"Let's rob and attempt to enslave the most mobile class in human history" 🙄
this California wealth tax thing is just incredible.
so in October 2025, a California healthcare union proposed a "one time" 5% tax on any billionaire's wealth. the ballot measure will be voted on next month. In early December 2025, California announced that it would be a retroactive tax if passed, applying to anyone with CA residency as of Jan 1, 2026. The objective is to pay for Medi-Cal, California's incredibly dysfunctional healthcare system. (which also functions as a glorified welfare apparatus, but that's a discussion for another time).
polls reveal that the CA electorate is split around 50/50 on the tax, but prediction markets only give Prop 40 a ~23% chance of passing (there are competing measures on the ballet that would nullify it).
there are 200 billionaires in CA, worth collectively somewhere between $2.3 and $2.6 trillion. 5% of that is around $100b, accounting for a small amount of attrition. proponents of the tax forecasted a significant windfall, assuming only a 10% departure rate.
supporters thought they were being clever with the retroactive concept. no billionares would leave CA in the few short months between the tax being proposed and the residency cutoff date, without the tax even having passed yet, right?
well as it turns out, they did.
a new paper [1] from two academics at Stanford called "The Net Present Value of the California Billionaire Tax Act", argues that not only did more billionaires leave due to the proposed tax than its proponents expected, but that the value of the tax, if it does pass, is actually negative.
proponents of the tax only assumed 10% avoidance, that is, only 10% of the taxable base of targeted billionaires would leave CA within the very short Oct-Dec 2025 window. Jaros and Rauh, the Stanford academics, find that a lower bound of 28.6% of the taxable base managed to leave.
just based on known departures, those departures total $577b in net worth, or $28.8b in potential tax lost. this of course does not capture everyone (many left without announcing their departure), so the authors normalize to departure rates estimated from other wealth tax events, and apply other discounts for litigation, arriving arrive at a figure of only $28-$36b in forecasted tax revenues.
but this isn't all! all of the income taxes paid by the departed billionaires are also forsaken. the authors estimate that CA billionaires pay $3.3- 5.8b in annual income taxes. they assume a midpoint of $2b lost in annual collections from departed billionaires. the net present value of that lost income stream is -$68.9 billion.
so when you add it all up you get:
+ $30b one-time tax windfall (estimated)
- $69b in present value of lost future income taxes
= - 38.9b overall net present value of the tax
of course, this analysis is highly estimate driven. the authors simulate the parameters 100k times and find a negative NPV of the implemented tax 85 percent of the time.
and if the tax doesn't pass, which it's not looking like it will, California still ended up losing at least $577b in departed net worth, and they have permanently poisoned the well with entrepreneurs who will now worry about future wealth taxes.
[1]: nber.org/system/files/workin…
He-Man, Principal MLE retweeted
Unfortunately Chinese models have too much botnet potential now and are a supply chain risk.
Time for the US Gov to fund open source models to remove this risk. – at San Francisco, CA
The China state-aligned threat actor Storm-0940 used the agentic AI penetration testing framework ARTEX to automate attacks targeting hundreds of organizations globally across a broad range of industries. Microsoft Threat Intelligence assesses that Storm-0940 was responsible for the recently disclosed ConnectWise supply chain compromise of a multitenant application. msft.it/6018aakEE
Our investigation found that Storm-0940's use of ARTEX accelerated the speed of compromised application abuse to authenticate to target environments through applications that had already been granted access, allowing the threat actor to operate without requiring an interactive user sign-in. AI-accelerated follow-on activity included reconnaissance, application credential changes, and attempts to modify privileges that could enable expanded or persistent access.
Read our blog for a detailed analysis and indicators of compromise (IOCs), along with Microsoft Defender detections, protection recommendations, and guidance on investigating service principal sign-ins and unexpected application credential or privilege changes.
He-Man, Principal MLE retweeted
China just cracked recursive self-improvement with a 27B model.
It’s called “Terminal-RSI”
right now, agents struggle to turn partial progress into complete task resolution. if they hit a wall, they get stuck.
Terminal-RSI fixes this by recursively coupling "verified trajectory rewriting" with "policy refinement."
what that actually means:
- it attempts a long-horizon terminal task.
- it takes the verifier outcomes and autonomously self-rewrites its own failed trajectories into successful ones.
- it extracts those new behaviors and permanently updates its own policy.
- it loops.
it isn't just brute-forcing retries or prompting its way out of an error. each retained stage consolidates the behaviors it just discovered, progressively advancing the agent from "partial local repairs" to "strict terminal verification."
it converts partial progress into complete task resolution, getting strictly better every single loop without a human engineer intervening.
we are watching agents move from simple automation scripts to actual recursive self-improvement.. the gap between human developers and ai is closing terrifyingly fast.
if you are building in the autonomous space, this is required reading.
He-Man, Principal MLE retweeted
“PhD in Bourgeois economics”
Replying to @sigmahamster2
You don't need a PhD in bourgeois economics to see that a system based on infinite growth and extraction is destroying the planet. Einstein was a genius who saw the big picture. Sometimes it takes an outsider to point out that the emperor has no clothes
He-Man, Principal MLE retweeted
There is an (unpatchable?) supply-chain zero-day in PyTorch autograd that allows a regularization term to use the model's weights as RAM and SGD as its CPU clock to run any computation during training.
He-Man, Principal MLE retweeted
"claude, take a serious stab at pretraining. here's $10m"
Today, we're sharing a major milestone in our mission to democratize frontier AI development:
Pretraining at scale without a pretraining team.
We set out to answer an ambitious question: Can we significantly reduce the time, human expertise, and compute needed to pretrain a competitive foundation model from scratch?
The result is Mirendil-Base, our first pretrained coding model, built by one researcher in just 3.5 months with less than $10M in compute, including all research, experimentation, and the final training run.
Our approach is already showing promising results as we scale to even larger models.
This is an important step toward a future where a single researcher, empowered by AI, can accomplish what once required an entire research organization. As these capabilities improve, we believe frontier AI development can become accessible to far more people and teams.
More to come soon, including independent evaluations, post-training benchmarks, and API access.
Pause icons: "China would pause too if we did! They love safety! They aren't advancing capabilities on their own they just follow our irresponsible frontier companies!"
China:
GPT-Rosalind surfaced a new lead for drug discovery: could restoring cholesterol transport protect neurons from degeneration?
DCTN1 is a gene linked to Perry syndrome, a rare neurodegenerative disorder. In mouse-brain data, disrupting it increased expression of genes involved in making cholesterol. GPT-Rosalind suggested a possible explanation: cholesterol may not be reaching where neurons need it.
That gives scientists a specific mechanism to test and, if confirmed, a direction for developing new treatments for neurodegenerative disorders.
With @perturbai_tx and @scrippsresearch, a team of thousands of GPT-Rosalind agents worked in parallel, connecting gene-expression results from millions of mouse brain cells with scientific literature in just 32 minutes. We’re now expanding this collaboration to identify new drug targets.
openai.com/index/scaling-ai-…
He-Man, Principal MLE retweeted
GitHub link:
github.com/NVIDIA/flare-loca…
He-Man, Principal MLE retweeted
NVIDIA just released Flare, a local inference server built for DGX Spark and RTX Blackwell GPUs
- Built for long chats, agents and tool calls
- Serves from Rust, no Python
- Plans memory up front
- FP8 KV cache by default
Interesting timing... right around when DeepSeek open-sourced their inference engine.
Link below.
He-Man, Principal MLE retweeted
Hoooooly shittttt SpaceX’s AI segment is now making profit also they made $2.3 billion dollars in profit off Starlink over the last 3 months
Our Q3 2026 earnings report is now public → ir.spacex.com/files/content_…
The live webcast with @ElonMusk, @Gwynne_Shotwell, and @BretWJ will begin in 30 minutes
He-Man, Principal MLE retweeted
“We found DeepSeek V4.1 Flash 1017 beats GPT 6 Sol on our multi-step offensive cybersecurity task.”
READ THAT AGAIN.
the same cyber capabilities that OpenAI kept behind a gate got beaten by a Chinese open model you can run locally.
buckle up guys.
We found DeepSeek V4.1 Flash 1017 beats GPT 6 Sol on our multi-step offensive cybersecurity task. Sol was tested via the Daybreak trusted access program.
We have been building a cybersecurity benchmark based on simulated corporate networks. This is a valuable indicator for the cyber-risk a model poses if in the hands of a bad actor. We plan to make the full evaluation public in the coming weeks but wanted to share our first immediate finding here.
He-Man, Principal MLE retweeted
META CONFIRMED BEHIND GIANT 2.2GW KENTUCKY DATA CENTER
Mason County officials named $META as the company behind the hyperscale site last week.
$10B for the first phase, with more phases planned. About 2,000 acres are rezoned so far. The plans show six buildings over 700K sq ft each and three substations, with power that could scale to 2.2GW.
Meta also gets a 40-year property tax break.
This is a county of 17K people. Those are the kinds of places hyperscalers are going now, because that's where the power and land are.








































