Sabitlenmiş Tweet
Will Reed
1.9K posts
Will Reed
@willreed
gp @sparkcapital
Marin County, CA Katılım Ekim 2015
1.6K Takip Edilen4.2K Takipçiler
Will Reed retweetledi
Abridge: 100M+ medical conversations, real-time prior auth, and the clinical intelligence layer latent.space/p/abridge
@AbridgeHQ is building the clinical intelligence layer for healthcare. In this episode, Janie Lee and @c_asawa explain why ambient documentation was only the first wedge, how Abridge is turning patient conversations into real-time clinical decision support, why healthcare may become one of AI’s most important proving grounds, and how 100M+ medical conversations, specialty-specific evals, and deep EHR integrations create a moat for AI-native healthcare.
English
Will Reed retweetledi
The work to make a clinical answer trustworthy includes figuring out which source applies, when the evidence is weak, and what the doctor is trying to decide. Charlie @oneill_c at @baseten on why AI companies train their own models. Must read.
Charlie O'Neill@oneill_c
English
Will Reed retweetledi
This month we turn 10.
The hard work started in 2016, and it hasn’t stopped.
Shortcuts are for losers. Winners welcome.
scale.com/careers
English
Will Reed retweetledi
"I see endless possibilities...this is the most excited I've been about something brought forward in the profession in my 25 to 30 years as a nurse.” – Misti Foust-Cofield MHA, BSN, RN, VP, CNO, Reid Health
Abridge for Nurses is now live, and Newsweek spoke with health system leaders about why deep collaboration with nursing teams is essential to building trust in clinical AI. Technology works best when it’s designed alongside the people who use it every day. By partnering closely with nurses throughout development, organizations can ensure AI is purpose-built for the realities of nursing workflows, supporting care teams in ways that are practical, intuitive, and meaningful.
English
Will Reed retweetledi
Krishna Rao is the CFO of Anthropic, and this is his first podcast appearance.
He joined the company two years ago when run-rate revenue was about $250M. Today it is $30B. He has helped raise ~$75B and is responsible for the procurement and allocation of compute.
I feel lucky we get to hear what it is like to sit inside a company this consequential at a moment this pivotal.
We discuss:
- The cone of uncertainty
- How he allocates compute across Trainium, TPUs, and GPUs
- What investors misunderstand about model companies
- Why the returns to frontier intelligence keep rising
- Platform vs application and where Anthropic builds its own products
- How Anthropic uses Claude internally
I have asked my closing question about the kindest thing more than 500 times. Krishna's answer is one I have never heard before.
Enjoy!
Timestamps:
0:00 Intro
2:38 The Compute Canvas
6:51 The "Cone of Uncertainty"
11:58 Why the Returns to Frontier Intelligence Are So High
16:45 Recursive Self-Improvement
20:20 Scaling Laws
23:30 Sourcing $100 Billion in Compute
28:05 Platform vs. Application Strategy
32:52 Pricing Dynamics
38:48 How Anthropic’s Finance Team Uses Claude
43:24 Raising Capital & Overcoming Investor Skepticism
52:32 Public Perception, Risks, and Government Regulation
57:25 Mythos Release
1:12:33 What Could Derail the AI Revolution?
1:13:47 Biotech and Healthcare
1:15:31 The Kindest Thing
English
Will Reed retweetledi
when I first heard of @baseten they were basically a competitor
but then I met @tuhinone and I liked him, and when I heard they were pivoting to inference, I was relieved because I didn't want to compete against him
and then he hired @DannieHerz and I was angry because I wish I had thought of it
and now they're a critical partner for us as we embrace our own many model future at @_hex_tech
I'm so happy for all their success and very excited to share what we've been working on with them!
Tuhin Srivastava@tuhinone
English
Will Reed retweetledi
The starting premise @Conviction was that AI (general models at scale) were a broad shift in computing. This has come to pass
But the way AI benefits many users more powerfully is going to be more distributed product/research work, in partnership with humans who do the work
Tuhin Srivastava@tuhinone
English
Will Reed retweetledi
Will Reed retweetledi
Will Reed retweetledi
Krishna on how Anthropic thinks about the platform vs. application layer, and when they decide to build their own products like Claude Code.
It’s the question every investor and founder is thinking about:
“Most of what we're building is platform. There's so many examples of where a platform can accrue a lot of value, but the customers who are building on that platform actually accrue even more value.
We will build our own applications on that same platform where a couple of things are true.
Number one, if we feel like we have a vision into where the models are going and we can demonstrate that and create customer value in that, that might be something like Claude Code.
The second is thinking about ways to demonstrate value for the ecosystem that others might emulate. If you think about Claude for financial services or Claude for life sciences, these are ways in which we've composed the platform.
We're building on the same platform as our customers. That creates a level playing field. We also think that there's so much value that's going to accrue in these areas that our customers can win and we can win as well.
So I think of our strategy as mostly horizontal. A lot of the value is going to accrue to the customers that are building on top of it.
Our goal is build the best models and then build the products and tools and services that allow that intelligence to proliferate within customers."
Patrick OShaughnessy@patrick_oshag
Krishna Rao is the CFO of Anthropic, and this is his first podcast appearance. He joined the company two years ago when run-rate revenue was about $250M. Today it is $30B. He has helped raise ~$75B and is responsible for the procurement and allocation of compute. I feel lucky we get to hear what it is like to sit inside a company this consequential at a moment this pivotal. We discuss: - The cone of uncertainty - How he allocates compute across Trainium, TPUs, and GPUs - What investors misunderstand about model companies - Why the returns to frontier intelligence keep rising - Platform vs application and where Anthropic builds its own products - How Anthropic uses Claude internally I have asked my closing question about the kindest thing more than 500 times. Krishna's answer is one I have never heard before. Enjoy! Timestamps: 0:00 Intro 2:38 The Compute Canvas 6:51 The "Cone of Uncertainty" 11:58 Why the Returns to Frontier Intelligence Are So High 16:45 Recursive Self-Improvement 20:20 Scaling Laws 23:30 Sourcing $100 Billion in Compute 28:05 Platform vs. Application Strategy 32:52 Pricing Dynamics 38:48 How Anthropic’s Finance Team Uses Claude 43:24 Raising Capital & Overcoming Investor Skepticism 52:32 Public Perception, Risks, and Government Regulation 57:25 Mythos Release 1:12:33 What Could Derail the AI Revolution? 1:13:47 Biotech and Healthcare 1:15:31 The Kindest Thing
English
Will Reed retweetledi
Will Reed retweetledi
I'm hiring an Entrepreneur, CEO Office at Handshake.
Handshake's grown to ~$1b in revenue in a year, and we've hired over 25 ex-founders along the way.
I'm looking for an ex-founder who wants to work with me to tackle the hardest problems at Handshake and build 0 to 1 motions in product, GTM, operations, and more.
If this sounds like a good fit for you or someone you know, DM me or check out the role below.
English
Will Reed retweetledi
Infra for running / training long-horizon agents getting better by the day. Congrats to @oneill_c @mudithj & the @baseten team on the launch!
Excited to be partnering with you all
Baseten@baseten
Open-source RL libraries break at frontier scale. We built Baseten Loops to fix this. Loops is a training SDK that takes you from your first RL run to production inference on a single platform: → Async RL so training and sampling overlap → 131K+ sequence length for agentic and long-horizon workflows → One command to promote your model to prod → Dedicated infra for predictable, repeatable performance We're excited to work with @harvey and @EvidenceOpen as early partners. Early access is open today: baseten.co/blog/introduci…
English
Will Reed retweetledi
Congrats to the @baseten team! Has been awesome working with @mudithj, @oneill_c, @mblau_ and team to post-train open weight models for long-horizon legal work. Early results on benchmarks like our Legal Agent Bench are super promising
Baseten@baseten
Open-source RL libraries break at frontier scale. We built Baseten Loops to fix this. Loops is a training SDK that takes you from your first RL run to production inference on a single platform: → Async RL so training and sampling overlap → 131K+ sequence length for agentic and long-horizon workflows → One command to promote your model to prod → Dedicated infra for predictable, repeatable performance We're excited to work with @harvey and @EvidenceOpen as early partners. Early access is open today: baseten.co/blog/introduci…
English
Will Reed retweetledi
Open-source RL libraries break in predictable places. Sampling pauses on every weight sync so agentic rollouts spend more time waiting than training, and if you try to mess with that you end up off-policy and degrading your model if you're not smart about how you manage it. Then your checkpoint lands and you spend two weeks merging LoRAs, quantising without destroying the signal you learned, and reconciling inference engine deployments that look nothing like what you trained against
We want ML teams thinking about data and reward shaping, not GPUs and parallelism strategies. More like painting with a brush than mixing the pigments yourself. Loops handles the rest, and one command promotes a checkpoint to Baseten's inference stack so the thing you trained is the thing you serve. This brings us a lot closer to closed-loop deploy→eval→retrain pipelines next!
Raymond Cano@vim_dzl
English
Will Reed retweetledi
Will Reed retweetledi
Will Reed retweetledi
so thrilled for our nursing team @AbridgeHQ to continue thoughtful AI deployment to nurses across the US
Newsweek@Newsweek
After two years of tweaks and slow-scaling pilots, Abridge is bringing AI to the nurses' stations. #Echobox=1778134418" target="_blank" rel="nofollow noopener">newsweek.com/ai-nurses-heal…
English
Will Reed retweetledi
New Anthropic research: Natural Language Autoencoders.
Models like Claude talk in words but think in numbers. The numbers—called activations—encode Claude’s thoughts, but not in a language we can read.
Here, we train Claude to translate its activations into human-readable text.
English
Will Reed retweetledi
"No post-training Pre-PMF"
Our CEO and Co-Founder @tuhinone sat down w/ @saranormous and @eladgil on @nopriorspod to discuss inference, the compute market, and how the app layer is using RL to win.
youtube.com/watch?v=XAbKfl…
x.com/saranormous/st…
YouTube
sarah guo@saranormous
"if we have all the compute, good luck running inference" from new @NoPriorsPod with @tuhinone, founder @baseten on the new AI compute landscape. full interview linked
English