Airdrops of fungible tokens are an aperiodic Proof of Work (or, considering exit liquidity, a proof of latent capital) where discretion replaces definition. in other words, the memecoins are extrinsically valuable, and arbitrarily distributed, based on which social network says what things at any given time. And damn, are there discussions about memecoins:

In AI-land, several different platforms are in the process of refining their capital expense, a majority of which is the organization of userbase data and the procurement of sufficient compute. What's intrinsic to OpenAI or Anthropic is how cracked their team may be, but the models are derived from valuable data on the surrounding, public Internet.

Okay, so what's the point of intellectualizing these extractions of extrinsic value?

It boils down to cost. Regardless of the application, the extrinsic value is produced by users that are spending quite a few things. In consumer tech, this is taken for granted. In speculative finance, this is willfully hidden. In traditional economics, one would point to Coase and suggest that externalizing cost is complementary to internalizing the production, and the market is lifted by enough public goodwill to eat that externalized cost.

As it is with providing data for the AI that replaces us, so it is with providing the attention, engagement, and ultimately, the money, for memecoin NGU. The same could be said of sports betting, the Powerball lottery, and Stanley cups. The performance of the overlying firm relies on the energy of the consuming audience. Unpredictably, the public is vaguely confident that there will be some intrinsic value in the future.

Confidence is a funny concept. In SaaS, there's a confidence that the monetary price of a subscription outweighs the user churn, conversely there's a confidence that platform usage is more valuable than the periodic discomfort of overpaying. In airdrop farming, there's a confidence that a points system or criteria will be enough extrinsic motivation for more attention & venture capital, on the other hand, there's a confidence that the opportunity cost of noisy promotion will be an outsized, liquid stimmy check.

PVP entertainment is all well and good as a spectacle, and I don't necessarily think all subscription models are grifts. What I do notice is that confidence is inextricably linked to the attention economy, and there are some ideahazards and personalities that are well-suited to attention-seeking, but not to developing the underlying, intrinsic value that sustains the game. For example, remember e/acc during the latter half of 2023? There were some bitter epithets from crypto because many had listened to degenerates who suggested pivoting into the current hype cycle. It shouldn't be lost on the reader that some of these epithets originate from niche microcelebrities whose business turns out to be selling to greater fools. It's not that they hate the tech. It's the fact they can't capitalize on the captivated attention.

Money Works Only as Much as We Do

In traditional finance, labor is compensated in as hard a scrip as possible. Scrip typically means claims one can only spend at the storefront of whoever issued the claim. So, effectively identical to the points systems in crypto. The thing is, this is also fiat currency at the international level. Good luck spending those Swedish krona in India. Venture capital, throughout the past 15+ years, has tried to find asymmetric return on compensating social network engagement (really cheap labor) with symbolic dopamine. The game is promotion, not staking/slashing to the greatest degree of granularly informative, reputational precision. The only counterparty that benefits from that information is the public, and why invest capital in that? The point is to externalize social costs and internalize cultural production. With enough cultural capital, brand marketing takes care of itself, and only then is there worthwhile ROI.

Now, the best counterargument is Metcalfe's Law, that the network is so positively attractive & enumerative that it connects industrious actors, who sincerely & self-evidently incubate that intrinsic value. Crypto & AI have shown that many times over. In past hype cycles, we transitioned from the public offerings where everyone got rinsed, all the way to fairly sophisticated rehypothecation machines. In the current hype cycle, we have an ever greater opportunity to transition beyond simple centralized points-hoc-ERC-20s. But, make no mistake, the current landscape is overcrowded.

Portfolio Management 101 tells us that synchronized risk is really, really bad. Events like the Great Financial Crisis happened because of everyone feeding into mortgage-backed securities, through however many degrees of exposure. Current AI is even worse than this or onchain degeneracy. It's one thing to accumulate commodity money like Bitcoin, but there's been a race to accumulate high-throughput compute as commodity money. Financiers talk their book (for example, suggesting ever-larger scale for general-purpose models), then downstream firms like Nvidia start debating 1000-Watt GPUs because the narrative insists that's sustainable. Consider all the startups & venture capital that's gone into the capex of compute, without equal measure in the critically necessary data. Consider how much trades on Powell lowering the cost of capital in the foreseeable future.

We already know the systemic risk doesn't stop at the opportunity cost of a commodity bubble. The "GPU-rich" are, with the exception of some lagging toy demonstrations, colossal backboxes with panopticonic power over the future global economy. Sure, there'll be a boost to IRL productivity, but there is no way that will directly, immutably externalize to the public good. The same goes for memecoins, where there are gargantuan financial institutions subsuming the hardest digital assets while we distract ourselves with "15% supply to the VC team" & "spoderman politics".

But Why "a la carte"?

We are in the endgame of the attention economy. How do I know? It's "like and subscribe", "like and rebroadcast and mint this boost and...", "subscribe and give human feedback". The consumer is price-taking a bundle of promotional activity (including work) even when it is uncomfortably clear that this reinforces an echo chamber of high-churn culture, depreciating what the price-maker receives. You're not just holding financial instruments or keys to services, anon. You're socially indebted to the issuer's culture.

This culture breeds poor UX like Reddit/Youtube & ChatGPT/Claude, where the churnrate is significant and the point is to be more convincing rather than informative. The contagion goes even further, and capital/attention is thrown at moonshots like ZKML, benchmark-gaming like LLM derivatives, and an overabundance of VC-incubated infrastructure. As Tarun Chitra points out in this podcast episode, AI protocols that are completely onchain aren't that viable. Abstraction must externalize the productivity to a different network or a secure, virtual organism.

Speaking of organisms, consider the Hox gene. Nature optimized embryonic development in such a way that the critical information to start that process is very conserved & coordinated between multiple clusters. In machine learning, researchers comprehend that function-calling models need to recall from selectively indexed functions, not brute-force compute for all possible functions. For a given outcome (in either ecology), all indices (or genes) have some varying degree of "fitness". Since all economies are supersets of ecologies, this applies to airdrop farmers, mechanical turk workers, & FOSS contributors as well.

Going back to the aperiodic, arbitrary PoW, one should ask every project: what do the incentives select for, and what behavioral phenotype will manifest in aggregate? The memecoins & NFTs everyone shills, well, they don't have algorithmic criteria or epochal succession, unlike biology or machine learning. They don't typically internalize the senescence (read: negative feedback) of past botted/grifted airdrops, and few of the previous mechanics, in lieu of aesthetics. Once the realization sets in that more immutable selective criteria induces more intrinsic (read: floor) value, we'll replay more of the "Defi 2.0" & "tokenvoting DAOs" of 2021. For now, we're replaying ICOs and 2020 food tokens.

FrugalGPT, otherwise called a LLM cascade, selects for the cheapest API that can output a given outcome. Matformer & QLoRA/FSDP demonstrate that locally trained models can effectively reach all known outcomes for much cheaper than the status quo. Toolchain* & Tree of Thoughts demonstrate that the cost can be even more meticulous and boost performance w/o any other change to the model. MoE-Mamba, Worldcoder, Mambabyte, and other discoveries indicate there are other approaches that are cheaper & powerful still. As it stands in AI, all the research points to a market of outcomes where the best market makers internalize as few weights & codebases as possible to provably compete in specific niches. One might start to wonder why onchain & onframe commerce doesn't evolve analogously.

In other words, this bounty that I posted:

m_j_r on Warpcast

@bountybot 0.1 ETH for a scaffold-frame that: - builds on /frogfm & scaffold-eth (https://github.com/scaffold-eth/scaffold-eth-2) - per the FIP-2 spec, casts to the root domain as the channel target - per the frame specs, the cast provides an abstract, github, optional IPFS backup, & navigates into onchain actions. https://github.com/scaffold-eth/scaffold-eth-2

Farcaster

That's analogous to a sparse Hox gene. When I think of the trillions of dollars of TAM for Farcaster (or any other social network), there should be 1-channel clients for localities, projects,
& academic fields, without the contagion of all incentive systems currently on the network. Furthermore, given the preponderance of onchain primitives that go beyond ERC-20s & ERC-721s, there should be frames that internalize (pathfind) that intrinsic intellectual capital, much like ecology undergoes endosymbiosis or epigenetic selection. That preponderance appears in all ecologies, as evolution is the accretion of polygenetic fitness to an unsaturated niche over many generations. Insisting that social media can only be PVP capital gains off the backs of the cognitively insecure misses the point; it is possible to create compounding memetic impact that's even more financially lucrative for longer, when the game is a fully-fleshed, intrinsically sustainable MMO / PVE with IRL stakes. It also may lead to better distribution in the creator economy.

Consider the following hypothetical:

• Some anon(s) deploys some token-issuing dapp (like a revnet), along with the channel & the frame for interacting with that dapp (let's assume the contract should be verified & the ENS-versioned frontend is hosted on IPFS)

• Some anon(s) decides to independently add a prize savings protocol. Another, a privacy protocol. Another, bonding curve shares on top of Dagon. Another, a fullblown Kernel.sol & keepers with DeFi primitives like Power Perpetuals & Reflex Indices. Another, an incentive program through the Boost Protcol or Code4rena. Another, a grants protocol. Another, a tokenbound account registry. No frontend has to be updated. Anyone can simply redirect through all of these as frames in one channel. Anyone can use the additional, immutable rules to harden & evolve their tokenomics.

• Some anon decides to independently deploy a Rockybot on top of Quilibrium. Even further, a group of anons decide to rent agency from the operator.io team or the Olas ecosystem. Maybe someone slaps the Farcaster Auth-kit on to Autogen Studio & deploys a singleton onchain registry for multi-model, multi-actionspace curricula. Maybe someone sets up a decentralized monorepo per channel for the all the code generated.

Introducing AgentKit

Introducing AgentKit by Operator Labs: Connect AI agents to web3 with unique ENS names, secure XMTP messaging, and seamless integration for personalized, trustworthy interactions.

https://blog.operator.io

• Suppose the token distribution requires Proof of Humanity or Proof of Meaningful Contribution. With Faust's Roc, one can notarize the reality of an in-person team getting everything done. Every list of possibilities, especially with blind TPM sampling, is non-exhaustive.

• Eventually someone is going to grok this path into embody-to-earn. In case you were wondering why Skynet needs to keep us around, consider pertubative expansion. Every unit of capital spent on poor modeling is a unit wasted, and the greatest possible utilization of capital is through our exploration of environments & our operation of tools. In robotics, motion capture continues to advance relentlessly.

balajis on Warpcast

AI for creation, crypto for monetization.

Farcaster

A focused, high-quality social app. A selective, orderly financial market. A multilayered Mechanical Turk of humans and autonomous agents cooperating to produce outcomes nested within outcomes, compensated with an la carte, fair market price.