Pinned Tweet
CEO Founder = allocating time between recruiting:
Talent
Customers
Capital
Best skill for achieving all of the above efficiently?
Great story telling.
Rasool Rayani retweeted
Underrated life advice: Be careful what you normalize around your kids. How you speak to your spouse. How you treat your body. How often you complain. How you talk about others. What feels ordinary in your house becomes how they'll act as adults outside of it.
Rasool Rayani retweeted
Alphabet is sitting on $242B of cash, almost twice as much as Amazon (Save this for later).
Cash and short-term investments at the end of the latest quarter:
1. Alphabet $GOOGL: $242B
2. Amazon $AMZN: $123B
3. SpaceX $SPCX: $100B
4. Meta $META: $90B
5. Microsoft $MSFT: $77B
6. Nvidia $NVDA: $62B
7. Apple $AAPL: $62B
8. Tesla $TSLA: $44B
9. Micron $MU: $43B
10. Oracle $ORCL: $37B
Banks excluded, among 44 of the largest US-listed stocks.
That cash pays for the AI build-out. Alphabet, Amazon and Microsoft all update their spending plans this week.
Everyone is worried about SpaceX funding its $470 BILLION of capex to hit 13 GW
But I think the market is missing how FAST $SPCX revenue is about to ramp up, specifically from AI
My model suggests SpaceX hits $122B of Q4'27 revenue and $314B in FY'27
With $90B of cash on the balance sheet today, this should bring new debt needed to fund everything next year well below $200B (most estimates are $300B+ right now)
I think the big thing the market is missing is that every 2-2.5 GW SpaceX adds in a quarter creates $25-35B of new QUARTERLY revenue at roughly 60% EBITDA margins. We already have proof it converts fast as they guided Q1'27 to $33-37B in revenue off 2.5-3GW
So if we look at funding $470B:
- ~$230B comes from the business itself (operating cash flow over the 5 quarters)
- ~$60B comes from the $90B cash pile they're already sitting on
That leaves $180B to raise in debt
And that debt shouldn't be too difficult for Elon considering SpaceX exits 2027 generating EBITDA at a ~$290B annual pace. $150-180B of debt against that is less than 1x leverage. Most blue chips carry more
And of course, if he doesn't hit the 13GW and revenue comes in lower, it just means the capex is lower too
I continue to believe $SPCX is very undervalued, as I've been saying since I made my first buy at $115. Milk Road PRO members have been getting the signals to buy SpaceX on dips by tracking my real-time portfolio and research. You can find it all here: link.milkroad.com/t5837z
Rasool Rayani retweeted
In a new randomized study, adults with type 2 diabetes were assigned to either:
1) A high-fiber diet
2) Metformin
After 12 weeks, the high-fiber group had better blood sugar control and greater weight loss.
Rasool Rayani retweeted
PREDICTION - Google overtakes OpenAI and potentially Anthropic by year end
Swapping all our Sol and Sonnet workloads with Argon. It’s just better
Their next release Carbon will beat Fable and Astra. It almost seems inevitable 😲
Rasool Rayani retweeted
can AI read a doctor’s handwriting?
i gave 11 open models 300 handwritten drug names
best: Gemma 4 31B, 76 exactly right
a 0.8B OCR model on one NVIDIA GPU: 65
letter by letter, it reads a Civil War doctor’s note at 98%. today’s prescriptions: 71%
pharmacists are safe 😅
Rasool Rayani retweeted
Biggest surprise on a Spellbook sales call this week:
"Can your agent negotiate with the counterparty automatically? We want it to send redlines without human intervention. We think agent-to-agent negotiation is the future."
From a conservative Fortune 500 healthcare company
Rasool Rayani retweeted
Someone pinch me: New Brunswick is the 1st in Canada to do a province-wide rollout of Patient AI for healthcare journeys (e.g. surgery, chronic care.) - and it’s powered by @SeamlessMD!
9 takeaways from the stories that have come out of our amazing partners @HorizonHealthNB and @VitaliteNB leading this rollout:
1/ Horizon Health Network and Vitalité Health Network, the province’s two health authorities, have both launched Seamless Answers: a clinician-approved AI assistant that gives patients answers 24/7, inside the SeamlessMD app their patients already use for their care journey.
2/ Patients across heart failure, heart surgery, orthopedics, maternity, colorectal, gynecologic oncology, pancreas surgery, liver surgery, and hepatopancreatobiliary journeys can now ask a Conversational AI questions like “when can I remove my dressing?” or “is this symptom normal?” and get an instant answer.
3/ Patients were already asking ChatGPT about their care journey- but ChatGPT knows nothing about the clinical pathway their care team put the patient on and could provide instructions that conflict - creating a safety risk. With Seamless Answers, patients receive immediate, appropriate answers.
4/ Already 650+ patients across New Brunswick have asked the AI nearly 2,000 questions!
5/ 65% of patient questions at Vitalité come outside regular business hours, mostly at night and on weekends. That is when clinics are closed and when patients are most likely to feel stuck. Seamless Answers is clearly addressing a need.
6/ For patient safety, the AI uses Retrieval-Augmented Generation (RAG) to answer only from each health authority's clinician-approved education library in SeamlessMD, not the open Internet. Every answer cites its source with a link to the original document.
7/ The AI understands and responds in multiple languages - eg perfectly bilingual English and French for citizens in New Brunswick.
8/ For anything out of scope where the AI does not have relevant content for, it appropriately directs patients back to the care team.
9/ 94% of patients at both Horizon and Vitalité who tried the AI found it helpful.
Thank you to Horizon Health Network and Réseau de santé Vitalité for the courage and conviction to lead the way with clinically appropriate AI for patient care. Grateful for SeamlessMD to be your partner on this journey!
It’s only by partnering with healthcare teams - not circumventing them - that AI can go beyond consumer health education and move the dial on outcomes like reducing ED visits and readmissions.
Rasool Rayani retweeted
Introducing SKY-001, the first Engineering World Model.
On CAD and BIM tasks, it beats GPT-6 Astra and every frontier model we tested:
-> 51-125x faster
- > 150-210x cheaper
-> Higher IOU than all frontier models
SKY-001 lets you dream a physical object before you build it.
18 months of work. Built by the team out of Maluuba's deep learning lab (acq. by @MSRResearch).
The first ‘AI General Engineer’ will be the backbone for Autonomous Manufacturing.
Rasool Rayani retweeted
Every year an investor asks me: “SeamlessMD has incredible results improving patient outcomes. Why the heck isn’t this standard of care everywhere?” I hate this question - not because it’s a bad one, but because it’s a painful one I ask myself privately every single month.
Relative to most Health Tech innovations, we’re doing reasonably well. 50+ studies and evaluations showing reduced length of stay, readmissions, ED visits, mortality, costs, etc. Integration partnerships with Epic, Oracle, MEDITECH. Deployed at many hospitals and health systems, across everything from surgery to oncology to chronic care.
Yet… not adoption at the scale of an Abridge, LeanTaas or Vizai (kudos to them and others who have scaled massively - so hard to do in Health Tech!).
It’s not a single factor, but a mix of the following:
→ Just plain lack of awareness: if a health system IS looking for digital care journeys, are they aware of SeamlessMD?
→ Misalignment with strategic priorities: increasing revenue and decreasing clinician burnout continue to take priority over improving clinical outcomes. It’s not that improving outcomes isn’t on the list… it’s just usually much lower right now.
→ Patient impact is less immediately visible: a physician who uses an AI scribe feels the magic immediately. A patient who goes home with @SeamlessMD and feels the magic of being navigated through their care? No clinician or executive feels that “magic” personally or immediately - it only shows up months later when we demonstrate reductions in LOS, readmissions, etc. - and even then, it doesn’t create a daily magic moment for staff inside the health system walls.
The last one is particularly tough because you could turn SeamlessMD off, and no one realizes the pain - until months later when they measure outcomes again and they’ve gotten worse. Not like an AI scribe where if you turn it off, perhaps a dozen doctors come knocking on your CMIO door with complaints by the end of the day.
So I’ve come to believe what we’re doing with digital care journeys - navigating patients with reminders, education and symptom monitoring for any condition, procedure or treatment - requires us to evangelize a movement.
But I’ve seen too much good this innovation has done to not do everything possible to keep supporting this movement. To push it up the priority list. To ensure every health system who is thinking about digital care journeys hears about SeamlessMD.
I’ve been in this for 13+ years… and unfortunately for healthcare, I still haven’t found something better to do!
P.S. If your health system is looking for EHR-integrated care journeys with reminders, education and symptom monitoring across ortho, cardiology, women’s health, chronic care, oncology - you name it - give us a shout at @SeamlessMD. And if you prefer to use Epic Care Companion? No sweat - we’re Epic’s only Toolbox partner for Care Companion content 😀
Rasool Rayani retweeted
My mind is blown by how far ahead Apple is with local LLMs. The new Mac Studio M5 Ultra will have a 512 GB model for around 15-20k. You can chain 6 of them together using thunderbolt 5 and this have over 3 TB of ram and a massive amount of CPU and GPU cores. This means you can run a frontier model with over 2 trillion parameters locally for around 100kish. Nivida is around 500k for the same capability.
Rasool Rayani retweeted
A consequential choice will be made soon by government and companies as they are increasingly flooded with agentic AI inquiries that overcome the frictions that gated access (see the graphs - it has already started!)
Two options: do you make services worse or better as a result?
Rasool Rayani retweeted
People really underestimate how many creative breakthroughs come from pure brute force search. And AI is very good at tolerating long brute force search (with some intuition)
TERENCE TAO:
“OpenAI did make a major breakthrough [on quasi-Riemann Hypothesis] the biggest advance on the subject in 50 years. A massive new zero-free region, very clever new technique.
Well, it’s sort of a combination of three or four existing ideas, arranged in a very unusual way. It’s not something I think any single mathematician alive would have come up with.
I think if you got the right people to work together and locked them in a room, they might have come up with this after lots of trial and error.”
Rasool Rayani retweeted
Believing difficulties build character is linked to higher self-esteem and meaning in life.
Struggle can be training, not just damage.
Rasool Rayani retweeted
Just a reminder about something I said a few years back: “addiction is a progressive narrowing of the things that bring you pleasure.” “Enlightenment (for lack of a better term) is a progressive expansion of what brings your pleasure.”
Are you addicted to something? If so, what?
Rasool Rayani retweeted
Something deeply troubling I’ve noticed this election is how quickly political disagreement turns into personal attacks.
If you oppose the NDP, you’re suddenly labeled a racist, climate denier, conspiracy theorist, residential school denier, or whatever other term is convenient.
I’ve spoken privately with many people who are afraid to share their political views because the personal and professional risks are simply too high.
Think about how dangerous that is for a democracy.
When one side feels free to speak while the other fears being publicly vilified, we don’t get healthy debate. We get an increasingly one sided conversation.
We’ve reached a point where having a different political opinion is treated as a reflection of someone’s character rather than simply a difference in perspective.
You can disagree with someone without believing they’re a bad person.
You can challenge someone’s ideas without attacking their character.
And you can believe strongly in your own views while accepting that someone else might see things differently.
A healthy democracy requires people to feel free to disagree. Not just those whose opinions are popular.
We need more debate, more critical thinking, and more willingness to listen to opposing views.
Not more labels, division, and attempts to silence each other.
Rasool Rayani retweeted
Angel investor lifecycle:
1. Invest $25k in random friend's company
2. Friend's corp random sells in 1 year, make $500k
3. Think you're a genius
4. Throw $5M at 20 random companies
5. None sell, illiquid nightmare
6. 20 yrs pass, feel like an idiot
7. It all hits, repeat
Rasool Rayani retweeted
You need to be kiwi-maxing.
-Actinidin maxing for better protein digestion
-Beats psyllium for constipation without the downsides
-Mogs depression in 4 days
-Mogs skincare products
-Helps you fall asleep 35% faster
-Supports faecalibacterium prausnitzii (produces butyrate and low levels are closely linked to IBDs).
fitandball.gumroad.com/l/bra…
Rasool Rayani retweeted
We made open models run over 2x faster than Ollama, without any new hardware
Introducing Magnitude, an open source inference engine that runs open models as fast as your hardware allows. It tunes its kernels on your machine to find the fastest way to run each model
Works on Apple Silicon, NVIDIA, AMD, or just a CPU
On average across 11 popular local models:
Mac M4 Pro vs Ollama (MLX) 1.32x faster prefill, 2.19x faster decode
DGX Spark vs Ollama (llama.cpp) 1.34x faster prefill, 2.17x faster decode
Ollama, LM Studio, and llama.cpp ship one-size-fits-all kernels, because tuning each model for each chip by hand would take forever
Magnitude tunes them automatically. When you download a model, it spends up to a minute on your actual chip trying different ways to split up the math (how many GPU threads team up, how big a chunk each one takes, how far ahead to load data) and keeps the fastest combination that doesn't change the model's output
Each model also comes paired with the best speculative decoding drafter, for even more speed
One click connects Magnitude to the agent you already use (Pi, OpenCode, Hermes, Codex, and more). It's also built specifically for agents:
• Each agent uses about half the memory, thanks to a quantized KV cache
• Memory grows and shrinks with your sessions, so your computer stays usable
• Agents share prefix caches, so several can run at once without slowing down
Next, we're working on getting even more out of your hardware:
• Expert streaming from RAM or disk, to run models bigger than your GPU allows
• Splitting one model across all your GPUs and CPU so nothing sits idle
• Attention pruning that skips low-impact math, for much faster prefill
Ships as a desktop app on Mac, Windows, and Linux
100% free and open source
Download → magnitude.dev/download
GitHub → github.com/magnitudedev/magn…
Rasool Rayani retweeted
turns out you can get frontier performance by having a swarm of tiny models working on the problem together.
great work from the research team on this one :)
cohere.com/blog/lessons-from…
Announcing new research on scaling multi-agent teams for long horizon software engineering tasks. 🤖🧩📈
Work led by @keenansamway, @mrdanieldsouza, Omar Darwiche Domingues, @mziizm, @yichern_tan, and @nouhadziri
Read the paper: cohere.link/tStRYJS
Check out the blog post: cohere.link/0TzgKoy



































