Tech optimist, recovering former CFA, livestreaming @AI_in_the_AM, future shocked self-aware neuron, once fooled by superconductors;
Join the community 👉
Joined May 2021
- Tweets27.2K
- Following3.3K
- Followers58.3K
- Likes12K
🤔 Fil is building real manufacturing know how… but I guess the average retiring factory owner just wants to get paid a bunch to go play golf in Florida.
How much tacit knowledge is in the emails, the maintenance schedules, employee rosters, quality control checklists of a business ?
Can AI absorb all that?
Sometime next year the labs will sell entire business management swarms…
Varying degrees of Amish is the future we are all resigned to. Because there will always be a limit to how much tech you can absorb and still relate to other humans.
xitter.cf/Hesamation/status/2114…
We should anti-doom pill.
That reminds me, I’m going to do an anti Doom Debate with @liron
The past and the future
$MSFT offers 60,000 employees a voluntary exit.
60K Y’all!!
First time this has ever been offered at Microsoft and it will affect employees in India, The United States, Canada, and the UK at level 70 or below…
Employees find out in a few hours if they are being targeted for this.
This just makes me think that H127 layoffs sound intense. Is this a pre-layoff round?
Also makes me wonder if they changed some scheduled layoffs due to be called out by @VP. Peace offering? 🤔 Timing is just interesting.
More to come.
Prakash retweeted
What?! P might equal NP after all...
The computer science collaboration will understand and explain a result concerning the Exponential Time Hypothesis (ETH), a longstanding conjecture about the computational complexity of Boolean formula satisfiability, a problem deeply related to the P vs. NP problem.
Surprisingly, our system has recently proved that the ETH is false, a result that was subsequently verified in Lean.
The refutation of the ETH reopens questions about how much faster these tasks could become, and may help guide the search for better algorithms. Finding such algorithms, and turning theoretical advances into practical improvements, will require significant further research.
Our system found its natural language disproof for the ETH we reached out to to @rrwilliams. Ryan will form a collaboration with Russell Impagliazzo and Ramamohan Paturi to verify and shepherd this result. Impagliazzo and Paturi, along with Francis Zane, are the original proposers of the Exponential Time Hypothesis, and Williams has refuted strong versions of the ETH, used ETH for hardness results, and connected algorithms and lower bounds.
This group asked us to make all raw data of the results public to facilitate independent verification by the community, which we have done by posting the natural language proof and Lean formalization to our GitHub repo. The group also invites the community to help organize an open examination of the ETH claim and to discuss the broader questions that AI-generated breakthroughs raise about verification, access, credit, and the future of research.
github.com/openai/math/blob/…
One way Anthropic could do AI safety is by having its employees use their IPO proceeds to buy up all the real estate is San Francisco thereby making everyone else homeless.
Checkmate
So that 80 billion loss is how much Amazon made?
Jassy put in 13 bil and made 190 bil.
And that’s after first dollar in Sept 2023, almost one year after ChatGPT… and while Claude was still in its lying liar phase.
Anthropic just reported an $80.7B loss for the first half of 2026.
Revenue was $16.3B. The operating loss was only $693M.
So where did the other $80B come from?
A few pieces of paper Amazon bought in 2023 and 2024.
Rewind to September 2023.
Anthropic was valued around $4B, a distant second to OpenAI.
Andy Jassy's Amazon wanted a frontier model of its own on AWS.
On Sept 19, 2023, Amazon signed a convertible note deal with Dario Amodei's company:
- $1.25B up front, another $2.75B in March 2024
- 8% interest, 10 year maturity
- Convertible at $20.93 a share
In November 2024, Amazon came back with $4B more:
- Paid in three pieces through November 2025
- 8.5% on the first piece, zero on the rest
- Convertible at $50.48 a share, a 10% discount to the next round
Then the valuation went vertical:
- $61.5B in March 2025
- $183B in September 2025
- $380B in February 2026
- $965B in May 2026, at $589 a share
Now do the math.
A note that converts at $20.93 into a $589 share is worth 28x its face value.
The 2024 notes are worth almost 12x.
Accounting rules make Anthropic carry these notes at fair value.
Every quarter it asks what Amazon would get if it converted today.
Any increase goes through the income statement as a loss.
$6.2B of principal is still outstanding:
- Carried at $41.7B on Dec 31, 2025
- Carried at $108.6B on June 30, 2026
- That $72.4B jump is most of the loss
Another $8.8B came from booking new funding deals signed with Google and Amazon this spring.
Same story in 2025: $34.3B of a $42B loss.
So why are the notes still there?
Amazon has an ownership cap.
Each new round triggered conversion, but anything above the cap rolled into a fresh zero interest note.
Then the next round marked those up too.
Strip the marks out and you see the real business:
- Q2 2026 operating profit of $974M, its first ever
- Positive $3.9B cash from operations for the half
But somebody does pay for this.
At the IPO, the notes convert into shares for Amazon.
Every other shareholder gets diluted.
Amazon has put about $13B into Anthropic.
Its own books now value the stake at roughly $190B.
Possibly the best trade of the AI era.
xitter.cf/hedgiemarkets/status/2…
ChatGPT ads just went into wide release worldwide. They will get the Q4 metrics for the holiday season, and go public next year as the consumer AI company. So perhaps end of Q1 IPO if the results are good.
It says “finding” (meaning not found yet) or “expansion” (meaning not under current constraints).
Ie the law as it is right now is fine, but S it maybe in the future may not be
2027 is the year of company in a box. You will be able to buy say a dropshipping company, add capital and an idea, and have it design, build, market, ship, account for the revenue. Small businesses will be the first affected.
My take: SaaS is dead. Software is charity now, make money elsewhere.
No one can make $$ selling the same features for years anymore.
Instead:
- Use good software to collect user data and sell data.
- Use good software to deliver outcome and sell outcome.
- Arbitrage tokens under the guise of selling software - buy bulk, sell retail.
- Use good software internally to drive your own margin higher for your services/products
- Build RSI agent in a domain that has the fastest learning loop so it has the most, latest expertise: e.g. a marketing agent that knows the latest tricks to get viral in X coz it's doing it for others every day and learn and evolve.
- Give away all your features for free, it's futile anyways. Open source them. Make money elsewhere.
obviously tons of military drone startups already have this tech. What exactly is the point of academics developing it? I often wonder how far behind academia is from industry.. in cases like this 3-4 years ?
Purdue University's IDEAS Lab built an autonomous drone that tracks runners through dense forests completely without GPS or cloud networks.
By running 3D LiDAR mapping and obstacle-avoidance AI entirely on-chip, it can weave under tree canopies and predict target movement through blind spots in real time.
Massive implications for next-gen search & rescue and edge AI.
When you think about it… Elon is absorbing an incredible amount of risk for Anthropic. He’s providing them flex capacity while they engage with others for baseload. If things go south on the AI trade, Elon’s left holding the bag first.
Their risk is that he eventually makes enough margin on his own business to take the capacity back.
They can prevent this by carefully hiring people from his team so that he can never build the talent base necessary.
Bearish for both chip firms and hyperscalars in the long run as both get commoditized
“Our Approach to Compute. We manage compute as a single, fungible pool from multiple strategic partners, and we dynamically direct available compute between workloads based on where the marginal unit is expected to be most productive at any given time.”
Ok given China has full access to all frontier lab research. Then every AI doomer and natsec person has to answer:
Why have all the China fears so loudly proclaimed not come true?
