Jon Rotbard retweeted
“We want them to be able to trade these names that are actually creating an impact.” - @theblackmanda
@jonrotbard sat down with @theblackmanda and @ichooserain from @mochatradeapp to discuss how Mochatrade democratizes opportunity for Indian investors.
We just led Mochatrade’s $5.5M seed. Read our thesis below ↓
nitter.cf/jonrotbard/status/2114…
Jon Rotbard retweeted
i agree with ryan and think there's a decent chance they leapfrog polymarket too
Jon Rotbard retweeted
We're excited to release OpenInstinct
Instinct but self-hostable, open source and supports every model. You control the code, data and passwords
Deploy on Vercel and never use the internet with your fingers again
github.com/Merit-Systems/Ope…
Jon Rotbard retweeted
all of fintech is becoming commoditized.
today you can launch a fully functional neobank by stitching together a handful of APIs:
> @dakota_xyz (global banking, virtual accounts & cards)
> @blend_money (yield & treasury infra)
> @moonpay (crypto rails & on/off ramps)
> @privy_io / @turnkeyhq (embedded wallets)
> @tryPortola (stablecoin credit as an API)
> @raincards (card issuance)
for well under ~ $60k, a small team can assemble infrastructure that would've taken years and millions of dollars to build from scratch just a few years ago.
the hard part is no longer building banking products.
as the underlying technology becomes more accessible, the moat shifts elsewhere:
the winners over the next decade won't necessarily have the most unique APIs.
they'll have:
> the best user experience
> the strongest distribution
> an obsession with customer support
> a brand people genuinely trust and want to be associated with
if you're building another generic B2B or B2C fintech with the exact same banking stack as everyone else, ask yourself:
why should a customer choose you over the other 200 companies built on the same APIs?
the answer usually lies between brand, experience & unique distribution.
Jon Rotbard retweeted
Unlike many investors in crypto, I did not pivot to AI in the last few years. However, since 2020, I built some of the deepest understanding in this industry on the intersection of AI and decentralized networks (crypto, web3).
From the start, it was very clear that AI models are a centralizing force and the biggest target for government control. That point became market fact last night, with @AnthropicAI’s export control compliance.
As an investor in decentralized AI, I know that d-networks are a counterbalance to this state of affairs. In particular, the starting point of sovereign, open, public, decentralized AI is the seemingly insurmountable compute problem.
How are people supposed to source more industrial compute for frontier training than these huge trillion dollar companies? The answer is simple: there is enough commodity GPU compute in the world to compete on the frontier, but to make use of it we need new algorithms for training.
That’s what a few companies like @gensynai @PrimeIntellect @bageldotcom @Pluralis @NousResearch @MacrocosmosAI @covenant_ai set out to research, while everyone on the planet told them it was impossible.
The result is that it is not only possible, but it can be cheaper and nearly as efficient as the alternative process.
The second major problem is economic sustainability. Open source models are great, however, they are not economically viable as they don’t have a business model. So far in decentralized AI, only @Pluralis has an answer — by breaking up the weights of the model among participants, we create a business model for tokenized AI models.
This is the moment of truth — will AI become fully centralized and fall under censorship and unilateral government control? Or will the AI world realize the importance of public AI on open decentralized networks?
Jon Rotbard retweeted
Systems like this must exist. This is the way out.
The 8B model currently training on Agora is 350B tokens in and continuing to converge. The top level metrics and evals look almost exactly like a centralised run. But;
- 133 external contributors total bringing 4090's, 5090's, L40S/RTX 6000 and RTX 6000 Pros. These are cards that people actually own - there are no H100, B200's etc.
- The max number of nodes the system can support (104) was filled almost immediately. The authorization layer is receiving approximately 100 requests/minute to join.
- The total tokens/per second processed moves directly with amount of compute in the swarm, with Agora constantly optimising to make most efficient use of what hardware is present.
- MFU is approximately 20%, TPS is 170k tok/s. There are near constant communication failures which Agora is completely absorbing without slowdown.
- The system is effectively on auto-pilot, requiring very little intervention from us. Bad nodes are purged immediately before training is affected and new nodes take their place.
Jon Rotbard retweeted
Today we're releasing Agora: the first ever pretraining stack that allows non-collocated consumer GPUs to be competitive with centralized clusters
Agora is 15x faster than Megatron-LM in this setting and is only 1.5x less efficient in terms of tokens per unit compute than TorchTitan on H100s, despite running on devices that have no NVLink or InfiniBand support.
Jon Rotbard retweeted
There's a lot of really exciting developments happening in decentralized AI training this year. Here's my take on why decentralized training is moving from "impossible" to "investable". 🧵👇
Jon Rotbard retweeted
NEWS: Kraken DeFi Earn has surpassed $100 million in total deposits.
It’s never been easier to earn on idle assets.
Powered by 3 Veda vaults:
Balanced + Boosted → curated by @chaoslabs
Advanced → curated by @sentoraHQ
Available on @krakenfx, @krakenpro, and @krak
Jon Rotbard retweeted
We’re making DeFi rewards simple.
Meet DeFi Earn, powered by @Veda_labs, @SentoraHQ and @ChaosLabs.
Earn up to 8% APY on your assets all within the Kraken app experience you already know & trust.
No complicated setup. No technical steps. Just start earning ⤵️
app.kraken.com/JDNW/defiearn
Jon Rotbard retweeted
This is directionally correct.
BUT — to say that blockchains are just for finance is too categorical. It may just be a question of timing. All those other use cases could come back down the line.
Famously, almost every idea that failed in the run up to the dot-com crash—online retail, grocery delivery, video broadcasting, and even digital currencies—ended up succeeding eventually.
The biggest problem with consumer crypto, for example, has been that the market is too small. The number of of people who use crypto wallets day-to-day is a rounding error compared to the scale of the internet. Consumer crypto has not yet had a real shot.
The real test might only happen once stablecoins/defi bring hundreds of millions or billions of users onchain.
It’s cool too see Vitalik clearly articulate his vision for Ethereum (now better than ever maybe?)
Personally, these ideas were riveting in 2014. It’s what inspired me to get into crypto. But today it feels like the world has moved on, unfortunately.
Blockchains are for finance. Finance is and will continue to get surprisingly more expressive because of blockchains, but it’s hard to see a mainstream renaissance in the types of applications described because they are skeuomorphic (decentralized) versions of existing products. That hasn’t been a formula for success.
I think a more interesting theme is to imagine: non-skeuomorphic manifestations of the vision he is describing?
For example, are we going to end up with 3 mega-datacenter corps, or are we going to find a way to leverage decentralized compute? And what *new* products can only be developed because of the new financial markets that blockchains enable?
Jon Rotbard retweeted
no crypto startup should feel pressured to do buybacks. in normal tech this would be insane. high growth startups should be reinvesting capital and preserving runway, not blowing it in the open market to placate a couple loud reply guys
the popularity of buybacks - and in some sense, the expectation by the market that every revenue positive protocol will do them - is unequivocally bad for the crypto startup ecosystem. founders should be focused on their users, product, and team - not managing a perpetual twap
buybacks are a symptom of a more fundamental malaise we find ourselves in as an industry. most tokens are worthless. the average app token is just a memecoin wearing a suit. it has “utility” but no ownership of anything meaningful
token holders have finally grown wise to these schemes and now as a result demand buybacks. why be forced to trust the team to steward a protocol’s capital when they can just immediately return it to investors via buybacks? the shorter the lag between revenue in and buybacks out, the better
this approach is fine for later stage teams. buybacks make sense for protocols like pump & hyperliquid that print money and may have no other productive place to park all that cash. but for your run of the mill crypto startup, buybacks are at best a distraction and at worse a waste of precious resources
the pendulum is finally starting to swing away from buybacks, as evidenced by helium’s decision to pause part of their buyback programme. metadao will only accelerate this shift
teams that launch ownership coins on metadao feel no pressure to do buybacks bc investors have actual ownership. this means ownership coin founders don’t need to virtue signal to their holders; instead they can focus on the important stuff - product, users, team
buybacks are a pre futarchy phenomenon
Jon Rotbard retweeted
Patrick (and by extension, @nic_carter) are correct. Stablecoin issuers will end up paying out all of their yield, one way or another. This will put a lot of pressure on banks that don't. They'll have to pay up.
The fact that banks don't pay competitive interest on deposits was never a pre-ordained optimal design for the economy. It was more of an accident of history, fueled by the lack of better technology and regulatory moats. Today, it's more of a policy decision.
The banking lobby argues this impedes credit creation, but banks are not non-profits. They make money from not paying more on deposits. So bank profits will have to go down.
The world doesn't yet recognize how economically impactful such a change will be. In its most recent earnings, JPM forecast $95 billion in net-interest margin income for the next 12 months. That's more than half of its 2024 revenues.
BoAs NIM is in the $60 billion range, Citi somewhat lower. Not all of it will disappear due to stablecoins, but most might, eventually.
Put together, I'd estimate a quarter trillion dollars that may someday not go to bank shareholders. The world is not ready for the economic impact of this - we are talking some of our largest and most profitable corporations.
More importantly, the quarter trillion that doesn't go to banks will go to corporations and individuals, the depositors. Imagine the economic impact of this - a massive stimulus if there ever was one.
But of course stablecoins aren't just desirable because they'll pay yield, they also offer instant settlement. The eventual disappearance of delayed settlement is another way stables will deprive banks and other payment providers--including Stripe--of income and give it back to customers.
For example, the US Government pays around $1.5 trilloin in social security benefits annually. Most of it goes out via ACH (T+2 business days), some of it via even slower methods like prepaid debit cards and checks.
That's $1.5 trillion dollars, the yield on which does not go to the US government or retirees, for several days a year. We are talking billions in profit for the banking system and card issuers.
Same goes for the even larger amount paid by all corporations in payroll, the lost float on your rent payment, wires that take 3 days to settle, etc etc.
The current architecture of banking and payments is predicated on someone else getting the yield on your money for some period of time, and charging you transaction fees on top of it for the privilege of not getting yield. Many FinTechs offer faster settlement, but for an additional fee.
The payments industry, which includes banks, has over $3 trillion in annual revenues. Thanks to stablecoins, but really public permissionless blockchains, some substantial percentage of that will someday stop going to the payments industry and instead go back to payers and payees.
One of the largest sources of economic rent, eliminated.
The world is not ready for this. But it will be glorious.
Good post on evolving stablecoin market structure. I would extend it further: yes, I think that stablecoin issuers are going to have to share yield with others, but this is just one instance. Everyone is going to have to share yield. Today, the average interest on US savings deposits is 0.40% (FDIC data), and $4T of US bank deposits earn 0% interest.* Things aren't better in the EU: 0.25% average interest on non-corporate deposits; corporate deposits just 0.51%.** In my view, this is going to change: depositors are going to (and should!) earn something closer to a market return on their capital.
(Some lobbies are currently pushing, post-GENIUS, to further restrict any kinds of rewards associated with stablecoin deposits. The business imperative here is clear -- cheap deposits are great -- but being so consumer hostile feels to me like a losing position.)
* See FRED's memorably-titled QBPBSTLKDPDOFFDPNIDP time series.
** MIR.M.U2.B.L21.A.R.A.2250.EUR.N and MIR.M.U2.B.L21.A.R.A.2240.EUR.N from the ECB.
Jon Rotbard retweeted
world app has been one of the craziest network bootstrapping plays ever...
- have a crazy retina scanning orb that people are down to experience
- pay people to open the app every month through expiring token claims
- use big user numbers to attract founders
- founders use big user numbers to attract capital
- more users, better products
we'll see what happens when the incentives dry up, but it's a heck of a starting point
We are looking for ambitious founders to build for the 35 million World App users.
I've published a piece with @variantfund outlining the opportunities we see
blog.variant.fund/the-world-…
Jon Rotbard retweeted
Major congrats to @alexanderlong & @PluralisHQ team, this has been over a year in the making and is literally a Google moment for AI.
This run is a live demonstration of a 7.5B model parallel training process running on commodity hardware.
The future of AI training is in giant permissionless swarms of heterogeneous, unreliable hardware.
Jon Rotbard retweeted
this take is
(a) categorically a mischaracterization of some of these companies including @ritualnet + assumes the quality bar of something like Templar is currently the same as @primeintellect/@NousResearch
(b) a good example of how people in crypto w/ bag bias over index on markets which creates poor incentives for real talent and good products in crypto x AI
so let's break this down
(1) @ritualnet is *not* a Bittensor competitor and doesn't plan to be. Bittensor is a network built for the incentivization of some compute/hardware task (i.e. GPU compute or inference).
Ritual is build to be a layer on which people build actual applications for the end user. On Ritual, people can easily link any sort of compute (AI models, web2 APIs) directly into the chain and access them from smart contracts.
Ritual would act as a layer where GPUs aggregated by subnets like Chute can be monetized for flow or models created by a subnet like Templar can be integrated and monetized via applications.
Bittensor, nor any of its subnets, is NOT built to be a layer for application development. Put another way, Ritual would be the layer where AI x crypto gets actual users, vs. incentive extraction that's happening on many subnets right now.
Of course, if @JosephJacks_ had read our docs, he would know this.
(2) In looking at the top subnets on Bittensor (Chutes, Gradients, Celium, Templar, Targon, etc.) the unified thematic is either compute/GPU offerings or distributed training/no-code training.
The commonality is almost all of this is incentivized, i.e. Chutes (#1 subnet) offers significant discounts on models like Deepseek R1/V3 to the official provider (almost a 67% haircut!), which is offset by $TAO emissions. It remains to be seen if this is sustainable (and fwiw, the Chutes team does seem to be trying this as an intermediary).
Tokens are great network bootstrapping tools. but $TAO emissions are being sold to offset multiple subnets. There were subnets that were explicitly farmed by external protocols to top up their treasuries (many that were venture-backed) for months that were just dumping $TAO.
Some of these teams are long-term oriented with $TAO, but not raising funding in $ or a numeraire thats not $TAO means that subnet teams need to sell $TAO to cover operational costs, creating some very perverse incentives
(3) A very important part of expanding crypto x AI is bringing in prime talent and attention *external* to crypto. For people in crypto, the narrative of not having VCs is great, but for people in AI they want to work with intelligent people and get compensate for it.
Getting and retaining this talent is a function of incentives + brand, which venture funding is overwhelmingly helpful.
@PrimeIntellect / @NousResearch / @ritualnet (and others) are no doubt able to attract strong AI talent in part because of this, which helps legitimize the crypto x AI space to people outside of the space far more that "having no VCs will".
This also shows: @ritualnet / @NousResearch / @PrimeIntellect all have a number of people who have consistently been publishing frontier research at ICML/ICLR/NeurIPS, and actively working on bringing relevant research (new cryptographic schemes for AI, watermarking, distributed training) to the attention of folks outside CT.
In comparison, there's been VERY little of this from the Bittensor ecosystem, as the focus seems to be primarily pushing the narrative vs. any meaningful expansion of crypto x AI into the mainstream (which is what gets net new attention, users and capital).
overall this bag bias maximalism is an uninformed take, paints all the VC backed projects as "bad", and willfully ignore the actual innovation many of these other projects are doing. Fair launch is not a compensation for quality whatsoever and we see this. As long as the optimization function seems to be around market dynamics, you're going to get the same shit quality crypto x AI projects that have been capturing attention for a while.
there are of course other projects on this list (i.e. Sahara) that have massive funding, a token, and no product to show that deserve the criticism. there are also quite a number of people working on Bittensor related things that I think are quite smart, and this is not a dig at Bittensor. I have exposure to protocols/assets in that ecosystem as do I Ritual.
Bittensor clone … there’s a few of these.
Here’s what I think:
There are no direct competitors to Bittensor itself, but there are many VC-funded competitors to many Bittensor subnets (specifically ones like subnet 4, Manifold’s Targon network)... The common pattern we observe across all of these below is pre-mine token sales (unfair distribution of economics) to VCs and no live mainnet launches on any, after years of effort and > $300M in funding.
This is very similar to what mainstream “crypto VC” has experienced over the last 5 years of deploying > $50 billion in capital into Web3 startups, producing negative IRR on that capital — which stands in sharp contrast to a fair launch network with no VC funding (Bitcoin) delivering > 10X liquid returns in the same timeframe (mid 2020 to mid 2025).
