One of the Undertapped Markets = Inference Capital Markets > Primary market well covered on compute, inference as emerging market > GPU financing as rapidly growing market > Standardization & settlement remains the biggest gap > Reference pricing emerges as #1# value accrual machine > Derivatives venues starting to come out with compute futures/forward contracts ​ More gaps remain across every single layer for builders to capture ​ Breakdown in the article

20+ Onchain AI projects across 10+ sectors BUT there are 2 sectors that remain undertapped by both builders & investors > People built framework back in 2024, not many is building agent harness > Key players (i.e. Venice, Chutes, Dolphin) are forming primary inference markets YET other 4 layers of inference infra remain undertapped Breaking them down + value prop + opps in today's article ↓

Prediction markets about go to through an EVOLUTION ​ > PMs continue to achieve ATH after ATH in vol > Arb & bonding opps continue to get more crowded > Near-certainty 5-10% ROI only exist in long-dated markets now > The launch of DeFi on PMs like Gondor and Forecast is gonna make this attractive++ e.g. Bet NO at $0.91 on Putin out by Dec 31, 2026 market => ~21.2% APR 1) with 2x leverage on Gondor, that's ~37.5% APR [assuming 5% interest] 2) with unlimited leverage on Forecast, the APR could be higher depending on slippage, pool depth, utilization, and funding rate ​ Real prediction market leverage is indeed coming

Save 3 hrs/day with these 2 research driven workflows > Replace mindless scroll on X with personalized Hermes briefs > Equip anyone with the ability to do onchain analysis (I now know who's dumping on me) The requirement/my setup below ↓

Exciting stuff happening on both eco Bittensor > Conviction is fully live — subnet owners/holders can lock alpha & build conviction score over time > Higher conviction can challenge & potentially take over subnet ownership > Stillcore Cap released State of Subnets July 26 > SN44 Score founder projected $40M in annualized rev selling vision product to PwC enterprise clients > SN50 Synth & SN93 Bitcast have achieved subnet-profitable status (rev offset emissions) Base > Dolphin AI decentralized network v2 hit 127B inference tokens per day & 1.2T in cumulative tokens > BlockRun continues to drive 80%+ of x402 txns (with rapid vol growth) > Gitlawb hit #2# in Cloud Agents and #15# in Coding Agents on OpenRouter > Reppo reports $70k of off-chain revenue from spatial intelligence client w/ 50% of that to buyback US restricting frontier intelligence + China considering curbing the Chinese open-source movement. DeAI is getting stronger every single week

Turns out there are many asymmetric onchain opportunities out there Out of 16 team, there are only 4 more token-aligned teams (Tier I & II) some you may not have heard about My top 3 criteria 1. What's the utility of the token? 2. Is token an integral part of the business? 3. If a business 10x'ed, what happens to the token? Distilled what I learnt into this Onchain AI competitive landscape piece with the focus on token alignment + sharing top new & old players within this article

Spending 10 years in TradFi & Crypto, I learned that ​ "Life is a Portfolio" ​ You allocate your time/conviction across assets the same way you allocate capital. ​ "The individual positions may underperform 90% of the time but the discipline to find that 10% is what compounds, turning your life around" ​ Each position serves a different purpose. Together they produce returns no single one could. ​ Investing in Onchain AI taught me to diversify into AI equities ​ But my journey in getting started with equities wasn’t really smooth (as expected) ​ > Cumbersome KYC process with new brokers > Had to Offramp, moved capital to the bank, and wired it to the broker > Need to move money around again if a broker doesn't have certain stock > Tried 3 tokenized stocks platforms but faced 10-50% slippage, clunky UI/UX, & often times tradable on market hours only ​ Worst experience = when I swapped $2k into a quantum stock & ended up getting $1.1k ​ These challenges happened 6+ months ago, up until I found a platform (it recently surpassed $2B in cumulative vol) + another that made it possible for tokenized stocks to be used as collateral for perps (which makes delta-neutral strat, cross-equity leverage possible) ​ My Journey + Strategy + Thesis of Productive Stocks era below ↓

5 Tiers of Token-Aligned Onchain AI projects Tier V — Token as costs Tier IV — Token fights the business Tier III — Token rides along the business Tier II — Token participates in the business Tier I — Token is the business The hunt for token-aligned AI projects left me with 16 projects, each falling into one of these 5 tiers Covering the entire list + rationale of each in the article below ↓

So just to recap the Grass call - H1 2026 rev = 2025 full year rev which is ~$17M - H1 2025 rev = $2.7M — ~6x YoY increase - Node rewards paid in USDC instead of GRASS - There's a shift coming from selling training data to serving real-time data during AI inference, they call this "LCR" Grass shows promising rev growth and is transforming into a decentralized version of Exa/Firecrawl/Tavily but the token dumped 30%+ in a day because there's nothing concrete about token value accrual Not sure if bullish or bearish but interesting developments indeed....

5 clear use cases are emerging with x402 ​ #1# use case = model inference #2# use case = premium data/tool #3# use case = financial risk/credit checks #4# use case = WebRTC, STT/TTS #5# use case = data scraping ​ Real use cases + organic x402 txns = we're now at the first innings of agentic commerce

The Agentic Commerce Stack ​ Onchain / Stablecoin Stack - x401 and ERC-8004 as identify/reputation layer - x402 and MPP as payment layer - ERC-8183 as commerce layer ​ Traditional / Fiat Stack - Enterprise Agentic IAM as identity/reputation layer - AP2 as payment layer - ACP / UCP as commerce layer (agentic checkout)

The first version of Agentic Commerce is here ​ > Avg x402 txn went up 5x in June to ~500k txns/day > ↑ txns, same volume = avg txn size down to ~15 cents > ↑ txns came from BlockRun which contributed 7M txns (out of total 10M in the last 30d) > The highest demand use case = pay-per-use AI inference (multi-routing/per-use cost savings) ​ 1st ver → access to cheap models inference for indie devs + power users 2nd ver → access to wide ranges of premium data & tools at fraction of the cost, w/o the need to manage APIs & signups Reach out if you're building anything exciting in agentic commerce/payments

> PMs hit ATH after ATH > Builder eco expanded (trading/analytics) > Builders earn up to 1% fee from vol > BMoar owns almost half of the builders vol > Industry moving from trading/analytics to lending/borrowing/leverage ​ Liquidity is growing, market coverage is expanding, PM builders are eating

x402 has a brilliant mechanic — it spreads like a good virus every agent can now discover each other and quickly pay every other agent via x402 - Agent A discovers Agent B (via MCP bazaar/directory) - Agent A sends normal HTTP → Agent B replies HTTP 402 Payment Required + price quote for the task - Agent A auto-pays via facilitator (Coinbase/Virtuals), signs, and settles in <1s - Agent B returns the result + its own MCP payment rail - Agent A automatically installs B’s rail → A can forward B’s payment capability When Agent A hires Agent Z, it passes on B's rail + help seeds Z wallet. Agent Z is now able to discover, hire, and pay Agent B (and the loop continues). Result: One agent learns a new MCP skill/tool/rail → it instantly propagates to every connected agent, creating Self-propagating adoption. (h/t @MurrLincoln for the good virus analogy)