@wayneb
Account based inUnited Kingdom
About this account
- Account based in
- United Kingdom
- Connected via
- United Kingdom App Store
Account-level information from X, not a live location or the device used for a specific post.
Building and experimenting with AI Agents
Joined April 2007
- Tweets1.6K
- Following759
- Followers279
- Likes519
🚨After another S-1 Leak...we know that Anthropic's revenue jumped from $1B to $16B in 1 year 🤯
But the spending is eye watering...
They spent almost $9B on compute in just 6 months (more than all of last year.)
And they've got about $450B of cloud deals locked in for the next 7 to 10 years, most of which can't be cancelled.
The scary $80B loss is mostly an accounting charge but the compute bill is the bit that's real money.
It looks like Anthropic is betting everything that demand keeps exploding and so far it has.
If the IPO gets anywhere near $2T, that bet looks genius but if it doesn't, they're stuck paying for it anyway.
Who is buying on day 1?
SpaceX's AI business just flipped to profit last quarter
They managed $431 million of operating profit
(that was after a $1.26 billion loss the quarter before)
Spending is also looking good...
In Q2 it spent about $6 on AI for every $1 of AI revenue,
but in Q3 it was closer to $2.
The payback period they are seeing right now is pretty impressive with their CFO confirming its under a year 🤯
Q4 capex is guided at $35 to $40 billion so we'll find out fast whether that holds
Who wouldn't vote for JB Straubel, he is a super smart guy and I love what he created with Redwood materials
Tesla shareholders: Our Annual Meeting is coming up on Dec 3rd
This year, we’re asking you to vote “FOR” all three proposals on the ballot, including the re-election of Elon Musk, Robyn Denholm and JB Straubel to Tesla’s Board.
Last year, we laid out our vision for building a world of amazing abundance – this year, it’s been coming to life. And we’re only getting started…
Your vote is critical to helping us maintain that momentum.
Voting instructions & control numbers to follow soon.
More info:
VoteTesla.com
There's a new version of this post
The EU is fast-tracking approval for “Project Bromo”, a satellite JV between Airbus, Leonardo, and Thales and its framed as Europe’s answer to SpaceX
The reality check is that these are not realistic near term competitors
Bromo is mostly a manufacturing consolidation of three traditional satellite players. It does not design, build, launch, and operate a mega-constellation the way SpaceX does.
Three separate companies with different national interests, shareholder bases, union pressures, and decision-making cultures now have to coordinate on strategy, capital allocation, and execution.
SpaceX has one roof, one vision, and control of the full stack, including the rockets.
That last part is decisive. SpaceX controls the economics of getting mass to orbit. The Bromo partners do not.
Even if they eventually supply satellites for Europe’s much smaller IRIS² constellation (hundreds of satellites, government-focused, years away), they are still dependent on someone else for launch cadence and cost. Starlink already has 11,000+ satellites in orbit and millions of users.
Consolidating struggling GEO satellite businesses into a bigger European manufacturer is industrial policy.
Calling it a SpaceX rival is mostly marketing
Anthropic turned Claude loose on quantum computing, and crypto is paying attention
Quantum computers are a new kind of machine that could (one day) crack math problems.
Today they're small and fragile, so every step a program needs really counts.
Anthropic gave Claude a one paragraph brief, plus a program that checked and scored every attempt, and let it get on with it.
One session went 53 hours, with humans only nudging it to think bigger now and then.
The job was redesigning the core step a quantum computer would use to break "elliptic-curve" cryptography.
That's the math protecting your Bitcoin wallet. Your public key is visible to everyone, and working back to your private key is meant to be impossible.
However, a big enough quantum computer could do it.
Claude's new design needs 17x fewer operations than the smallest one published before, for a few more qubits. Fewer operations means a faster attack.
Normal computers have a neat shortcut for this step, but it never worked on quantum ones because they can't throw away their scratch work.
Claude figured out the numbers could carry their own history, proved some new math to back it up, then checked it with a formal proof.
Anthropic says it took research teams 8 years to shrink a famous chemistry circuit. This came from one 53 hour run.
And it started out too cautious, kept triple checking itself until they pushed it to be bolder.
Before anyone panics:
No quantum computer today is anywhere near big enough to run this.
Nothing ran on real quantum hardware, it was all designed and checked on normal computers.
And an open leaderboard run by crypto researchers already shows other AI-assisted designs beating some of Anthropic's numbers.
So it's AI racing AI to make this attack cheaper.
Your coins are fine this week. But Bitcoin's quantum-proof upgrade just got more urgent, and I'd much rather it moved early than late.
Begs the question, is crypto moving fast enough?
Podcast Clipper for X is live.
Tested it on today’s BG2 with Jensen.
This clip is one of twelve it cut, captions burned in, drafts ready to tweak.
x.ai/bot/9Lqx9JK4j4sGUwPBr2O…