with contrast and nobody knows it’s a thing? This is everyone’s worst nightmare!!!! What is going on here?!
English
@tlrdrkmrgn @BillAckman @SpaceX @xai @anduriltech @elonmusk @PalmerLuckey @Albedo @TopherHaddad Not to mention it’s the only commercially available source of 10cm visual images combined with 2m thermal images.
English
@BillAckman @SpaceX @xai @anduriltech @elonmusk @PalmerLuckey Great idea! I would invest. You should consider adding @Albedo into the tech stack mix. @TopherHaddad and his team are doing incredible work by bringing down the cost of 10cm resolution VLEO satellite imagery by over ~133x (sat build cost down from $2B each to $15M each).
Topher Haddad@TopherHaddad
This past weekend, @Albedo turned 4 years old. 5 years ago, President @realDonaldTrump tweeted a satellite image that - based on its 10 cm resolution - experts concluded was from a top secret reconnaissance satellite. 2 years ago today, that same image was officially declassified by NGA. People speculate that those satellites cost taxpayers $1B-$2B each. In 4 months, Albedo’s first satellite Clarity-1 will launch into VLEO and capture 10 cm resolution. Our Clarity satellites are $15M each to build. In 2 years, the target for completion of @DOGE, I'm hopeful that Albedo's innovation is being leveraged to its fullest - for our Nation, our Allies, and our tax dollars 🇺🇸
English
Putting out fires before they spread does not seem like a serious technological challenge in a world with @SpaceX, @xai and @anduriltech.
So @elonmusk and @PalmerLuckey, why don’t we start a company that uses satellites, AI, and drone technology to put out fires before they spread?
One can envision satellite monitoring, rapid drone investigation, and drone swarm water delivery. The number of drones sent to extinguish the fire and the nature of the extinguishing agent (water, foam, dry or wet chemicals etc) would be a function of the nature and scale of the fire. No human lives would be put a risk and there is no risk of DEI involvement in drone selection.
What do you think? I would love to invest.
And it would be very good for real estate values in California.
English
@nikillinit First with the sous vide steaks and now with the oysters- why you tryna attack my favorite foods!
English
As I’ve gotten older, a few things have tilted in my risk-reward assessment and are no longer worth doing. Curious what things have changed for you all.
Honestly I think raw oysters have officially into the “not worth it” territory for me - I’ve never loved oysters and feel like vibrio outbreaks seem to be getting worse with climate change
English
Much like in large software projects, institutions encode logic that might make sense at the time, but as times change, they become outdated. Still, they're hard to change because no one remembers why we added it in the first place or knows what would happen if we remove it.
Quite Interesting@qikipedia
From 1264 to 1827, anyone who received a Master of Arts degree from Oxford had to swear that they would never forgive Henry Symeonis. By 1608, no one knew who he was or what he had done.
English
@traestephens Zuck is largely apolitical with Meta decisions / wants to win / this take has so much explanatory power
English
Trump’s anti-TikTok-ban argument has been that it would turn over too much power to legacy social media platforms that censored conservative voices.
Seems like Zuck just massively increased the probability of a ban, which is great for America, but also very very good for Meta Platforms.
English
sam thinks before he talks and he definitely thinks before he tweets. this is a huge flex for the quality of the offering and a play to bring on more subscribers- because while it's likely true, they're also halving their compute costs every 6 months
youtu.be/SImm15uF_3Q?fe…
YouTube
Sam Altman@sama
insane thing: we are currently losing money on openai pro subscriptions! people use it much more than we expected.
English
@MabreyTed Palantir is the first job for many and they don’t know how high the bar is set until they’ve left. It provided
- access to work on huge problems at the enterprise
- product leverage to solve them
But Palums who stay long enough to learn the trade become absolutely cracked!
English
All this Palantir alumni love is great…but I want to make sure hobbits know you can come back to the OG shire too.
Boomerangs have been popping recently and we are into it if you want to do it for the right reasons (let’s chat if this pulls at your heart strings).
And to be clear, you didn’t “miss it” by leaving. It’s still right here and as much at the very beginning as when you left. We have a fundamental insight with AIP akin to the “decisions not data” secret that made Foundry so special that we are just beginning with. We have infinite space to build the future of the product into. We have access to virtually every single one of the most important problems affecting society. We will be a trillion dollar company.
The next phase is going to be the fun one that dwarfs all the others. Come back into the journey to Mordor.
Sebastian Caliri@SebastianCaliri
Interesting observation!
English
@packyM This article was absolutely based. Great job, Packy.
English
Aggregators won the Internet era by owning demand and not owning supply.
The exact opposite model will win the Techno-Industrial era.
Vertical integration is back.
English
@mttgrmm Campus 805 is a lot of fun if you have some time in HSV! Also Gold Sprint is a great coffee place.
English
Kinda an inauspicious corporate sponsor but nonetheless we have made it to space camp
Matt Grimm@mttgrmm
I'm taking my oldest to Space Camp today 🧑🏼🚀, and we're teaching some independence by giving him the mandate to navigate us through the airport, make a connection, getting lunch, etc... He's in charge! We'll see what city we end up in 😂
English
@dvasishtha Definitely, but you could also extend to point-of-care CDS and help clinicians perform better. I think CDS / clinical summarization / AI scribes will bundle together over time.
English
If we see commoditization of ambient scribing coming so soon then I wouldn't be surprised if we start seeing the scribe businesses starting to acquire tuck-in AI capabilities in order to build on top of their feature set.
The obvious place would be HCCs / autonomous coding.
English
I'm going to be very curious how difficult or easy it will be to transition from Nabla to Abridge at Kaiser.
If it's difficult then you have to assume politics was involved in switching from a tool physicians were already finding value from.
If easy then you have to worry that low switching costs for ambient scribes will commoditize pricing.
English
@bryan_johnson @DavidSHolz Can you also run the numbers on optimal temperature / time effects? Lot of variables to measure, could be worth recruiting some additional folks
English
@DavidSHolz same and why I'm thinking of doing a self experimentation study myself
English
Sauna is not currently part of my routine as there's not great evidence in healthy people. It seems great if one cannot exercise due to injury or illness. Unclear if sauna benefits stack on top of exercise. I may experiment with sauna to explore whether it can improve my cardiovascular health as measured by ultrasound. What's your take?
Tim Ferriss@tferriss
Why Dr. Peter Attia (@PeterAttiaMD) Changed His Mind About Saunas👇
English
@rargulati @GPaullJr Y’all keep doing your thing and we’ll keep building in ATL.
English
SF: a boring, expensive city where you have to actively find ways to entertain yourself, and the default activity is to work on a difficult problem / mission to impress your peers and get invited to cool things.
NYC: an exciting, expensive city where you actively have to fight against things constantly happening to focus on your mission, but get invited to things regardless of what you're working on, as long as you went to the right school, belong to the right peer group, or family. Or are hot.
Silicon Valley: a boring, expensive suburb where you can only work and build. Where many of the greatest companies of our generation exist or were started.
English
We spend a ton of time writing this newsletter every week. Definitely give it a shot if you want to hear our nerdy takes!
Dhruv Vasishtha@dvasishtha
The Elion newsletter is one of the best resources I've seen to keep up with all things healthcare AI. In particular their interviews with buyers on their current initiatives and asks from vendors is something I haven't seen from any other healthcare newsletters
English