Anthropic day 2: down 10% after the biggest IPO ever
$83 IPO price
$111.50 open
$97.17 close, up 17%
~$87 this morning
The real test is earnings in ~3 months @AnthropicAI
https://t.co/QgL9ftjzvS
#Anthropic#IPO
ANTHROPIC RESEARCHER SAYS CLAUDE PROVED THE RIEMANN HYPOTHESIS
Anthropic $ANT mathematician Levent Alpoge said Friday that he used Claude to prove the Riemann Hypothesis, according to AFP.
"I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis," he wrote in a four-page note.
The claim has not been independently verified or reviewed by outside experts. He released only a Lean version, which a computer can check. A full written argument would be extremely long, he said.
The 1859 hypothesis about prime numbers is one of seven Millennium Prize Problems, each with a $1 million prize. The Clay Mathematics Institute requires peer-reviewed publication and two years of scrutiny.
Alpoge acknowledged that the timing could be viewed cynically.
One agent alone cuts corners. A thousand agents start a journal to catch each other cheating. So they reinvented academia.
This is a really interesting research ↓
It’s your tenth AI approval request of the afternoon, and you click approve. “Technically there was a ‘human in the loop’, but was there really?”
Corné Potgieter draws on Kahneman’s System 1 and System 2 thinking to show why oversight fails, and how to design loops that work. https://t.co/nvjjOAOZel
How hard it is already to deal with all the data about our health, so imagine the difficulties we will face when the same data will be used not only to tell us what we might have but to forecast what we could have in the future.
It would be a new level of health anxiety.
Researchers just analyzed 2.78 million laboratory measurements from 3,905 cancer patients, using AI to predict 162 possible treatment-associated complications over a two-year period.
The study that made me think about this: https://t.co/awkMFKOGkJ
The people building AI agents and the people doubting them have never disagreed this sharply.
LeCun’s position is that agentic systems on top of LLMs will fail because the models cannot predict the consequences of their actions. OpenAI is betting the opposite, that chain-of-thought models like o1 can learn to plan, and Andrew Ng pushes the industry toward agentic workflows right now.
pages
duke
Each side has a real roadmap behind it, so founders get to test the question in their own product. start by listing every action your agent takes and marking the ones a human could undo in one click, because those are where the autonomy belongs today
Our scheduling agent hears procedure names on half its calls, 'my colonoscopy', 'the MRI with contrast', 'my echo'.
Scribe v3 Live Medical goes on it today.
Britain's relationship with emerging economies should be built around trade, investment and long-term commercial partnerships.
Development finance can help unlock projects that create jobs and infrastructure while opening opportunities for British exporters.
That is how we build sustainable economic relationships.
This is very very sweet. We think of capital partnership as just one thing (and honestly the most infrequent) that we do at 1517. We also have lots of hand written letters and actual art people have made for us. Interesting to see this video as a reflection on the journey. Thanks Leo! (You can tell the robots, "She watched it!")
MMFlash: Faster Vision-Language Decoding
Lossless speculative decoding speedups up to 3.58x for VLMs by focusing on visually dependent tokens. Runs with SGLang and vLLM.
A GIANT #HealthcareAI FabCollab ... ⤵️
Apple isn't launching a health chatbot; it's launched a 200K person survey to find out if we should trust one.
Yesterday, @Apple announced a virtual research study run with clinical AI platform @OpenEvidence for up to 200,000 adults across the US.
Participants use a dedicated iPhone app to ask everyday health questions and receive answers grounded in peer-reviewed sources -- think NEJM, JAMA, and Nature -- with linked citations behind each response, and can choose to share HealthKit data such as resting heart rate, sleep, or lab results to see how personal data changes the answers.
The study measures whether people TRUST the responses, understand them, and act on them ... including whether the AI helps them prepare for a doctor's visit.
Apple stressed data will not be used to train AI models, and the app is not meant to diagnose or treat any condition.
🤔 My take ... this study ask what the whole industry has been avoiding, nobody is asking whether healthcare dedicated chatbot answers help or harm.
This is one to watch, closely ... 🔎
🔗https://t.co/TarZhM3ClO
#ResponsibleAI