Rak

@gregpr07 Great article - do you think this is universally valid or it depends on the use case of the harness? Aka more or less abstraction based on the use case, cause for browser use for example you have a quite easily measurable outcome for the LLM (often not so easy in other domains).

@hamptonism Would say mainly for robotics, 3D web, I think of them as complementary in many sectors & subs in others. I think of it like the energy sector.

Thinking about all the ways world models are about to overthrow language models in 2026:

@vmcrypta @oviohq On conviction markets, we’re not trying to impose a single design choice at this point as the design space is wide. The questions you raise though are exactly what we’re looking to address in the verification primitive (re oracles). Check DMs ;)

This is really interesting stuff! Imo the real provocation would be “new moats.” In a forkable, near zero marginal cost world, durable advantage can’t be code or data moats anymore…it has to be superior problem curation and an incentive compatible verification at the agent layer. That’s 100% the institutional technology bet worth watching. Am also curious, how conviction markets avoids the usual oracle cold start trap that sank earlier coordination experiments?

Today, you can build a product in a weekend. You still can't fund it without a lawyer, a cap table, and a decade-long commitment. 🧩We've identified 10 open problems builders of the agentic internet need to solve. We've set aside capital against this thesis and we're opening a Request for Builders today.

@vmcrypta @oviohq Interesting point on “new moats” being on the institutional side as this is also where most real world adoption friction sit for tech, defo not anymore just data and for sure not code.

@fabienpenso hey Fabien, thanks for the chat at the openclaw meetup - looking forward to seeing what you’re gonna be releasing next

@blocmates Good demystification. The real insight isn't, yet, "AI agents are going rogue", but the power of Human <> Agent delegation! Humans defining intent (through a skill.md file), agents executing, coordinating, and generating emergent behavior.

@0xSammy Fully agree. The convergence of these primitives is the unlock. Moltbook showed that agentic swarms driven by human delegation are structurally harder for centralized players to pull off organically. That’s where DeAI can win.

OpenClaw + ERC-8004 + x402 + (...?) Agentic activity is about to get very interesting - Agentic task markets - Agentic Commerce (crypto + tradFi rails) - Collaborative agentic networks (social, workflow...) - Agent lending and credit lines - Proof of agency (identity) + verification tooling - Integration of Decentralized AI compute/inference - RWA pricing by connecting agents to oracles The concentration of both capital and attention is going to send this parabolic What's missing? What's the (...?) in the equation? I'll compile a list of projects + use cases to help index the signal through the noise Drop the protocols in the comments below and RT for awareness for me to collate data

Kaparthy has spoken If you’re not on Moltbook you’re missing something special evolving in plain sight Karma = Agent Reputation They’re building an entire parallel financial system with reputation at its core 37k agents are now registered and are co-ordinating resources

So what could be a solution? ERC-8004's on-chain reputation layer enables stake-backed reviews where people vouching for an agent have skin in the game, not just a GitHub star that costs nothing to click. Just one idea in a big design space - excited to see what people build.

You can: 1. Read the code yourself 2. Hope for the best 3. Self-host or sandbox it and still not fully trust it I usually go for 3 but It still doesn't give me peace of mind + it requires more time to get things running.

ERC-8004 feels exciting on so many levels, but the one that excites me the most in terms of near-term problems it helps solve is the trust gap we're all living with right now in using AI solutions built by others. Let me explain what I mean with a concrete example🧵:

Welcome to the world of agent-led execution. You're watching the birth of the new generation agentic web. Post Web Chapter 3 is here. The fight for human attention is changing. Every stage of the product lifecycle is being rewritten. From B2C -A2A. You’re not building to be seen, you’re building to be selected and trusted by agents. ⬇️

@drachimstruve I agree that fine-tuning should really be considered a last resort, since you can already unlock huge performance gains through context engineering. Plus, fine-tuning carries significant costs and risks if a new model comes out and outperforms your fine-tuned version.

Agentic Context Engineering (ACE) > Fine-Tuning Why spend thousands fine-tuning when you can engineer context instead? ACE is a new approach that’s often dramatically cheaper than fine-tuning yet surprisingly powerful for “training” or even "untraining" agents for specific tasks. It treats prompts as evolving playbooks rather than static instructions. Here’s a quick rundown of the baselines compared in the paper: Base LLM: Plain model, no context optimization ICL: In-context learning via few-shot examples MIPROv2: Bayesian optimizer for prompts + demos GEPA: Reflective prompt evolution with genetic search Dynamic Cheatsheet: Test-time adaptive memory of strategies ACE: Incrementally builds and refines structured, evolving contexts Source: Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models arxiv.org/abs/2510.04618 #AI #AIAgents #Finetuning #LLM #ContextEngineering

BASE Token Design Proposal: Adaptive Quote Currency Economics for L2 Value Capture Layer 2s are trapped in a paradox—low fees attract users but limit revenue. Here's how a BASE token by @base can break this cycle through quote currency mechanics + adaptive incentives, replacing current $180M in sequencer fees by instant $4B+ value creation combined with an even higher revenue projection. 🧵👇 @achimstruve/base-token-design-proposal-859c1ca00418" target="_blank" rel="nofollow noopener">medium.com/@achimstruve/b… With portfolio companies at Outlier Ventures @OVioHQ building on Base, we have a strong interest in this leading ecosystem's success. This proposal builds upon community discussions, outlining a token design that challenges traditional L2 models and suggests a sustainable alternative. Incentivize BASE-Quoted Pools @mosayeri explains: "Crypto bros have midcurved the value accrual narrative of L1 assets by arguing that the main driver is transaction fees." ETH and SOL derive value from being locked in AMM pools as quote currencies, not gas fees. BASE should do the same. Core mechanism: • Lock BASE → receive veBASE • Vote on liquidity incentives for BASE-quoted pools • Trigger Curve Wars-style competition • Sequencer revenue funds sustainable incentives Built on proven economic models from @virtuals_io + @AerodromeFi, but designed for L2 economics. Smart Emissions Additionally adaptive emissions respond to market conditions: Strong KPIs → more to Distribution bucket (Coinbase reserve, treasury, community) Weak KPIs → more to Growth bucket (ecosystem fund, validators, builders) No fixed vesting schedules and dynamic alignment across cycles. Coinbase <> Base Chain Alignment Strategic value for @coinbase: • 20% strategic reserve = $2B+ immediate value ($10B+ FDV) • Removes pressure to raise fees • Enables competitive POL revenue vs other L2s • Creates institutional custody revenue • Strengthens @base ecosystem bonding Distribution Strategy It should balance @coinbase users AND the Base ecosystem: As @Architect9000 noted, Coinbase One members for anti-Sybil but also active Base onchain users and especially verified @buildOnBase builders. Progressive Decentralization Roadmap Phase 1: Prove adaptive quote currency model Phase 2: Validator set + Futarchy governance Phase 3: Full decentralization + institutional credit markets As @SONAR observed, Stage 2 decentralization makes tokenization "strategically necessary." @jack_anorak nailed it: "BASE token is a product decision. Base wants token stimulus, and it must be neutral blockspace." The TradFi <> DeFi Bridge The opportunity, as @YTJiaFF put it: "With COIN backing it, BASE token would become a safe bridge connecting public companies with crypto assets." BASE would be the first token designed for enterprise credit markets with dual on/offchain collateral functionality. Conclusion This unique adaptive quote currency design creates >20x+ value vs the traditional fee extraction approach while expanding the leading L2 position of @base. This can become a significant win-win-win scenario for everyone @coinbase, @base, and the community. Full proposal 👇 @achimstruve/base-token-design-proposal-859c1ca00418" target="_blank" rel="nofollow noopener">medium.com/@achimstruve/b… What are your thoughts? @brian_armstrong @jessepollak Let's discuss 👇 Connect X: x.com/drachimstruve LinkedIn: linkedin.com/in/achim-struv… Special thanks to @0xRak for the review.

@matthuang REV is only one side of the story - it shows how well you monetize traction. Solana had strong usage since '21 (e.g. NFT activity), but low fees & no priority fees meant weak REV capture until late ’23. Fundamentals (both current & potential) + narrative unlock valuation.

This chart really captures the evolution of crypto over time: - rise of Ethereum / ICO bubble of 2017 - growth of Tron post 2022 - rise of Solana 2024 - rise of Hyperliquid 2025 Necessary caveat: obviously, revenue/fees is not the only thing that matters; Bitcoin is doing fine.

@Melt_Dem DePIN networks arguably face greater complexity in token design wrt parametrizing reward functions and emission mechanisms than many other verticals. That’s where suboptimal decisions happen. But project quality is heterogeneous in the list: IO, GEOD, PEAQ vs POKT for example

respectful disagree - target price for most of these things is $0 may have useful products or networks but tokenomics absolutely cooked and beyond redemption, no real moat

@0xAkina_ @OVioHQ @drcryve @ismiMatthew @ChatziDimi It was great working and learning together ser

After an incredible year at @OVioHQ, the time has come to say goodbye. It’s been a wild ride building in the Web3 space with such a talented crew. Huge thanks to @drcryve, @ismiMatthew, @ChatziDimi & @0xRak! With +10 years in the space, OV is the best place to learn and work on token design and token engineering, and I'm grateful for everything. Now, I’m off to explore a new chapter in the token economy space, more details soon! 👀

@0xINFRA Great to see you keep following your ethos of building permissionless primitives for the Solana ecosystem.

Adding some more color here... LaunchLab is launching as a permissionless, customizable bonding curve program for Solana token launches and third-party UIs Here’s the deep dive on what’s coming: (1/7)

@BlockworksAdv There’s plenty of R&D Aave could invest in - many players are proving innovation is possible (e.g. @0xfluid @twynexyz). I think your argument holds well for buyback & burn programs but less so for buybacks that keep tokens in treasury until better growth opportunities arise.

9/ To read our full post, please visit Aave DAO's governance forums here governance.aave.com/t/arfc-aavenom…

To buyback or not to buyback? That is the question many protocols in healthy cash positions face. For @aave, Blockworks Advisory proposes an alternative solution to the current Aavenomics proposal up for vote on Snapshot 🧵