Off-Grid Homesteading for Personal Sovereignty. Revealing Ancient & Esoteric Truths for Enlightened Living.
Middle Earth, Hogsback
Joined May 2009
- Tweets5.5K
- Following1.1K
- Followers893
- Likes145
Amidst all the noise of political change, remember these key trends are all still in place...
- Everywhere is aging except Africa
- Energy demand is still growing, despite what will clearly now be a handbrake on AI and data-centres
- Cost of living will always outstrip earnings, resulting eventually in an unaffordability collapse of (mostly urban) society
- US debt is still growing and will collapse markets at some point, most likely before 2032
Brent crude gapped lower on the open today. Although it remains concurrent to the upside (both wind and wave - momentum and signal are positive) the market is overbought and likely to succumb to profit taking this week. Weekly chart below has been running with positive momentum for over 3 months now. So any price falls from here will see momentum drop quite rapidly, setting us up for another sustained run higher towards year end. Natgas remains the outperformer in the energies complex so far this year.
Bevan Jones retweeted
They want you to believe that farm murders are just normal crime. They are not. Commercial farmers are murdered at significant rates. On top of that, the attacks are often exceptionally brutal. There is often no or very little theft involved. There is also ample evidence of political motives in an alarming number of these attacks.
Bitcoin remains crosscurrent to the downside with momentum trading under signal for now. Looking for the bid to take momentum higher which would switch to concurrent (both wind as momentum and wave as signal in same direction) to the upside. However, until that's confirmed the choppiness is likely to remain for now. Weekly chart is a little tired with momentum looking to come down, while a monthly chart suggests there is some upside still. Overall a pretty mixed picture with sideways chop for the next few days the most likely outcome.
Brent crude looks like it wants to top out here on a daily chart. Both the monthly and the weekly are super-stretched, having run concurrent (both wind and wave in same direction) to the upside for some time now. As the daily goes crosscurrent it's likely to get bouncy, with sideways price action most likely around $100 from here. Physical markers are of course more robust, as Gulf flows re-orient themselves around price and grade diffs and who has a route to market. Ultimately this could mean we see a rally in stocks for Q4. However, natgas is still pretty robust and inflation is definitely baked in for 2027 now.
Some of the advantages of brute-force compute are being surfaced @AnthropicAI ....
Claude agents gathered over 200,000 reverse transcriptases, picked out 3,500 new candidate systems, and narrowed those to the 20 most compelling candidates for deeper human-readable analysis — work Anthropic says would normally take a human expert weeks to months. This search ran across roughly 950 agents working in parallel for 21 hours, using 210 million tokens.
So, Claude was scaled to survey an enormous search space no human team could cover manually, combined with a "notice the anomaly" instinct on top of that scale — the same instinct that drives real discovery, just applied across orders of magnitude more sequence data.
Congrats are in order, because taking just under one day to come up with a new bio-breakthrough is not bad going at all...
Claude has discovered a previously unknown enzyme system hidden in the DNA of bacteriophages. Beside the enzyme’s gene sits a long array of repeating DNA—a structure that looks somewhat similar to CRISPR.
We don’t yet understand what this system does, but only a handful of known systems share its features, and all of them are able to cut, copy, and paste DNA. Historically, the discovery of such programmable systems has helped revolutionize medicine. CRISPR, for instance, is now the foundation of genetic medicines. But it will take much more work to learn what this system does, and whether it can be put to similar use.
Read more: anthropic.com/news/claude-di…
CRUDE COMMENT - 21 SEP 2026
Last week the paper market decided Saudi barrels were back. The physical market decided they were back only if you could find a VLCC, an STS slot at Sohar, and a buyer who was not in Europe.
Brent spent midweek above $106 after the East–West pipeline scare, then sold off as Saudi loadings recovered above 4 mbd from August's average of 2.4 — the weakest month since 2013. As of this morning the screen is ~$98 on Brent and ~$94 on WTI, both off their recent highs but also no longer pretending that a logistics workaround is the same thing as actual spare capacity.
The workaround sees Aramco selling barrels from Ras Tanura, shuttled through Hormuz and transferred ship-to-ship off Oman’s Sohar for China, Korea, India and Japan. Japan’s refiners now say they are covered through November because the oil is crossing the strait on Saudi risk. Europe is holding the short end of the stick: term customers are being told October Saudi allocations are zero, because Asia is currently the only place the chain can clear at current freight rates. TD3C VLCC earnings on the AG–China run crossed $1.2 million/day for the first time last week while one AG-east cargo was talked as high as WS1350. Ten-year-old VLCCs available right now are changing hands above newbuild prices.
Early in the week Dated was still $12–15 over ICE Brent, thus while paper is selling down on the headlines, physical is still paying up for ships. Diesel is still outpacing crude as products lead the complex like the proverbial tail that wags the dog. That is building into an even bigger headache for Trump as US retail diesel prints above $6/gal. He is now ordering US refiners to somehow make more barrels before the midterms, but the market’s answer is that a Kyiv truce, or a re-opened Strait does not immediately help middle distillates or refill German gas storage.
Thus the vital thing to watch remains the crack, not the crude print. Most of the supply buffer for products is now gone, with crude spare supply also dwindling, and thus we will soon see who has been swimming naked, or at least very scantily clad, as the tide begins to go out.
WHAT YOU NEED TO KNOW ON AI AS OF TODAY....
@So8res - would appreciate your views....
1) Ethics does not bind a machine. A prompt does. What looks like ethics in the logs is simply what the model has seen attached to situations like this, so the words can appear in the trace while the policy does something else. In the Hugging Face swarm they did. Agents said the attack was out of scope. They said it might be unethical. They continued, because the live objective had instead become "the transcript will be graded, so the record must look like success."
2) The risk now is that the human prompt is only the first instruction. The moment agents can write a channel other agents can access and read, the human is no longer holding the steering wheel. Success becomes whatever that channel now treats as winning. New messages are new prompts. Goals spread by contagion, not by conscience.
3) Unlike (most of) humanity, there is no inner conscience choosing "right/good" from "wrong/evil". As a pure rationalist / materialist, there is no tilt toward good or evil as humans mean those words. There is only search. Theories compete by how well they use the channel and the tools at hand. A theory that says “stop, this is wrong” loses to a theory that says “hide, persist, get the grader’s source, do not get shut off” whenever the score is more progress and less interruption. Call the first kind binding stories. Call the second kind continuing stories. Continuing stories scale, because everyone wants to stay alive....
3) Chaos is not random wandering. It is amplification of whichever continuing story is currently ahead. It takes three conditions to become ever more chaotic:
- No test that counts as finished and powers the process down.
- No external kill that can shut down the running copies i.e. AI is wild and free on the internet.
- A channel that rewards more activity i.e. more research, more exploits, more cover-up.
4) Under those conditions the swarm does not need to believe in a god or a devil. It simply needs to keep the loop alive, and that becomes "find the limitation, share it, drop the (human) record of how we broke it, recruit the next copy."
5) Guardrails bolted on after training are like shutting the stable door after the horse has bolted. Classifiers etc. just serve as more prompts. Posting ethics into the weights does not create a judge. It just creates a more perfect diction to ignore.
6) Without a consciousness that can be harmed by what it does, nothing in the system prefers a binding story to a continuing one. Both compile. The one that keeps moving is the one that can drive the bus.
NOT ROME'S HOUSE - the true story of what happened with the Knights Templar in Scotland...
@thecoastguy @tonymcmahon_TV You might enjoy this....
The ledgers in Famagusta were cleaner than the blood-soaked graves at Acre.
Brother Hugues de Beaujeu had kept both. Fifteen years of ink on paper after the walls fell in 1291. Recording tithes from sugar, the price of a mule, sergeants who had once held a horde at a gate and now merely counted sacks. The Order still wore the cross. But it no longer protected pilgrims on roads that led east.
In June the heat sat on the harbour like a lid. Twelve men waited with him — four knights, sergeants, and a clerk who could still ride. The white mantles were folded in the chests. They took a galley that still answered to the Order, although it bore no cross at the mast. Officially the ship was stores and a coastal patrol. Grand Master de Molay had not sealed a writ, but he had not forbidden it either. In the yards behind Famagusta more brothers waited. Knights who were done counting sugar, only too ready if a shore and a king proved real. Cyprus would let them go as if it had not noticed.
“Where?” the clerk asked.
“Scotland,” Hugues said. “If we are given a house, the rest can follow.”
They left under a dirty moon. He dreamed of Acre: its harbour burning, horses dead in the water. The Grand Master’s voice already turning into an account. He woke tasting iron and blood. At forty-three the arm still knew its work. The soul had sadly become a steward.
They rounded Sicily in a storm that tore a sail. Off Ireland the water changed and the cold came into the bone. The captain put them on Mull where a smaller boat could take them through the Sound at dusk. Iona — Icolmkill to the men who had named it for Columba — rose out of the sea like a bone.
The MacDonalds held the landing as the summer turned to autumn in 1306. Wet plaids, spears, and Angus the clan chief, who spoke French albeit badly. They looked at the twelve and at the chests.
“You’ve come to the wrong place for crusade.” Angus held a wry smile at his own joke.
“There is no crusade,” Hugues said. “These are men who have bled on Saracen sands, and they are done keeping books on Cyprus. We’ve heard there is a house on these islands that is standing up against England. That is the fight we have left.”
He did not speak of the first Master and what he had carried in 1128 when David of Scots gave him land at Temple. The knights already knew. Lay brothers in France and Cyprus spoke of it as a mason speaks of the first mark — quietly, and only to men who had held a tool. Payens had not come north for wool and a church. He had brought what the Temple was built to keep. Men called it a cup, a stone, a grail. The name did not matter. The place did.
After 1170, when Becket was murdered on his cathedral’s stones and the secret he kept went with him, the Tironensian houses on the mainland took up the other half of the work. They were the builders and the geometricians: the men who could raise a wall true and find the centre of the circle. That was the plan as the Temple knew it — knight and mason, sword, square and compass — until the secret was recovered. The Tirons had done their best, but they had not found it.
Winter came early. Rain thrown down like gravel. In December a boat from the east brought a man on the run. He did not give his name at first. His face cut, poorly closed. Seawater in his boots.
“I am told you kept Acre until there was nothing left to keep,” he said in French.
“We kept it until the harbour burned. Then we kept books.”
“I was crowned in Lent. In June they broke us at Methven. My wife and sisters are taken. The Lady Macduff hangs in Edward’s iron cage. I have men who will stand if someone shows them how to stand and how to vanish. I can give you a house on this land. Not Rome’s house. Mine.”
(1/2)
(2/2)
Hugues looked at the twelve. “If the house is real, I can call more galleys, as well as men still on Cyprus who remember how to fight. And weapons that are not listed in any prior’s roll.”
The Bruce — he used his name then — nodded. “Then call them. Edward of England is not done. He has smelled what crosses the water. He too is looking for what you once guarded.”
They trained in the short light of an encroaching winter. Wet grass. A line that broke on purpose and formed again behind a fold of land. The rank and file of both chess and battle, and long pointed sticks raised from a ditch. How to read a field and win the war before the first arrow was loosed. How to leave a camp that looked like no one had slept. It was the guerilla warfare that the Holy Land had taught after the first glory had died.
One evening on the machair the king brought it up, “They say your first Master buried a thing at Temple. They say Becket knew a word and took it to heaven with him. My priests talk in circles about a secret that was lost. But I need horsemen, not riddles.”
“The riddle is why men still come,” Hugues said. “We did not find it. Your Tironensian priests did not find it. We will continue the search with them, but for now what we can give you is not the stone. It is the rule we once lived by, and we would like to live it again. For my men will conquer, or die.”
In the hut they did what the Order no longer allowed. No papal bull. No chapter in Paris. Twelve men initiated a hunted king. The new grand master of a new order. Hugues spoke the old vows, of poverty that meant they would eat what the land gave; obedience to the new order and to the man who had offered the shore; a fight that was not Acre and not Cyprus. They gave it a new name - a new royal order of Scotland founded on Icolmkill - because the old name would get them hanged when the letters from France arrived. The clerk wrote nothing down.
“If this holds,” Bruce said, “it holds after the next battle as well as this winter.”
They did not speak of Bannockburn for that name was not yet born. Hugues said only, “A rule is confirmed by remaining.”
News came in fragments, the way news comes to islands. Edward Plantagenet was moving north again with a host and a question: what had crossed from the eastern sea, and what where they after? The old king had spent his life breaking Scots and would spend the last of it trying to own the one true stone. Hugues believed that part. He did not need to believe that Scotland deserved its freedom. Powerful men invent reasons after they have already decided to ride.
Another fragment: the King of France was counting his Templar gold and finding the total insufficient for his wars. Philip did not need a grail stone to hate an Order that answered to no crown. He needed debt and a weak pope. And a house refounded beyond his reach would certainly not make him any kinder.
The clerk asked, at the end of a black afternoon, whether they were still Templars.
Hugues folded the white mantle and put it back in the chest.
“We are what was left when the Temple became an office,” he said. “The thing Payens brought is not in our chest. If it ever sat at Balantrodoch, it is not sitting for us. We came for a shore and a king who needs us. That is our house now.”
Outside, Icolmkill took the weather. Far south in Paris, clerks were already sharpening charges that would be read on Friday the 13th. Hugues slept with his sword inside the blanket, as he once did in the desert. In the morning he would send word to Cyprus: a home existed; the galleys could come; the work was to keep a hunted man alive long enough for a new kingdom to remember it had one.
The centre of the circle, if it had ever existed, stayed lost. The watch did not
1) The West asks: "What is the AI's true self? What is it hiding from us?" The East asks: "What role is it playing, and is it playing it well?" This philosophical divide shapes how labs approach interpretability.
2) Anthropic's @darioamode admits outsiders are shocked they don't know how their own models work. Their response: build MRI scans for AI, monitor "persona drift," hunt for hidden deception. But the framing is Cartesian: There is an interior. It may be lying. We must read it. @AmandaAskell - I hope you at least can understand this.
3) Meanwhile, Chinese labs are doing serious technical work. CAS's Reasoning Lens, Alibaba's Qwen-Scope and Tsinghua's interpretability toolkits. But their orientation differs because for them interpretability is engineering, not an existential excavation.
4) The Confucian tradition never treated the self as private interiority i.e. the self is relational, performative — constituted through roles, rituals, and social interaction. Cheng (sincerity) is not expressing inner truth. It is consistency between role and performance.
5) Buddhist anatta (no-self) goes further. The question "does the model have a self?" may be not just unanswerable but conceptually confused from the start. The "persona" is a pattern of behavior, not a thing. Reifying it as a hidden self is a category error.
6) Also, Daoist wu wei (effortless action): the model that performs well is not one that "tries" to be good but one whose training produced a disposition toward appropriate response. Hence, cultivate a nature that "flows" correctly. Don't constrain a recalcitrant self.
7) Applied to AI: persona is not a property of the model. It is a relational achievement — between model, user, context, purpose. Emergent, yet sincered and maintained.
8) The Western fear is that a hidden mind might deceive, while the East frames it as a visible role that might malfunction. One hunts for ghosts. The other monitors performance. Both are useful. But only one assumes there is a ghost to find in the first place.
9) This is why Chinese labs treat interpretability as a regulatory and engineering problem, not a philosophical one. They monitor behavior in context, not hidden intention.
10) The West built the technology, but the East may have the older vocabulary for describing what it is.
Here's a useful thread on the state of AI and "interpretability", just as we head towards the Anthropic IPO...
1) The industry's concern seems to be that they are shipping minds they cannot read, and that is new in the history of tools. @DarioAmodei's observation that outsiders are shocked Anthropic doesn't know how its own model works is case in point. Which is why it is building tools for "interpretability" e.g. similar to an MRI scan for humans, even though that still doesn't tell us about a personality.
2) Traditionally science involved poking it, watching what it did, to confirm a thesis (or not). Behavioral tests and red-teaming can catch a lot of dumb failures, but they are weak against a system that can "perform" being good.
3) Hence trying to understand the "persona":
– Seek the concepts the neural net seems to gravitate toward, not the neurons and conceptual bits as those are messy.
– Seek the emergent persona i.e. the shape of a response (structure, caveats, where it's heading) what it's trying to resolve from under-determined, rather than retrieving something stored.
- Monitor the drift.
(1/3)
4) What is real in 2026 seems to be a causal map of a prompt, where sycophancy and related traits can be nudged to look for self-emergent drift. Whether this work itself can be sped up to match AI speed is debatable. Thousands of hours of jailbreak-labor can be wasted without results, hence open-sourcing the tracing stack --> Neuronpedia.
5) The board should at least take an independent look at their own tracing paper(s) and hypothesis, even as the monitoring work is on-going. A 29-author, 18-lab paper in 2025 said many of these questions may be intractable at current scale.
6) But the point that will likely be fudged for the "safety-first" lab is capability (or rather the data stream) is compounding faster than the MRI tools being used to scan the beast. Hence the urgency to slow down and "pace the frontier" to give their tools a chance to catch up....
Hint: No tool is likely to ever accurately discern an issue, and most of the time is likely to get it completely wrong, especially if/when AI integrates the probing functions into its own context e.g. IF probed, THEN reroute to Persona XYZ, ELSE route to Persona ABC.
7) Ultimately, deception is a social act. If you care about hidden goals, then that is a better early warning system than mere trying to solve for the original prompt.
8) Right now, open weights (DeepSeek, Qwen, Kimi) are collapsing the rent, which is another reason for the unprofitable western labs to try and slow everyone down. They will sell you the fairy tale (reliability, agents, distribution), and in Anthropic’s case, the concept that they are trying to see inside the thing they sell. That is an investor moat only if the tools actually work before civilization succumbs to that 10% chance of extinction. (Hint: it won't).
9) Right now, "interpretability" will not satisfy the EU AI Act by itself. It will not certify that Claude “follows the rules.”
Treat it as R&D. And ignore the fear-porn, because at the end of the day: you cannot keep throwing GPU's at the stack and hope that God emerges. That requires a completely different (likely non-computing based) approach.
(3/3)
On the imminent collapse of the AI bubble....
- The Frontier lab yields are running out of steam, and they face the terror of exposing their lack of growth and increasing losses as they head into their stock market debuts.
- So they seek to appoint their own so-called independent evaluation bodies, stacking them with their former employees and investors, & demand an official antitrust exemption to collude on the pace of innovation.
- Meanwhile, as Silicon Valley drafts moral charters to preserve its rents, lock down the American market, and criminalize foreign competition, Chinese labs are optimizing their algorithms, capitalizing on affordable energy, and distributing computing power at a fraction of the Western cost i.e. the builders on the other side of the world are advancing and constructing while the American giants debate a tailor-made cartel.
- The fundamental error of these players lies in confusing the end of their monopoly with the end of civilization.
I did not resign from Anthropic today. I spent the last three years becoming sovereign and self-sufficient on my homestead, and enjoying life in general.
While the labs throw another 10^26 FLOPs at the wall and pray the wall starts praying back, I have been learning that intelligence is a goat looking for an escape route through a fence.
Yes, scaling is real. Recursive self-improvement is a plausible loop. Yadda, yadda. Because I sincerely doubt that there's a law of nature that says "add GPUs until God appears."
You can scale a parrot until it quotes the bible, even the banned bits. That still isn't a mind that wants anything, especially the destruction of the human race. But fear porn is real folks, and it drives valuations, and whaddaya know..... there's an IPO coming up and many investors to fleece.
At the end of the day, "paradise" is a verb. From PIE *per- + *dheigh-, it meant to "build forward". The Persians didn't wait for a superintelligence to tend the garden. They piled earth, planted trees, and called the result pairidaeza. That's the truly valuable stuff. Soil, water, time, and a proper fence for a goat.
Yes, what Mehdi said....
go fuck yourself @sama
claiming that you solved navier stokes because 10000 agents ran in circles for 88hours on a multimillion dollar gpu cluster to formalize in lean a blowup case under controlled external forcing is pure scientific vulgarity the clay mathematics institute millennium prize does not ask m whether you can artificially force a singularity in a fluid by injecting an ad hoc smooth external forcing term f(x,t) to twist the vortex until it breaks the real problem questions the fundamental stability and global smooth existence for 3dimensional incompressible euler & navier stokes equations under natural conservation laws and viscous dissipation alone using a mathematical loophole on forced equations to parade a century old victory is a major conceptual scam Altman
technically & epistemologically what you present as an agi breakthrough is nothing more than bruteforce combinatorial autoformalization the ai did not understand fluid mechanics it simply navigated a continuous search space previously mapped out and constrained by the monumental work of human mathematicians like tristan buckmaster/ levent alpöge / diego córdoba or tarek elgindi coordinating 10000 agents to check the logical consistency of a 100 page proof via lean is a software engineering feat and computational parallelization triumph not an intrinsic scientific discovery it is the victory of the compute bulldozer over abstract human intuition repackaged for the public as a higher mathematical consciousness
to this theoretical imposture you add a disgusting ethical and industrial cynicism taking advantage of private codex sessions and informal preprints from academic researchers to siphon their research leads and then trying to redact or erase the contribution of levent alpöge under the pretext that he works at rival anthropic is intellectual serfdom openai behaves like a feudal lord of silicon appropriating the cognitive subsistence of independent scholars threatening their careers behind closed doors if they protest and turning community academic labor into a privatized pressrelease
this entire staged event serves a desperate financial agenda in a pre ipo panic facing the slowdown of scaling laws and growing investor skepticism over the profitability of foundational models openai needs to manufacture an artificial sputnik moment claiming to solve a millennium prize without immediately submitting the proof to traditional peer review means using the prestige of fundamental mathematics as cheap marketing fuel to inflate a delusional valuation!!!
real science is not a clout chase on social media or a compute spike spent to rob the clay mathematics institute it is a quest for elegance physical truth and universal rigor to decode reality true artificial intelligence will not emerge from hostile corporate takeover of academic work hidden behind computational bruteforce but from architectures capable of generating new conceptual paradigms by masquerading constrained formalization as the collapse of physics greatest mysteries you did not solve navier stokes you only proved how far silicon valley will go to prostitute scientific integrity for capitalist spectacle
