Tim

@BowTiedLoon @MartinShkreli @xintanifae Pharma (and investors) don't profit from the grey market.

@MartinShkreli @xintanifae Why? (Honest question). There’s no public risk, as there are with things like meth.

suck my dick

@vitrupo His predictions are way too aggressive in the short term. LLMs have too many major shortcomings to automate jobs in the next couple years.

Emad Mostaque says the civil unrest may come before the job loss. Lawmakers are already getting death threats for approving data centers. “The math is really unpleasant.” “$60,000 all-in job gets replaced by $600. Next year it’ll be $60.”

@wokler_ @scaling01 Probably by far the best LLM model, but still with the many shortcomings of LLMs. Too expensive to release, so they made up a story about danger for hype and fundraising.

@scaling01 Anthropic Mythos will be a nothing burger until I see otherwise. I'm not believing any of the AI companies because all's they do is lie.

there's something weird going on I call it the "AI psychosis psychosis" Claude Mythos was literally the strongest proof that scaling AI models still works wonderfully, but these people are in complete denial they think everyone that is speaking highly of AI is in AI psychosis, but don't realize that they are in psychosis hence AI psychosis psychosis

@haider1 If it is truthful, the person it emotionally disturbed. This is just the next "what did Ilya see?" nonsense.

wtf is going on internally at anthropic? and by the way, this is coming from an openai researcher

@signulll Super intelligence just means greater than human intelligence. If the highest IQ human 400, the bar for super intelligence would be much higher.

humans are state of the art hardware running kinda ancient software. the wetware is absolutely extraordinary.. 86 billion neurons, all massively parallel, & energy efficient beyond anything we can remotely engineer. the hardware is genuinely best in class. so then why did super intelligence not naturally evolve within us?

@DaveShapi The model is too expensive to inference, but they wanted to use it to raise funds, so they made up a concern about security.

My personal view is that Anthropic has been ideologically captured by Effective Altruism and that the leadership of Anthropic explicitly and sincerely believes that: 1. Superintelligence is intrinsically uncontrollable. The best we can do is create it in such a way that it will be "friendly." 2. That they (and only they) are skilled and competent enough to land this outcome, and that they should have unique privileges and authority to shape this technology and humanity should just sit back and let them. 3. That "steering from within" via regulatory capture, integration with the establishment, and narrative control are the correct and appropriate tools, because they explicitly believe "democracy is too slow." That's what they believe, as far as I can tell, having studied Effective Altruism and AI Doomerism and the LessWrong "rationalism" for the last few years.

@Mayhem4Markets yep, obviously true

"Mythos is too dangerous to release to the public" Or, perhaps, more accurately, Anthropic is too compute-starved to serve a 10T parameter model at any reasonable scale. Hence the hype fest. That's not to say it isn't an impressive model. But it's not as powerful as promoted.

@wintonARK If it isn't economically viable and thus not released to the public, it doesn't count.

re: Anthropic's Mythos Yes, its a substantial advance, though roughly just a year ahead of what one would naively anticipate. People continue to over-index on current capability rather than anticipating continued advances. The entire compute demand curve is misunderstood and underestimated.

@Alexsei88 The projections are way off (revenue, profit, EPS) based on the assumed APT and production by year (even with payable % of APT and fx). They're way too low. The near term upside is much higher than shown. 65,000 people saw this so far, which is annoying.

Huomasin nyt vasta katsoa paremmin näitä kuvia. EQR:stä saadaan tehtyä näin bulli laskelma, vaikka tuotantokapasiteetissa on varaa parantaa vielä merkittävästi, esim. nostamalla syöttöä, recoveryä tai gradea. Loistava laskelma, eikä ole mielestäni mitenkään yliampuva. #Tungsten

@BULLOFBRITAIN There are errors resulting in significant underestimations (don't trust LLMs to do financials). Based on current APT prices and 2,500t production, FWD PE is ~3.5. If APT prices rise (they've been rising a crazy 7% per week). The FWD PE drops. Huge near-term upside.

$EQR.AX Updated - Final model with 3,000% upside I spent 100+ hours building what I think is the most detailed $EQR.AX model in existence. 820,000 words of research. 13 sheets. 129 live formulas. Every input traced to a CEO quote, SEC filing, or customs document. The price projection based on CURRENT trends since 2025, projected scaling, all known catalysts, comments from every known Tungsten/industry professional with a 3-5 year timeline. The earnings table showing A$66M revenue transforming into A$1.5B at peak. The stock price scenario table from A$0.30 to A$4.50 base / A$12.00 Nasdaq. The probability-weighted EV of A$2.54 in the green box. And the model stats bar at the bottom. I've used all relatable Nasdaq uplisted Tungsten companies and their respective average volume increases and applied it to $EQR.AX as an industry standard on where it would put the stock relative to its intrinsic value. This also includes other industry bottlenecks and what the market prices their premiums as such. It has transcipts from CEO Craig Bradshaw in various seminars, youtube videos and earnings reports. This includes sections of Production, Contracts and Commercial info, Balance sheet / Corporate speak, Exploration and expansion (such as Wolfram), CEO sourced Market Data and Competitor data (ALM, China buying African/Bolivia mines to squeeze the West more) Lastly Third-party data from industry professionals and community members such as @Alexsei88, @UnnajaSnoop, @METhompson72, @el_miikka, @Fibonazzo, @AdventusCEO and @88_becks. Why the Western supply demand is broken and why Tungsten has no real ceiling based on price inelasticity. The West has completly relied on the back of Chinese material pricing and supply for so long, that it got so complacent that it shut down its majority of domestic supply. These are things such as exploration, funding and governmental backing, which have stemmed from net-zero and eco friendly narratives. Now China are shutting the tank, the West has finally realises the situation it has created. They had no substitute for this outcome from sheer ego and lack of intuition to pursuit social issues rather than physical. This entire situation has left a gap which can be mathematically obvious if only you just look at a few numbers. TOTAL WESTERN SUPPLY 2026: 17,940 2027: 22,100 2028: 26,000 2029: 31,000 2030: 34,300 TOTAL WESTERN DEMAND 2026: 24,900 2027: 31,100 2028: 34,900 2029: 37,250 2030: 40,200 Defecit: 28% Tungsten market is growing by a CAGR of 8%. Western defence being 7-8%. AI being 19-30%. CAGR of production volume? 4.7% to 4.9%. If you know anything about compounding, you can see where this is going. Supply EQR (Australia + Spain) ALM — Panasqueira (Portugal) ALM — Sangdong Ph1 (S. Korea) ALM — Sangdong Ph2 (S. Korea) ALM — Gentung (Montana) Wolfram Bergbau (Austria) Group 6 Metals (Tasmania) Tungsten West (UK — Hemerdon) Guardian / BMM (USA) Mactung (Canada — Fireweed) Recycling (Western) Vietnam (Masan — grey zone) End-use and % of demand Defence / munitions (11-14%): W = 10-20% of shell cost. NDAA mandates procurement regardless of price. Zero recycling. 1% military increase = 2% market increase because munitions aren't recycled. Oil & gas drilling (10%): W cost = 0.2-0.5% of driller profits at $300. At $3,000 = 2-5%. But oil doubled too, offsetting. No drill hole without tungsten carbide. Semiconductors — WF6 (3-4%): WF6 per wafer = tiny vs $20K+ wafer value. 200-layer NAND = 200 depositions. Samsung reportedly out by June 2026. No substitute for CVD tungsten in sub-22nm. Semiconductors — wire cutting: Tungsten wire yields more chips per wafer than diamond wire. Less swarf, thinner kerf. PV industry switched because economics were better even BEFORE the price spike. CNC machine inserts (25%): Insert = fraction of $500/hr machine rate. Nobody stops a factory over a $5 insert becoming a $50 insert. Redesign to non-W takes 3-5 years. Nuclear fusion (new): ITER divertor = 432 tonnes of tungsten. Government-funded. No substitute — only material surviving plasma-facing conditions. ITER started series production. DEMO, STEP, SPARC follow. PV tungsten wire (new, +198%): 4,500t demand in one year from near-zero. 20%→60% penetration. Solar deployment doesn't stop for tungsten prices, it's mandated by climate policy. Li-tungstate batteries (new, +300%): 12,000t demand. Solid-state battery core material. EV transition depends on it. Li₂WO₄ as electrolyte additive. Mining & construction (26%): Every drill hole needs tungsten carbide bits. Every excavator tooth. Every crusher liner. Mining doesn't stop because tungsten is expensive, it's the input cost of extracting everything else. Catalysts and timelines 2026 May-June: Samsung WF6 supply reportedly exhausted. Semiconductor tungsten crisis headlines. June: EQR debt fully repaid. Bradshaw: 'Expect to generate significant cash. July: Mt Carbine expansion board approval + construction begins. A$30M spend, equipment purchased. Sep-Oct: H2 FY2026 full earnings. EBITDA massively positive. Revenue A$200-350M+ for H2. Sep-Oct 2026: Bank coverage initiations. BofA did ALM in March, EQR is the logical next call. 2027 Nov 26-Feb 27: Mt Carbine expansion operational. Production steps from 2,500→3,500-4,000t. Jan 1, 2027: NDAA goes live. Chinese tungsten BANNED from all US defence contracts. No workaround. H1: AVDS/Cadence/institutional accumulation. ETF inclusion. Possible analyst day. Possible Nasdaq/NYSE uplisting announcement. GMTL +43% in 2 weeks on listing. ALM +200%. H2: Peak scarcity. Defence + semis + fusion. Thompson: 'rhodium-like move'. APT $6-10K+. 2028 New supply: ALM Ph2, TUN, Fireweed/Mactung beginning. Recycling at 6-7K tonnes. Takeover risk. Major miner or defence prime bids for EQR. When supply chain = 2 companies, buy them. 2030 Fusion fleet procurement (DEMO, STEP, SPARC). Permanent demand Pool 3. The general consensus from all the talks I've had with industry professionals and rare metal investors gives a timeline of 3-5 years minimum to re-engineer away from W in any application. Most applications have NO substitute at all. Companies will pay anything for 3-5 years rather than redesign. (AI Capex, 1.5 Trillion U.S. Defence spending) Once again. This entire post has been modelling on a framework of CURRENT day trends. Any fluctuation upward or downward will compound these results massively and will vary differently. Thank you for reading and enjoy the gains. This will most likely be my last long form Tungsten post unless anything major happens. Let's ride. 💥#Tungsten 🍀 NFA. DYOR.

@AbHomineDeus You're missing the point. Scaling LLMs will not fix their many deficiencies. They're already jagged/narrow ASIs in some respects, but that doesn't make them AGI.

Because they're going to blow right past it. By the time we have something that's agreed to be AGI it'll be ASI. The anti-AI psychosis is brutal.

@CyberrBandera @BULLOFBRITAIN Being that China is short tungsten (their mines are depleting and they are net buyers, to my knowledge) and it is a critical mineral to military (China/US security competition), it seems extremely unlikely they'll lift export restrictions.

@BULLOFBRITAIN Have you considered that China may lift the export controls on tungsten as part of broader deal with US? Trump is coming to China soon, so that is something on my mind.

@mreflow @MatthewBerman I think there is a strange incentive to claim psychological distress. We still haven't solved the many weaknesses of LLMs (hallucinations, novel problem solving, subjective taste, world modelling, etc.), so this just the next incremental step. Still a long ways from AGI.

@MatthewBerman Here's an article from 2019, 7-years ago... The gist of it: "GPT-2 is too dangerous to release to the public" slate.com/technology/201…

i'm on vacation with my family. i read about mythos and couldn't relax the rest of the day. i am completely stunned. i already have a severe case of ai psychosis. i dont know what to call this now. i'm up late right now (late for me). i can't stop reading about anthropic's new model that they can't even release publicly because it's so good. this feels different. words like "frightening" and "uneasy" and "scary" are being throw around by the anthropic team. i feel all of those things. i knew this moment was coming. i didn't know it'd be so soon. i'm generally optimistic. i don't feel as optimistic today. i was shell-shocked most of the day. my mind was stuck on it. i kept looking around at people enjoying their vacations with their families and...i just felt weird. like i had been told aliens are real, they're coming, and soon...and no one else knows. it's true though, practically no one knows what's happening in AI right now. where does this go from here? how quickly? is software solved? is all software vulnerable now? am i even asking the right questions? what about anthropic? this is an enormous amount of power for one company, one man (dario), to have. i've said this before but now it's more real than ever: can any company catch up to anthropic? opus likely helped build mythos, mythos will help build the next model after that. recursive self improvement is here. the "intelligence explosion" as leopold aschenbrenner put it, is here. i knew the frontier labs were racing towards ASI. i knew it. but i didn't fully grasp what it meant. the first company to reach it wins. period. full stop. nothing else matters. dario knew that and his bet on coding was right. on the one hand, imaging all science, math, coding, climate problems being solved. imagine cancer being cured. imagine going to the stars. on the other hand - imagine concentration of power, political and economic change happening so fast, society can't adapt. how do we go on like things are the same?

@atmoio @chebibu @stephen_wolfram People confuse intelligence with consciousness or think consciousness emerges from intelligence. Obliviously, this is wrong. Bugs are conscious.

@chebibu @stephen_wolfram you can’t look at some alien primordial technology you don’t understand (carbon consciousness), pick apart a few constituents (matter), then say, hey look, anything with matter is also this same stuffs! it’s preposterous. primordial technology != human 21st century technology

Claude Mythos is Delusional

@BecomingCritter It absolutely is about hype. Model companies have been "leaking" internal AGI for years and raised massive amounts of cash in the process. Remember the "what did Ilya see" nonsense?

Some are saying that pretending like your frontier model is extremely dangerous is a stunt to draw enterprise sales. I strongly doubt this is the case with Mythos Did it find 10k firefox vulnerabilities or not? Did it find decades old exploits or not?

@haider1 All of them have a 0% of chance of being conscious as will Claude 10 unless it is specifically being targetted. Consciousness does not emerge from intelligence. Bugs are conscious.

quick questions: if anthropic already puts opus 4.6 at a "20%" chance of being conscious, where does mythos score on that eval? and if gpt-5.4 and opus 4.6 are already helping with phd-level research alongside people like terence tao, what will spud and mythos be capable of?

@adamtaggart @Haymaker_0 Well, I didn't sell.

Was speaking with @Haymaker_0 about the fragility of the Iran ceasefire He raises a good question: If you don't expect it to last, should you be thinking about picking up energy stocks? (especially after today's big drop)

@sporadica Yes, you are wrong. Until hallucinations are novel problem solving are solved (they haven't been solved), you will not see large economic impacts.

really interesting "boy who cried wolf" thing happening in the AI space rn yea, a lot of earlier releases (GPT-3, o1, o3, Opus, etc) were met with insane fanfare (and crazy dooming, thanks yudkowksy and co) and so far have changed little. i myself have been responsible for some of that insane fanfare, too, i know it but now people are so tired of the hype they will instinctually gut react that any major step change in capabilities is really a nothing-burger. the AI hype-ists have cried wolf too many times apparently. i don't really know what to make of this, but i'll be honest that to me this time feels different. maybe i'm wrong. time will tell.

@Rory_Johnston were they actually passing through?

🤷🏻‍♂️

@_tenZdhon_ Why would they? It's an internal model.

No one on LinkedIn is talking about Claude Mythos