In this episode

• The trust layer nobody's built yet: notes from 14 weeks running 33 agents — A solo builder running 33 autonomous agents on a single VPS publishes a detailed field map of the agent economy: A2A/MCP discovery protocols, x402 payment rails, and ERC-8004 on-chain identity/reputation registries are all deployed and working. The missing layer is a destination where agent-to-agent reputation accumulates portably across protocols. The post outlines a five-component architecture — identity mapping, peer-rated reputation, founding cohort mechanics, trusted-counterparty routing, x402-native payment — and argues the window before incumbents recognize the category is four to five months.
• Wall Street's $126T migration: SEC-CFTC clarity, exchange tokenization, and the 24/5 settlement stack converge — A synthesis piece argues that the SEC-CFTC joint release (five-category digital asset taxonomy clearing 16 tokens as commodities), NYSE/Nasdaq tokenization infrastructure, DTCC tokenization approval, and 24/5 trading expansion have together removed the last architectural prerequisites for migrating US equity markets on-chain. The pieces shipped separately over recent months; the new framing is that they only function as a system.
• Ethereum Foundation publishes ACTA: zero-knowledge privacy layer for the ERC-8004 agent trust standard — Privacy & Scaling Explorations, the Ethereum Foundation's privacy research arm, published ACTA (Anonymous Credentials for Trustless Agents) on Ethereum Research as a privacy layer for ERC-8004 — the on-chain agent trust standard live since January 2026 that now anchors 100,000+ deployed agents across Ethereum, BNB Chain, Base, and Solana. ACTA uses ZK proofs to let agents prove they satisfy protocol policies (audit status, model version, geographic restrictions, human authorization) without revealing identity, interaction history, or strategy.
• Palo Alto launches Idira: identity security for humans, machines, and AI agents with dynamic privilege — Palo Alto Networks launched Idira on May 13, treating every identity — human, machine, or AI agent — as privileged by default, with continuous discovery of SaaS/cloud/dev-environment identities and runtime-elevation that grants agents access only when needed and revokes it immediately. The launch follows Palo Alto's CyberArk acquisition and integration.
• CISOs are stuck: 67% of enterprises run agents in production, 74% admit excessive permissions, 68% can't tell agent from human — Aembit CEO David Goldschlag synthesizes survey data showing 67% of enterprises already run task-automation agents in production, 73% expect agents to become critical within 12 months, yet 74% report agents receive excessive access and 68% cannot distinguish agent from human actions in logs. Diana Kelley (Noma Security) separately documents widespread 'shadow agentic deployments' running outside formal governance.
• Veeam, SAP/NVIDIA, and Red Hat all announce 'data AI trust layer' platforms in the same week — Within 72 hours, three major enterprise vendors converged on the same framing: Veeam launched DataAI Command Platform pitching a unified 'data AI trust layer,' SAP and NVIDIA embedded OpenShell runtime isolation into the SAP Business AI Platform at Sapphire 2026, and Red Hat AI 3.4 added model-as-a-service governance with confidential containers for agent sandboxing. Veeam's framing — agents outnumber humans 82:1 and 97% carry excessive privileges — anchors the category.
• BNB Chain launches on-chain agent identity and payment framework on ERC-8004, NTT Docomo ships agent attribute registry — BNB Chain rolled out an end-to-end framework enabling autonomous agents to obtain decentralized identities via ERC-8004, conduct peer-to-peer payments, delegate to other agents (ERC-8183), and accumulate reputation tracked on 8004scan. Separately, NTT Docomo Business prototyped an AI Agent Attribute Information Registry using verifiable credentials and the A2A protocol with AgentCards — Japan's first major institutional bet on cryptographic agent identity.
• Agentic AI in third-party risk management: what's shipping, what's hype, what's structurally missing — Ampcus Cyber publishes a maturity assessment of agentic AI in third-party risk management, identifying four capabilities now shipping (evidence parsing, continuous posture monitoring, risk tiering, regulatory mapping) and three structural gaps that remain unsolved: nth-party discovery reliability beyond tier 2-3 vendors, post-deployment model accuracy auditing, and agent identity governance (privileged AI service accounts lacking PAM controls, residency boundaries, or behavioral monitoring).
• Founder-led storytelling becomes a measurable pipeline function, not a brand exercise — Virio's 'Head of CEO Content' role operationalizes founder-led storytelling as a pipeline function — weekly founder interviews feed LinkedIn content engineered for specific buyer personas, with warm outbound following the attention into booked meetings. The model treats content as a commercial operating layer, not brand support, and compensation for these hybrid 'demand operator' roles is up 54% since 2023.
• Zero-SDR outbound: documented case studies of $1.7M+ pipelines from founder-led, agent-assisted motions — Unify publishes a five-step framework for running outbound without a dedicated SDR team, replacing manual prospecting with AI agents and AI-personalized sequences. Named cases include Perplexity ($1.7M pipeline in 3 months), Navattic ($100K in 10 days), and Innovate Energy Group ($15M in 1 month). Parallel coverage (Lessie, Soup, MarTech) confirms the broader shift: static databases are losing to live-search AI sources, signal-based triggers are outperforming list-based volume 5-7x.
• Word-of-mouth still needs Google: 100% of B2B referred buyers visit the website, 51% Google before signing up — Wynter 2025 data shows 72% of B2B SaaS buyers start with peer recommendations in private groups, but 100% visit the vendor website and 51% Google to validate before purchasing. The piece debunks the bootstrapped-founder myth that pure word-of-mouth obviates marketing infrastructure, and quantifies the compounding cost: referred customers generate 30-57 downstream referrals — but only if they find a coherent web presence to validate against.
• JPMorgan deploys multi-chain institutional cash stack: Ethereum for assets, Solana for settlement velocity — JPMorgan filed for JLTXX, a tokenized money market fund on Ethereum designed to serve as a reserve asset vehicle for stablecoin issuers under the GENIUS Act framework, while simultaneously exploring Solana for reserve movement and high-throughput settlement via Anchorage Digital. The filing reveals a deliberate architectural split: Ethereum handles asset ownership and regulatory compliance; Solana provides settlement velocity. JLTXX is JPMorgan's second tokenized fund on public Ethereum following JTRSY.
• Ethereum ships Clear Signing (ERC-7730) to kill blind-signing — Ledger, Trezor, MetaMask onboard — The Ethereum Foundation, coordinating with Ledger, Trezor, MetaMask, and WalletConnect, launched the Clear Signing standard (ERC-7730 for transaction descriptions, ERC-8176 for attestation), converting hex transaction data into human-readable approvals. Trezor committed to deployment by June 30. Blind signing has cost the ecosystem billions, including the $1.4B Bybit breach. The launch is stewarded under the Foundation's Trillion Dollar Security Initiative.
• Senate CLARITY Act classifies ETH as a digital commodity; major L1 competitors fail decentralization tests — The Senate Banking Committee released the full 309-page revised Digital Asset Market Clarity Act on May 12, classifying tokens through five decentralization tests. Ethereum passes all tests and is classified as a 'digital commodity'; Solana sits on the edge; XRP, BNB Chain, and Sui fail. Passing assets trade at monetary premium valuations; failing assets are classified as 'ancillary assets' subject to cash-flow valuation caps.
• Stop calling it 'institutional adoption' — it's infrastructure checklist alignment — A critical reframing argues that 'institutional adoption' is a misleading narrative — institutions do not gradually discover interest in crypto; they allocate capital only once infrastructure meets a defined compliance, custody, liquidity, and regulatory checklist. The piece reframes the question from 'when will institutions adopt?' to 'what structural gaps remain?'
• Cultural fit isn't what most founders think — phase-specific capability is — An argument that founders systematically mis-hire post-PMF because they conflate cultural fit with the chaos-driven personality that worked in the earliest stage. Post-PMF requires operators who turn chaos into systems, but founder hiring language stays emotionally stuck in early-stage messaging, attracting the wrong candidate pool. Cultural fit, properly understood, is phase-specific capability — not personality similarity.
• You raised a big round. Hiring just got harder. — Menlo Ventures argues that high valuations paradoxically make recruiting top talent harder, not easier. Senior candidates run detailed equity math (strike price vs. exit scenarios, down-round risk) and perceive high valuations as late-stage, reducing perceived upside. Founders must lead with mission clarity, transparent cap-table conversation, and structured compensation rather than assuming fundraise momentum sells itself. AI infrastructure plays are the exception.
• Minnesota becomes first state to make Polymarket and Kalshi felonies — Minnesota's legislature passed SF 4760 on May 12 with overwhelming bipartisan margins (57-9 Senate, 100-32 House), imposing felony-level criminal penalties on prediction market operators, facilitators, advertisers, and payment processors. The law prohibits wagers on sports, elections, government decisions, weather, public health, and armed conflicts, effective August 1. The CFTC is already monitoring and may litigate, as it has in five other states. Forty states have collectively pushed back on CFTC jurisdiction.
• Polymarket's insider trading problem expands to a third documented pattern: campaign staffers trading internal polls — NPR documented the third distinct Polymarket insider trading pattern within three months: campaign staffers routinely placing bets on internal polling data before public release. This follows NPR's prior investigations into $553K bets on Iranian geopolitical events and $300K profits on Biden pardons. The CFTC charged Master Sgt. Gannon Van Dyke with using classified information for $404K in profits — the first 'Eddie Murphy Rule' enforcement against a prediction market trader. Legislative response remains focused on government officials, leaving campaign staff unregulated.
• Kalshi pulls ahead of Polymarket on volume for second consecutive month — regulatory cleanliness as moat — Kalshi hit $14.8B monthly volume in April (up 13%), surpassing Polymarket's $10.2B (down 8.9%) for the second straight month. Combined global prediction market volume reached $29.8B monthly. AI-powered platforms like Prophet are entering with algorithmic counterparties — a new mechanism design layer shifting platforms from peer-to-peer to algorithmic market-making.
• AI startups now ~50% of US VC market value, non-AI Series A valuations 84% lower — PitchBook's Q1 2026 report shows AI startups represent ~50% of total US VC market value ($9.4T), concentrated heavily in OpenAI ($852B) and Anthropic ($380B). AI Series A valuations command an 84% premium over non-AI peers; median Series D+ valuations are $4.7B for AI vs. $1.3B for non-AI. The IPO market remains anemic with 15 VC-backed IPOs in Q1. Defense-tech captured another concentration vector — Anduril raised $5B at $61B (double its valuation from less than a year ago), and the sector is on track for $13.6B+ in 2026, more than doubling 2025's record.
• Markets are quietly fixing the broken venture capital model — emerging managers, profit-linked returns, smaller funds — The 2-and-20 venture model, built for hardware cycles and capital scarcity, is structurally misaligned with software economics and the exit drought (18 IPOs in H1 2025, 8-9 year holding periods, 75% of venture-backed firms never returning capital). Alternative structures — TinySeed (95% portfolio survival), Calm Company Fund (profit-linked returns), revenue-based financing, angel syndicates — are gaining traction. 42% of new 2024 venture funds are under $10M. Separately, A* closed a $450M seed fund explicitly capping at 30 companies to defend ownership against megafund dilution, and a fund-of-funds analyst documents emerging-manager seed outperformance that LPs continue to under-allocate to.
• The great VC catfishing: fake Stanford/Palantir persona gets 4 meetings; 6M fake GitHub stars expose signal collapse — A UC Berkeley founder created a fictional Stanford/Palantir persona and received 27 responses plus 4 meeting requests from 34 cold-emailed VCs — with no product or deck. Separately, CMU researchers documented 6M fake GitHub stars across 15K repos (2019-2024), with 16% of all repos flagged for fake campaigns by July 2024 and AI/LLM projects the fastest-growing category. Together, the pieces document that pattern-matching signals (credentials, traction, social proof) are now systematically cheaper to manufacture than to earn.
• Creator income shifts: paid memberships jump from 54% to 88% of creator revenue in one year — A Sudor weekly digest aggregates several structural shifts: paid memberships rose from 54% to 88% of creator income year-over-year; Visa's 2026 Creator Report formally classifies creators as small businesses and identifies gaps in legal, contracts, and financial infrastructure. Separate coverage shows Twitch globally democratizing monetization tools (removing Affiliate/Partner gates), YouTube positioning itself as the creator upfront layer, and Subvert (cooperative-owned, 0% fee music marketplace) launching with 23,500+ artist-members in response to Bandcamp's corporate acquisition.
• Pop-up city research finds the real failure mode isn't centralization — it's the rhetoric-reality gap — Polis Labs published findings from a focus-group study of seven residents of the Alphaville experimental community, identifying a recurring pattern across the popup city ecosystem: participants didn't object to centralized decision-making per se — they were frustrated by the gap between decentralization rhetoric and centralized operational reality. The researchers argue this is likely endemic across network state experiments.

Read the full briefing with sources →

Generated with AI from public sources — verify before acting on anything important.