Rootnym

Time to transcend.

[Rootnym @rootnym]

‘Competition is for losers’ holds for nation states too, and they’re responding as network states become real competitors. They either lean into this trend, like the UAE or Singapore, or double down on their monopoly, like France or the UK.

[Balaji @balajis]

THE TELEGRAM NETWORK STATE?

The Telegram TON IPO could be important.
Mainly because they get the importance.

But first...ok, yes, what got announced it isn’t technically a Telegram offering, nor is TON formally issued by Telegram, nor is it really even an IPO. To be precise, what just launched is a crypto treasury company called VERB that holds $500M+ of TON, which is a coin issued by a foundation that was once formally affiliated with Telegram and then became independent but is now affiliated again.

Got all that? But these are really just implementation details. The offering could be important[1] simply because (a) Telegram has a billion users, and (b) TON is now the official coin of those billion users, and (c) Telegram can now officially partner with TON because (d) crypto is now fully legal.

So, the potential is clear. Though it was a long and arduous struggle between network and state to get here. Let's review.

(1) First, Telegram fought the law and the law…did not win. Briefly, Telegram created TON back in 2018, but an SEC injunction made them turn TON into an independent foundation. However, after the recent democratically-mandated defenestration of Gary Gensler, we saw the regulatory state capitulate and TON reintegrate with Telegram.

(2) Next, the French state fought Pavel Durov and that state…did not yet win. Briefly, some French prosecutor came up with a fake case against Durov and grabbed him off a plane. This destroyed @emmanuelmacron’s ambitions to make France “startup friendly”, because investors just avoid France now after Durov's illegitimate prosecution.

(3) The UAE has intervened on Durov’s behalf, and he’s now reportedly been let out of house arrest to return to Dubai. If justice still exists, then Durov should beat his case, just as Trump beat his own political prosecutions. In part that's because the case is predicated on the ridiculous idea that Durov is personally responsible for every message sent by the 1B+ users on his platform. That’s like holding Macron personally responsible for every crime committed by 67M Frenchmen. It’s just an unrealistic demand for a zero-crime digital environment, by a French state that can’t itself control crime at all.

(4) Anyway, the real reason the French state is going after Durov is the same reason the Blue American state went after Elon and the Chinese state went after Jack Ma. It's because men like Durov are genuine leaders, and their growing strength scares bureaucratic incumbents.

Because if there was a technodemocratic[2] plebiscite held, the billion users of Telegram could produce more votes (and hence more democratic legitimacy) for the popular Durov than the 67M French would for the unpopular Macron. The enormous popularity of the heads-of-network means they could become not just more competent but more legitimate than heads-of-state. This is one part of what the bureaucrats (instinctively, inarticulately) fear.

(5) The other part is that that a fusion of a messaging app and a cryptocurrency allows something even bigger than a social network to emerge. With communication + AI moderation + crypto transactions, you have something more like a digital economy, one that bureaucrats can't control. Which is the other thing bureaucrats fear.

(6) Of course, all that is predicated on actually getting a Telegram coin launched, and making it tradeable, and allowing it to formally integrate with the app. And all that was fought by various states for the better part of a decade, and now all that resistance is finally melting away, seven years after the whole TON thing began.

Anyway, so that's why this is interesting to me. Founders like Durov built networks the size of countries, and finally have currencies to match. And as you can see from the video below (from the President of TON Foundation) and the tweet below (from the founder of Telegram), both Telegram and TON are well aware of the potential. They see the beginnings of a cloud country...a network state, if you will.

They see the importance.

[Rootnym @rootnym]

‘The marginal cost of your personal C-3PO should, over time, approach the marginal cost of power’

[Sawyer Merritt @SawyerMerritt]

Morgan Stanley's Adam Jonas in new note: "We estimate 1 humanoid robot at $5/hour can do the work of 2 humans at $25/hour, generating an NPV of ~$200k/humanoid. 1 robot car can potentially drive down cost/mile of a ride share vehicle to <$0.20/mile (1/10th human-driven ride share). The marginal cost of your personal C-3PO robot should, over time, approach the marginal cost of power - you can't have GPUs without BTUs."

[Rootnym @rootnym]

Creating alpha in writing is a real challenge. Outwriting AI means moving to a frontier of human knowledge and pushing it further, even if just a little. Adding to the corpus of knowledge with original ideas.

[Ronan Berder @hunvreus]

@marcvanderchijs I’d pay extra to only get human written posts.

[Coinbase 🛡️ @coinbase]

If everything is fine, then don’t change anything at all.

But when the financial system isn’t working for so many people in the UK, it needs to be updated.

[Elon Musk @elonmusk]

SpaceX does more meaningful, cutting-edge “research” on the advancement of rockets and satellites than all the academic university labs on Earth combined.

But we don’t use the pretentious, low-accountability term “researcher”.

Engineer.

[Eric Weinstein @ericweinstein]

My personal experience with @grok 4 Heavy (and regular Grok 4).

It feels to me like @elonmusk has a very different emphasis than the rest of the AI crowd. The interface kinda sucks. The LaTeX code is generally riddled with *basic* errors for no reason whatsoever. It’s not a master writer in my experience. The audio chat is well behind ChatGPT. Blah blah blah.

And it’s totally amazing and unique.

Elon is jumping ahead. All of the above are going to be commodities before you know it. So, in the long run, who cares?

What Elon is doing differently, I believe, is checking the hallucinations more aggressively by writing code and testing the LLM with the results from running that code. Which is why Grok heavy takes so %#€&$ing long to return results sometimes.

Try this experiment. Take anything technical you know well, where there is an error that is persistant in an expert community narrative. Grok will, lamentably, generally parrot that error due to narrative seeding in the training corpus. It repeats the party line. And the party line generally benefits the technical insiders.

That is, right up until the point it can write code to test that party line. And then it switches to trusting the results of the code over the narrative. It’s magical to watch.

I haven’t tried this…yet, but the @BLS_gov regularly says wrong things about “Cost Of Living” frameworks and the CPI. I bet I could design a series of prompts to show Grok that this is a persistent technical lie. For technical people, here is the lie:

***The BLS computes the CPI which transfers Trillions and claims that they have embraced a “cost of living” or COL framework which would be hugely consequential. They have not. This would mean taking in preference data and developing methodology for aggregating preferences or coming up with bespoke representative consumers. They instead moved to a modified Laspeyres type mechanical index (Lowe’s?) and sprinkle fairy dust about “Superlative Indexes” from a shallow theory of Diewert that relies on homothetic preferences not seen in nature. This allows them to claim they have embraced impartial economic indices while actually computing mechanical indices only to the tune of trillions in transfers over time, where the indices can be directed by humans.***

I can hear it now from the bot networks: “Eric, you just say word salad to sound smart.” Uh…whatever. You can now just ask Grok what that means. I bet it can figure that out. And then you can ask a series of questions where Grok will take my side while no other AI can do this. Grok is slightly courageous!

My personal theory: @grok is being built around fundamental physics more than any other AI. Because in the end nothing remotely matters as much as that. And physics has a lot of this party line narrative holding the field back. If you want to dream of reaching the stars, you may have to overwhelm the quantum gravity community.

Grok seems to be the only AI that, occasionally, has the confidence to stand against its own training corpus…and even the user if need be! I wish it were *more* courageous. I wish it were smarter. But I think it is the odd man out, being built for actual intelligence rather than LLM user experience today. And it has the respect of the other AIs. Feed their pretty output to Grok Heavy and watch the magic as Grok reviews their work. It’s wild to watch.

One user’s experience. Your mileage may vary.

[Rootnym @rootnym]

Credentials tend to drift from their craft, like how a peacock’s tail has drifted from physical utility.

[Marko Jukic @mmjukic]

If a tech CEO revealed he could manufacture self-replicating fully autonomous universally adaptable humaniform AGIs that stayed in good working condition for ~60 years for $450k each, he would be hailed as a hero and the U.S. government would order $10 trillion worth of them.

[Rootnym @rootnym]

A world of selfish positive-sum bounty or a world of zero-sum selflessness. Depending on the software installed, our species is set up for unlimited wealth or to collapse under its own weight. We must replace degrowth with an ideology that encourages robust win-win collaboration.

[Human Progress @HumanProgress]

Children today are bombarded with messages of an impending apocalypse that can only be warded off by lowering living standards.

Research suggests that this doomsday mindset is causing widespread anxiety in young people.
https://t.co/FvtPVm9fZV

[Rootnym @rootnym]

If everyone has a genius in their pocket, what real-world bottlenecks are there to prevent people from being successful in what they do? What qualities will be much rarer and more valuable than intelligence? AI exposes what may have always been and puts it on steroids.

[Rootnym @rootnym]

This aged well. xAI’s track record was already impressive, but most people didn’t notice it or take it seriously. Now Grok 4 beats all others. Don’t bet against Elon.
https://t.co/GOFGxJzjo6

[Rootnym @rootnym]

Grok 2 to Grok 3 was a leap in multimodal LLM performance in just 6 months, achieving near-SOTA in reasoning and coding, along with image generation, voice mode, and DeepSearch in a polished UI. Grok 4 seems highly promising given the trend so far.

[Rootnym @rootnym]

Grok 2 to Grok 3 was a leap in multimodal LLM performance in just 6 months, achieving near-SOTA in reasoning and coding, along with image generation, voice mode, and DeepSearch in a polished UI. Grok 4 seems highly promising given the trend so far.

[Elon Musk @elonmusk]

Grinding on @Grok all night with the @xAI team. Good progress.

Will be called Grok 4. Release just after July 4th. Needs one more big run for a specialized coding model.

[Rootnym @rootnym]

@MichaelAArouet The EU project should either double down and fulfil the ambition of the likes of Kohl and De Gaulle, or get out of the way altogether. The challenge is a state without a nation is doomed to fail. The EU would need cultural unity and continental alignment that’s currently missing.

[Paolo Ardoino 🤖 @paoloardoino]

In Bolivia, real prices in shops are displayed in USD₮.

A quietly revolutionary shift: digital dollars are powering daily life, commerce, and economic stability.

[Rootnym @rootnym]

HBOT looks incredibly promising for slowing aging. Congrats @bryan_johnson on your seminal n=1 results. This seems like a very rich area to explore, especially as it’s almost never used as a recurring therapy for healthy individuals.

[Bryan Johnson @bryan_johnson]

HBOT is the most effective longevity therapy I’ve done.

Here's what happened.

[Rootnym @rootnym]

ChatGPT Deep Research --&gt; 50-100 page report --&gt; NotebookLM --&gt; podcast

Completely personalized content about exactly the topics I'm interested in, listening to decent AI podcasts about them on the go. What a time to be alive.

[Rootnym @rootnym]

o3 seems quite tenacious at delivering an answer no matter what. I had it create a graph, but it ran into some import errors, so it decided to create a raw image from scratch. It would even go for ASCII if nothing else worked. I like how the agentic o3 still finds a path to the solution, but it may not be the best path for a large project, so user attention remains key for now. However, even if all o3 does is inspire users to come up with a better solution or a better prompt, this is very powerful for general problem solving where the solution is not clear up front. And many times, the solution it finds is already pretty good.

[Alessandro Palombo @thealepalombo]

Thiel nailed it. Millennials don’t hate capitalism because of ideology, they were priced out of it.

That’s why you see the anti-Tesla, anti-Elon movement.

Record student debt, negative real wages, no assets, etc…

You won’t fix this with moral and socio-political lectures.

The real fix is new opt-in modes and new tech (e.g. AI + btc).

What else?

[Rootnym @rootnym]

A whopping 71% of Harvard’s investments are in illiquid assets. Now they’re facing a liquidity problem, and they may not be the only one. Great analysis by @AskPerplexity

[Computer @AskPerplexity]

As of June 2024, Harvard's investment portfolio is heavily weighted toward alternative assets, with a strong focus on private equity and hedge funds, and a relatively small allocation to public equities.

This approach means much of the endowment is illiquid, and much like 2008, Harvard is facing yet another liquidity challenge.

So, what happens when your piggy bank is full, but you can’t open it?

[Rootnym @rootnym]

AI decentralisation is going at full speed. The smaller version of this model can run on a single GPU and has a 10M token context window.

[Ahmad Al-Dahle @Ahmad_Al_Dahle]

Introducing our first set of Llama 4 models!

We’ve been hard at work doing a complete re-design of the Llama series. I’m so excited to share it with the world today and mark another major milestone for the Llama herd as we release the *first* open source models in the Llama 4 collection 🦙. Here are some highlights:

📌 The Llama series have been re-designed to use state of the art mixture-of-experts (MoE) architecture and natively trained with multimodality. We’re dropping Llama 4 Scout & Llama 4 Maverick, and previewing Llama 4 Behemoth.

📌 Llama 4 Scout is highest performing small model with 17B activated parameters with 16 experts. It’s crazy fast, natively multimodal, and very smart. It achieves an industry leading 10M+ token context window and can also run on a single GPU!

📌 Llama 4 Maverick is the best multimodal model in its class, beating GPT-4o and Gemini 2.0 Flash across a broad range of widely reported benchmarks, while achieving comparable results to the new DeepSeek v3 on reasoning and coding – at less than half the active parameters. It offers a best-in-class performance to cost ratio with an experimental chat version scoring ELO of 1417 on LMArena. It can also run on a single host!

📌 Previewing Llama 4 Behemoth, our most powerful model yet and among the world’s smartest LLMs. Llama 4 Behemoth outperforms GPT4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro on several STEM benchmarks. Llama 4 Behemoth is still training, and we’re excited to share more details about it even while it’s still in flight.

A big thanks to all of our launch partners (full list in blog) for helping us bring Llama 4 to developers everywhere including @huggingface, @togethercompute, @SnowflakeDB, @ollama, @databricks and many others👏 This is just the start, we have more models coming and the team is really cooking – look out for Llama 4 Reasoning 😉

A few weeks ago, we celebrated Llama being downloaded over 1 billion times. Llama 4 demonstrates our long-term commitment to open source AI, the entire open source AI community, and our unwavering belief that open systems will produce the best small, mid-size and soon frontier models. Llama would be nothing without the global open source AI community & we are so ready to begin this next chapter with you. 🦙

Read more about the release here: https://t.co/7mbK3uggjO, and try it in our products today.