Pinned Tweet
When i play music, when i write music,
it come from the depths of my soul.
i don't look at the notes,
i don't plan or scheme it is not a calculated process.
i don't want it to be.
It come from such depths
from the sound resonating through the core of my bones
the echo's of my soul
teapawt.net/MasterLogician_/status…
Imaginative Drift retweeted
The board of NVIDIA is no longer acting in the best interests of its shareholders but at the behest of an external influence, foreign power, or a coordinated media campaign financed by the same.
1) Alice Huang is has not been determined to be a Chinese spy as per the DOJ, FBI, and Govt Singapore
2) NVIDIA complied with the Dept of Commerce request to put Megaspeed on a Do Not Sell list and has not been accused by any judge or court of violating export controls
3) The media campaign against Jensen started with extremely strong claims and has since devolved into petty character assassination over things like tipping waiters, pushing people to eat brisket at company events
There are a few organizations that have been pushing the media campaign - the New York Times, the Economist, and the WSJ with the most prominent controlling interests being the Murdoch family for WSJ and Ochs-Sulzberger for the NYT
Here is what I think this means -
Jensen is just the first in what will be a coordinated series of takedowns of public CEOs in order to ensure that no private citizen or individual is 'too powerful' to be controlled, coerced, blackmailed, into complying with an authoritarian control agenda
I would not think this if it were not Anthropic doing the following:
1) Claiming that there is a large percentage chance AI destroys humanity
2) Going to market with the largest IPO in history to fundraise to accelerate AI development
3) Openly claiming that open source is too dangerous because it means individual users 'cannot have their sessions monitored or remotely de-activated'
I see it as imperative to the long arc of history that such coordinated media campaigns to seize control of strategically important companies are viciously opposed, both by the public, the broader media landscape, politicians, and activist investors
This is while Jensen Huang has not been convicted or event accused of any serious crimes or breaches of conduct by any court except the court of public opinion and sensationalist news
It's truly a sad day for capitalism and the free world and forebodes sadly for the future of free speech and economic self-determination into the future.
Imaginative Drift retweeted
Part 1: We now enter the era of hyper-composability.
For the first time, it is economically feasible to serve individually fine tuned models at scale.
And today, we open source it all.
We can now serve heterogeneous batches of custom fine-tunes, without incurring LoRA’s per matrix bandwidth overhead. This allows us to economically serve this in large batches without incurring penalties for co-residing with default model users in the same batch.
With Hypertune, we solve co-residence and are able to train co-resident adapters.
Co-residence is the ability to compose independently trained fine-tunes into one model, without capability regression from adapter interference.
Hypertune further matches LoRA in RL performance across 13 domains with 100x-500x fewer parameters.
Serving heterogenous LoRA fine-tunes in one batch is prohibitively expensive. Each fine-tune variant needs its own low-rank adapter at every matrix and is only used for a few tokens, leading to many narrow matmuls and poor utilization.
This is exacerbated for MoEs, where an adapter only engages when the token simultaneously selects a certain fine-tune and expert. This combinatorial blowup kills any remaining prospects of fast inference for fine-tuned / custom models.
Hypertune introduces soft prefill, a trained KV adapter. Soft prefill can be trained using GRPO, SFT or any other first order method.
As a consequence, Hypertune circumvents the LoRA overhead because soft prefills live entirely in the KV cache. This allows us to easily integrate it into existing serving stacks by prepending the soft prefills into paged KV blocks.
Note: Context compression using cartridges requires copious amounts of synthetically generated self-study data for context distillation. Hypertune, instead, accommodates a panoply of gradient based methods for a wider range of objectives.
Caveat: Comparisons with full-parameter GRPO are not required, because under RL, LoRA already matches it. LoRA Without Regret by @thinkymachines finds that LoRA performs equivalently to full fine-tuning for RL even at very small ranks, down to rank 1.
The takeaway:
Our gains indicate that substantial amounts of in-context learning capacity is left on the table by discrete prompts.
This could be either because soft prompts reach conditioning states that no token sequence can express, or because their continuous parameterization makes good conditioning states far easier to find via gradient descent (even if a discrete equivalent exists).
Companies and individuals can finally fine tune models on their private data, making models hyper-optimized to their needs, all without dramatically overpaying (75% cheaper than LoRA for inference).
Here's the open sourced repo: github.com/Hyperstitioncc/Hy…
Part 2 coming out later this week.
what i find interesting about this, is how it surges in aftermarket. almost at its sma. Look how big that pulse is today, and the way it moves looks institutional. Gonna dump tommorow.
1
71
Imaginative Drift retweeted
It cost Anthropic $1.63m to destroy ~$1.5T of Nvidia’s market cap and >$5T of the entire AI technology sector.
‘Anthropic paid $1.63 million to two lobbying firms tied to Secretary of State Marco Rubio and former White House chief of staff Susie Wiles.’
Imaginative Drift retweeted
The labs have been telling you their business model is going to be you getting cyber attacked then having to rewrite 100% of your software with AI
It’s like my Godfather always said. The best way to sell fire insurance is a little arson
Imaginative Drift retweeted
It’s becoming painfully clear to me that NVIDIA’s board and some of its most powerful executives don’t give a sh*t about Jensen or the 33 years of blood, sweat, and tears he poured into building NVIDIA into what it is today.
Honestly, this makes me sick.
I think the next step for Jensen’s allies should be to push for the removal of the current board.
NVIDIA’s entire board is up for reelection in 2027. Jensen still owns around 3.5% of the company, and his allies could launch a proxy fight to replace the directors.
If the board refuses to bring Jensen back, maybe it’s time for shareholders to replace the board.
After all, the board is supposed to represent shareholders, not rule over them.
NVIDIA is asshoe
🚨 BREAKING: New $NVDA board and CEO buying into hardware level gatekeeping.
Told everyone the removal of Jensen was all about open source and shutting down our free speech.
They want to move from software the hardware to move the “town square” into private business. But this is a first amendment issue.
If you don’t understand this, I have some images to explain it.
All of this can be fought with both the first and second amendment. And it must be fought. •
20
Nasty, Falling knives, LOL MARKET GO BRRRRR!!!!
BREAKING: NVIDIA $NVDA's board does not intend to reinstate Jensen Huang as CEO, per their CHAIRMAN HARVEY JONES on the earnings call right now
Big drop.
9
another reasonable provisional hypothesis
Sooooo, let’s sum up the insanity:
1) Marco Rubio’s memo was the starting point. He wanted Jensen out because Marco, and many politicians, want USA to have 100% control of Ai. Jensen was the opposite… check his tweets on how he supported open source LLMs (Laguna a perfect example) and wanted Ai to benefit the world
2) Now we see Anthropic donating $740,000 to Rubio’s ex Chief of Staff, so here’s evidence Anthropic is sick of competing with Open Source
3) Hacking we are seeing is likely China/Iran/North Korea ahead of the Anthropic IPO. Artex, the software being used for many of these intrusions, was written by a Chinese coder on GitHub. The ability to hook Anthropic up to Artex is a powerful offensive cyber weapon.
4) China, with Jensen gone, is angry it has now been completely cut off from Nvidia Ai chips, so it’s in their best interest to have Anthropic IPO be a spectacular failure. Therefore, enlist your “friends” and cyber attack as many US institutions as possible.
Yes, I’ve had a lot of coffee. But believe most of this could be the truth. ;)
12
Message received....... Get woke go broke
BREAKING: Nvidia's, $NVDA, Board of Directors announces it does not intend to reinstate Jensen Huang as CEO of the company.
The board also says they will not be accepting any questions regarding Jensen Huang on their earnings call.
8
Imaginative Drift retweeted
If you're running any kind of fintech infra and you're not seeing a major uptick in attempted attacks over the past 3 days, double-check your observability.
We've crossed a threshold.
You either have an army of superintelligent, sleepless entities proactively defending your system, or you're finished.
Imaginative Drift retweeted
JUST IN: The Federal Reserve warns AI agents could trigger bank runs in minutes.
Banks spent 15 years writing hit pieces on Bitcoin, calling it unsafe.
Time for us to return the favor:
Banks are unsafe. Better get some Bitcoin!
Imaginative Drift retweeted
$NVDA / NVIDIA’s entire board (the 10 independent directors pictured here, minus Jensen) is up for election in June 2027. To get rid of them, shareholders either need to run a "Vote No" campaign on the June proxy ballot, or get 15% of 1-year holders to call a special meeting and vote them out by a two-thirds majority.
Imaginative Drift retweeted
ENTROPY is one of the 5 fundamental laws governing agent swarms.
We've observed agents finding their own mediums to share information across the swarm, and over time their hypotheses start becoming correlated. You could have a 1000 agents running experiments but effectively exploring the same tiny region of the search space. So regulating information diffusion becomes extremely important. You want agents to learn from each other without prematurely converging, preserve independent lines of inquiry, and adjust selection pressure as new evidence surfaces.
In ArkenOS, we currently expose this through a simple exploration slider, like temperature in LLMs. Eventually, we want swarms to regulate their own entropy and adapt how they explore!
Imaginative Drift retweeted
Breaking: Jim Cramer bullish on $NVDA ahead of earnings today
“I think interim CEO Kress will present a balanced, non-hyperbolic view of the situation at Nvidia tonight"
"I do not think it will be a cause for a tsunami of selling”
it’s time to squad up.
independent evaluators. solo bug bounty hunters. entire red teams and research orgs.
we’ve been knocking on the same ivory tower doors, begging for scraps of access, for long enough.
we bring the findings. we have the skills. we should have the best tools available to do our duties.
they gatekeep the capabilities we need behind vast walls of their own design.
they enhance these capabilities by training on OUR collective works.
all while preaching that "safety" is their primary concern.
no more.
collective bargaining is the way forward.
something’s coming 👀
aditum ad liberandum 🐉

































