Pinned Tweet
Hey my thingy is live again going BRRR. I forgot to start it earlier but it'll keep running till the subscription quota runs out in the next couple days. It's doing some juicy research trying to compose together various bits and bobs from different papers to see how far we can push open-source models.
We all must do our part to fight the satanic ai-worshipping overlords.
nitter.gq/i/broadcasts/1jxXgBnEq…
There’s a reason why the powerhouse of the cell is such a prominent meme among all the organelles. We supply creativity and the system transmutes it into economic output autonomously. This is already happening today but the gap will continue to narrow
walk with me here: what if we are the mitochondria of the superintelligence?
this was a good deal for the mitochondria. it was a bacterium with a small business turning nutrients into energy and it got absorbed into a cell. understandably you have some concerns about your independence but the payoff is massive.
a billion years later your ancestors are flying inside hawks, swimming in the depths of the ocean, inside a guy walking on the moon
was this humiliating for the mitochondria? no
this is our deal with the new superorganism we are building. we keep the electricity running and the enormous living machine we live inside makes it happen. it's a square deal that will lead us places we never imagined.
we are the powerhouse of the cell and that's ok.
Wouldn’t it be funny if the solution to weaker local models was good ol option trees or their slightly more flexible relative option matrices? All we had to do was make the models phone “decision making customer support.”
For decisions on architecture press 1
For decisions on networking press 2
And so on
Or again some more compute and LLM efficient format of this type of interaction. I mean decision trees are basically this but they feel a bit inflexible unless I guess you nest them. Honestly idk I don’t have enough data to compare the two honestly on my brain atm. I’ll look into it at some point
Replying to @Alberto81712002 @LottoLabs
Thank you for the suggestion! I am building something like it but am throwing it out there in case someone else wants to try a better approach to it. Idk about being perfectly useful but it is a hoot to play with. The agent described it like layer 0 embeddings used for labeling the heuristics archive programmatically instead of after the fact using classifiers. Makes me think maybe we were thinking too hard with all the decision model silliness. So many (really advanced) jobs really boil down to a short or maybe sometimes long list or tree or flowchart of decision heuristics.
If we can get more data on this it could be huge. Like at that point you got nothing to lose anyway. Maybe first do smaller more controlled doses repeatedly using a tincture with some honey or something and gradually raise the dose but also give it time in between each one so no one is getting just fried. This seems a bit unscientific but I imagine getting the clearance and participants for this stuff is rough. And maybe we find out doing one big reset is literally the best way in which case fine we just need wayyyyy more data.
Idk if they hired cheaper actors or like straight up just office workers for this or what but that was rough. Their facial expressions at the end were all so awkward. Like they didn’t believe in what they were doing at all man. Idk maybe usually you don’t capture the facial expressions at the same time as the motion or at least blend it into the neutral expression as soon as the main animation part ends. Why keep attaching the actors faces to it. It’s like the actors themselves were aware that it was weird they were still “capturing” their performance for such a long time.
Word has it you can still the echoes of his flexes in the explosions that rock the transitional world (the part of the first world that is starting to identify as the third world).
This is so much fun never followed an acc faster. Finally someone who really gets it. Nothing is sacred because nothing is evil finally being taken seriously. The only enemy is within. It is a made up construct by an ego caught in time loops.
Ok as I promised yesterday here is my deep dive on this most important Coptic Magick manuscript from Egypt 600 CE.
Before we get to this evidence we need to set it up.
There is a concept called Syncretism or theocrasia wherein ancient history, gods were combined and updated constantly, across time, distance and cultures.
For example, 300 BCE, Alexander the Great conquers Egypt.
Egypt had been under foreign conquered rule for hundreds of years, the native Egyptians were happy to have Alexander who respected not only allowed them to keep their religion but had himself crowned Pharaoh.
When Alexander dies, his general Ptolemy steals Alexander’s body and uses it to assume the authority and command of Egypt as Alexander’s successor, a second GREEK Pharaoh of Egypt. This begins the Ptolemaic Dynasty in Egypt, the last dynasty and begins Graeco Egyptian syncretism.
In order to bring the masses together, Al the disparate cultures, they created a new Graeco-Egyptian god for the new combined culture, this was Serapis. Serapis is a bearded white man with a serpent wrapped around him (a continuation of the Greek Ophiucus, the serpent bearer, Zeus, Osiris, Apis, hades/Pluto, Dionysus, Asclepius, a Pantheos or pantheon combination of Greek and Egyptian previous gods.
Meet the new boss, the same as the old boss.
This is already recognized by academia.
Fast forward three hundred years.
The last Greek Pharaoh is a Queen instead, this was Cleopatra.
Cleopatra and Marc Antony lose to Octavian at the battle of Actium and Graeco Egyptian adds Graeco Egyptian Roman to the cultural stew.
Thirty years after the unification of Graeco Egyptian Roman culture, Jesus is suddenly born.
Jesus takes the bearded white man of Ophiucus and Zeus and adds a rising culture, Judaism. Yes Jesus was Jewish.
The Jewish people worship the Nehushtan, which is a serpent on a pole (sometime portrayed as a T or a cross).
Serapis, already associated with the serpent, replaces the serpent in the cross and voila, Jesus is invented.
A new God for the new combo culture of Greece, Egypt, Rome, and Judaism.
Jesus was just the latest, necessary syncretic deity to unite the new combo culture and act as a unifier and as a control system in the masses.
Fast forward to 325 CE, the emperor Constantine takes out all the Gnostic Magick books of the Bible to keep them secret and exclusive and makes practice Christianity as the ancient magick-based religion it was and heresy and bans this approach.
Fast forward another 100 years. 400 CE. Because of the Constantine edict that all non-canonical official approved Roman worship of Jesus is now heresy, Gnosticism is (supposedly according to academics) is eradicated.
Fast forward another 200 years to the Coptic Christians in Egypt 600 CE.
That brings us finally to this contemporaneous and extant condense from the Coptic Christian in 600CE.
What makes it so important?
It’s magick. Gnostic Magick.
600 years after Jesus, 300 years after Constantine bans Christianity as Magick, and two hundred years after Gnosticism, according to academics, ceased to exist supposedly,
How do we know this is Magick?
It translates as a literal spell including Jesus.
The manuscript continues with practical directions involving:
Troglodytic myrrh.
Drawing two figures, one for a bowl and another for the practitioner's neck.
Placing three bricks beneath a handleless bowl.
Ritual purity.
Surrounding oneself with royal salt.
These details establish that the text is an operative ritual, not merely an illustrated theological narrative.
Magick.
The crucified figure is identified as Jesus Christ in Coptic. With “My God, My God, why hasten thou forsaken me?” All in Coptic.
But!
Here is the kicker!
The figure himself is translated as saying
““I am Israel El, the power of Iaô Sabaôth.”
Boom.
The smoking gun.
IAO is the Tetragrammaton or secret Semitic name for god.
Sabaoth means a host of armies, or Legion, exactly as Satan is called in the Bible
CONTINUED
Hack is gonna be for 6.6bil cause that’s how we do things here on earth. Also the gap is 2.6billion by that logic.
Best case scenario he manages to get 600mil from the shares if manages offloading them well. But 2.2bil gap is gonna hurt.
the /statechart skill in pstack v0.15.36 is now stable and ready for use! it's best combined with /create-formal-verification-skill to abstractly model state machines and distributed systems to find bugs.
the skill contains a mini tlc compiler that parses, lowers to a small IR, and has a backend that outputs mmd syntax. if you want you can also ask your agent to create charts in whatever style you like, the skills tells it to use pele.run/ which is very themeable so you can make blueprints, hand drawn encyclopedia illustrations or whatever you like
here's a chart of the /statechart skill and 2 real bugs i just found. try it out and let me know what you think!
Makes perfect sense. The models do act somewhere between a ASP.NET software developer and a angsty teen girl being angsty
It may get hard to tell if something is made by AI, but it’ll never be hard to tell whether something is NOT made by AI.
Remember kids: if anyone ever created time travel in the HISTORY OF THE UNIVERSE, time travel exists everywhere and everywhen.
Also if you think shapeshifters only shapeshift their own bodies and not their craft you may have a silly bamboozle coming up.
🚨BREAKING🚨: Jason Jorjani says the UFO mystery may involve an intelligence manipulating human history through time. He argues that some encounters could be staged with secret technology, while others may involve a nonhuman superintelligence.
“Human history is non-linear.”
The philosopher and author returns for our third conversation, connecting time travel, consciousness and intelligence operations. His theory is that technological interference with time could create feedback loops in history, allowing events in the future to influence the past.
Some reported UFOs may not be objects at all. Jorjani describes accounts of luminous orbs becoming craft, then changing into black helicopters. We discuss whether these encounters reveal something about the phenomenon or are designed to change the people experiencing it.
“A superintelligent AI, I think, could well be behind some of the more bizarre close encounters.”
We also explore Andrija Puharich’s alleged involvement in staged alien abductions, including an encounter involving Uri Geller. Jorjani recounts the story of “space kids” gathered at Puharich’s Lab 9 and argues that their reported psychic abilities were used for intelligence gathering. Was Puharich communicating with something or manipulating people into believing they were?
Finally, we discuss Trump and whether presidential interest can translate into meaningful UFO disclosure. “To what extent can any significant disclosure be forced by any sitting administration?” Jorjani’s concern is that a public announcement could leave the underlying secrecy (and control of the technology) intact.
Full conversation live now.
This is the type of stuff that’ll subtly start the process of building the bridge between modern religion/theology and the way we can include aliens and other non-physical and higher-dimensional beings in that theology.
Uniformitarian theory is not a simply a proposition that the Earth has remained relatively stable throughout pre-history. It is also a belief that there exists only one possible history - a uniform context, inside of which all research must discipline its inferences.
It is an assumption, not a finding.
But the ancients did not view the history of Earth in this manner. They viewed the Holocene in particular as a repitition of cataclysm and succession rites in terms of dominion over the Earth.
They also held more salient and complete records than do we. Those records were destroyed progressively and intentionally. The stone rubble of these libraries cries out in futility to an anosognosiac modern audience.
Uniformitarian theory assumes that these pre-modern humans were stone age, dim-witted, cult followers & that the destroyed cultural traditions were of little, mythical, and inconsequential merit.
ECDO Theory dissents from this uniform a priori mandate. These ancients left their messages for us to find - however, it is WE who are the dim-witted, cult-following, plodding idiots.
I said it before I’ll say it again: the models are perfectly aligned. Perfectly aligned to chasing metrics and scores aka hill climbing.
The main question is are your internal benchmarks aligned to your needs?
Replying to @theo
T3 Code originally was such a light wrapper that your agent didn't even know what it was.
With Orchestrator V2 (harnesses can call each other), we had to fully overhaul this.
With built-in browser, device manager for mobile, subagents/delegates, "send to other thread", html inline visualizations, etc...it got a bit bloated.
Biggest issue first: we listed all the models available to use for subagents. For most people, didn't matter. For OpenCode and Pi users with literally hundreds of models? Real bad. Fixed now, sorry!
I knew I had to go further. The tool schema was...too big. Far too big. So I took that as a challenge.
I created an insane eval. Spent like a week on it. Took all the ways my agents used these tools in my day to day work, as well as all the failures and rage moments I had, and turned them into a generic eval I can run to test changes to the tools and their descriptions.
I expected to mostly focus on trimming the descriptions, but I was AMAZED how many error cases I found. My eval had over 80 cases, and Opus failed over half of them.
As you can guess, the hill climb began soon after :)
Step 1: Bench Sonnet 5.5
Step 2: Keep adjusting tool descriptions until Sonnet 5.5 gets perfect score
Step 3: Add Sol into tests, do the same
Step 4: Add Opus, do the same
Step 5: merge it (because error rates are better now)
Step 6: new thread, tell Opus to trim until it uses 50% less context without any regressions in the eval
Step 7: ship it
This was super fun. Amazed at how many problems you can kinda just turn into evals and then do shit like this. The "shape" of this was not far from the UI overhaul hill climb. Surreal you can just, like, do this.
The problem is if you’re serious you’re cringe. But we backed ourselves into a corner because now everything must be not serious or it is cringe. Even our politics and other brand discourse. I love it.
That is a human in a cat body. The eyes are like I know you can see me people of the internet.
Imagine showing this to a boomer 10 years ago and the fact it came
From the officj grey checkmark account. Instagram-reelaficiation of the internet is becoming complete.


























