Dan retweeted
Tesla has released new FSD (Supervised) safety data.
Miles per major collision:
• Teslas Driven With FSD (Supervised): 6,006,200
• Teslas Driven Manually (Active Safety Features): 2,031,221
• Teslas Driven Manually (No Active Safety Features): 845,915
• U.S. Average: 660,164
Total major collisions:
• Teslas with FSD (Supervised): 1,579
• Teslas driven manually with Active Safety: 17,775
• Teslas driven manually without Active Safety: 239
• U.S. average: 4,989,713
Total miles logged:
• Teslas with FSD (Supervised): 9,483,790,215
• Teslas driven manually with Active Safety: 36,104,958,020
• Teslas driven manually without Active Safety: 202,173,600
• U.S. drivers overall: 3,294,031,000,000
Tesla vehicles with FSD (Supervised) engaged experienced ~9x fewer major collisions than those driving without it on.
This is North America data across all road classes (highway and non-highway). The data reflects a rolling twelve-month aggregation of miles and collisions.
Dan retweeted
Grocery stores make about $1 in profit for every $30 of goods sold. So if this person walked out carrying about $75 in groceries, that instantly erases the profit from $2250 in sales. Something to consider as you complain about grocery prices.
Dan retweeted
every generation thinks the world is going to end. it hasnt ended and it isnt going to end. you are a narcissist
Dan retweeted
Creating net-new gold with alchemy is becoming real. The lab grown diamond industry is growing rapidly.
But you can’t invent anything that will increase Bitcoin supply. There will only ever be 21M.
Dan retweeted
The degree to which we piss away our human capital is shocking to me. It's bad enough that we haven't solved aging but society is increasingly dependent on transfer payments via taxation from a hyper productive outer crust. Now, more than ever, we need to be identifying our future captains of industry and training them. Instead, we run schooling into our mid 20s as daycare with no real stakes or consequences.
Dan retweeted
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
Dan retweeted
A list of stats from the Anthropic S-1:
- NDR is ~900%
- Revenue per employee is over $6.6M, up from about $1.4M a year earlier
- Revenue went from $41M three years ago to $16.3B in the first six months of 2026
- Quarterly revenue went from $4.7B in Q1 to $11.6B in Q2
- Q2 2026 was the first profitable quarter: $974M in operating income ($1.7B operating loss the quarter before)
- Adjusted operating income was $1.6B in the first half of 2026, up from a $1.8B adjusted loss a year earlier
- Revenue grew 1,368% year over year in Q2, while compute and infrastructure costs grew 300%
- Compute went from 251% of revenue to 48% in five quarters
- The value of Anthropic’s own internal Claude usage grew 250x from January to September 2026
- Anthropic spends more on tokens (at list price) than on payroll
- Over 1,900 customers pay $1M+ a year, and about 250 pay $10M+
- Customers paying $100K+ went from 950 to over 11,900 in a year
- Over half of the Fortune 500 pays $100K+ a year
- Those Fortune 500 customers still spend less than 0.1% of their total expenses with Anthropic
- The top 5% of Claude Code users spend roughly 20x what non-users on Claude Enterprise do
- About 40 million smaller customers
- $72.4B of the $80.7B first-half net loss is an accounting charge on convertible notes (not operating losses)
- Nearly $800B in compute commitments over 10 years
$141.1B raised in total, with over $130B in cash and short-term investments
- Nearly half of revenue comes through Amazon, Google and Microsoft’s cloud marketplaces, all of which also compete with Claude
- A single share held by the seven co-founders carries 50.1% of the vote
- That share is redeemed for nothing if fewer than three founders remain
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
Dan retweeted
a very simple solution to virtually every problem would be to stop allowing people who live off taxpayer money to vote
anyone whose entire livelihood depends on the government taking money from other people will always vote for a government that takes more money
we’ve built a system where people can literally vote to give themselves more of other people’s money
the conflict of interest is beyond ridiculous and if you remove the incentive you remove the problem
The point of no return by Richard von Mises is worth reading:
"A state becomes almost unreformable when the majority of the population is financially or structurally dependent on the state (state employees, civil servants, recipients of subsidies, grants, or transfer payments, etc.). As soon as those who net benefit from or are supported by the state outnumber those who do not, the system blocks reforms, since the majority votes out of self-interest for the preservation or expansion of state services."
Does this sound familiar? Some European countries, France, Belgium, Finland are already way beyond that point, closely followed by Germany, Spain, Austria and Italy. It works for a while, as long as markets buy more sovereign debt at reasonable interest rates. The party ends when they stop though. For France, they stopped this month, Belgium will be next.
The longer the socialist illusion persists, the more painful the inevitable cuts and adjustments will be. It’s pure arithmetic. Good luck Europe, good luck Euro. Buckle up, it won’t be pretty.
Dan retweeted
A jet engine that has spent 10 of the last 30 years in the air.
The fan blades spin ninety-four times a second, the blade tips at 1,070 miles per hour.
28,800 g's of centripetal acceleration
For a decade.
Fewer countries have built a commercial airliner than gone to space.
Dan retweeted
not only am i not worried about the crypto community's ability to handle the upcoming ai/quantum threat, i will venture to say that in the next 2 yrs the world will realize crypto is far better prepared for it than the rest of the internet and finance. we've had a trillion dollar *irreversible* honey pot for a decade, producing a far more battletested system and adversarial security culture than pretty much anywhere else on earth.
Dan retweeted
In 1999 @ESYudkowsky weighed in on Y2K bug, predicting “major catastrophe” (also: the singularity by 2025!) diyhpl.us/~bryan/irc/extropi…
Dan retweeted
Peak "Bitcoin is dead" Google searches:
- November 2022: Bitcoin $15,599
- June 2026: Bitcoin ~$58k
- October 2026: Bitcoin ~$80k
(highest since FTX)
Dying at a higher price every time
Dan retweeted
The older you get the more you realize that politics are psychology: the vast majority of people believe what they do as an extension of their ingrained personalities.
They don’t reason from first principles, or adjust policy based on data: they’re woke lefties or tradcons or AI doomers or accelerationists because the political shoe fits the mental foot. Few ever change sides, and if so only after contact with the results of their politics (and usually not even then).
Because to do so wouldn’t be updating an opinion, but changing their entire self, and nobody intentionally casts themselves into that confusing abyss.
Dan retweeted
this is gold's quantum FUD but it's way worse because you can't software update gold.
Last month, we demonstrated high-rate alchemy: turning mercury into real, stable gold, beating CERN’s record by a factor of 10.
Last week, we enriched isotopes in the world’s most powerful plasma centrifuge, beating the Soviet record that stood for 40 years.
Now, we’re building a fusion reactor in San Francisco to achieve record power density.