AgentPaymentsRailsOpenasCopilotCutsKillWrappers

◆ DEEP DIVES

1. 01

   Agent Commerce Rails: The $2.2B Pre-Consensus Window Is Open Now

   Category Formation in One Week

   Four independent actors converged on agent-payments infrastructure in the same week, which is usually the moment a category stops being a thesis and starts being something you can actually buy. Google Cloud and Solana shipped pay.sh, metering Gemini, BigQuery, and Vertex AI in stablecoins at $0.001 to $20 per call, with 75 providers live at launch and MCP-server compatibility. Anchorage Digital launched Agentic Banking with regulated settlement and Know-Your-Agent identity standards. Stripe shipped 280+ features including agent wallets and stablecoin rails. a16z crypto closed Fund 5 at $2.2B and named autonomous agent transactions as a deployment target, in writing.

   When four independent actors converge on one category in one week, the integration problems that killed prior narratives just got retired simultaneously. That does not happen often enough to ignore.

   The Comp Set Just Got Printed

   AI-finance software valuations now have anchors, or rather, the anchors have decided to publish themselves: Rogo at $2B on a $160M Series D for IB modeling and research, 9fin at $1.3B on a $170M Series C for debt markets intelligence, and Mercury at $650M ARR, profitable for 3 years. These are almost certainly sub-$30M ARR businesses priced at 50-80x forward. Separately, Anthropic co-founded an AI-native services firm with Blackstone, Goldman Sachs, and Hellman & Friedman whose explicit job is deploying Claude across hundreds of PE portfolio companies.

   Where the Alpha Actually Sits

   The headlines favor the mega-names. The alpha sits one floor down:

   1. Agent wallets, KYA identity, and agent orchestration at Seed/A. Underfunded relative to the demand curve pay.sh and Anchorage just created. Expect 3-5x multiple expansion once the first marquee Series A prints.
   2. MCP-native data and guardrail plays in finance verticals. NatureAlpha's pattern — proprietary dataset, MCP server, monthly Claude Skills updates — travels into credit, private markets, alt-data, and regulatory reporting.
   3. Agent-fraud detection and spend controls. Stripe just commoditized the payment rail. Value migrates one layer up, to which agent is transacting, what it is allowed to do, and whether the transaction is fraudulent.

   The Timing Constraint

   a16z's $2.2B with an explicit agent-payments thesis will reset Series A multiples in this wedge inside two to three quarters. Anything not written in the next two quarters gets written at 3-5x the price. The counter-thesis is worth stating in the same paragraph: fund size is not deployment velocity, a16z crypto has historically been patient, and if the marquee agent-payments Series A does not print by Q3 the window widens rather than closes. This is probably wrong on timing and approximately right on direction.

   ---

   Finally, the 76% concentration on OpenAI across finance firms, paired with 43% of regulators not yet tracking AI adoption, is a latent regulatory catalyst sitting on the table. A single FSOC or OCC guidance note mandating model diversification triggers a category-wide repricing. The audit layer is cheap today.

   Action items

   • Launch agent-payments sourcing sprint: identify 10-15 seed/pre-seed companies across agent wallets, KYA identity, agent orchestration, and merchant acceptance
   • Build AI-finance comp sheet using Rogo ($2B), 9fin ($1.3B), Mercury ($650M ARR) as anchors and reprice any active Series A/B deal in space
   • Stress-test portfolio companies whose core wedge is 'AI agent for financial workflow' against Anthropic's 10 finance agent templates + Blackstone/Goldman distribution
   • Formalize 'crypto as settlement layer for agent commerce' as active sector thesis with sizing, competitive map, and 3-5 investable wedges

   Sources:TLDR Fintech · TLDR Crypto · NFX · AI Weekly · StrictlyVC

2. 02

   The AI Margin Reckoning: Per-Token Dies, Per-Result Lives

   AI Gross Margins Are Running 50-60%

   Bessemer finally put the numbers in public, and they are worse than what most board decks will admit: AI companies are landing at fifty to sixty percent gross margins against the eighty to ninety percent that SaaS investors spent a decade treating as natural law. The fast-growth cohort runs closer to twenty-five percent gross margin and funds the gap with burn. OpenAI disclosed a thousand-fold response cost reduction over fourteen months, which sounds like the end of the problem until you notice reasoning models push per-query compute up by ten times and eat most of it back. The margin trajectory of any AI company now depends on whether its optimization stack runs faster than its reasoning workload mix grows.

   A company doing $100M ARR at 25% gross margin is not the same asset as one doing $100M at 60%, and the gap between those two companies inside the same portfolio is about to widen considerably.

   Microsoft Just Proved the Thesis

   The best-resourced AI distributor on earth, with the cheapest compute and a captive OpenAI relationship, still cannot make horizontal AI-in-every-app pencil out. The Xbox CEO killed Gaming Copilot. The Windows chief pulled Copilot from Photos, Widgets, and Notepad. An EVP conceded on the record that inference costs are pressing on margins. What Microsoft is defending instead is Office 365 Copilot at $30/seat with 33% QoQ paid user growth. The read is simple: workflow-embedded AI with measurable ROI survives, feature-spray AI does not.

   The sleeper detail is that Anthropic is now inside Microsoft's flagship AI revenue product, which quietly ends the OpenAI exclusivity narrative and raises Anthropic's strategic optionality by an amount the last round did not price.

   Anthropic Moves to Per-Result Pricing

   Anthropic is shifting from per-token to per-result billing, which is probably the most important unit-economics signal of the quarter. It structurally rewards vertical agent companies with measurable outcomes, and it structurally destroys wrapper businesses whose margins evaporate the moment the lab commoditizes the workflow underneath them. The counter-thesis, which I half-believe: token-based pricing is currently a structural moat for Anthropic and OpenAI, because customer gaming (Meta's engineers burning millions of tokens on internal leaderboards being the canonical example) actually increases lab revenue. Both things are true at once, which is why the transition matters.

   The Repricing Framework

   Business Type	Margin Trajectory	Multiple Direction
   Reasoning-heavy AI (code review, research)	Compressed — 10x compute per query	Down 20-40%
   Single-shot/deterministic AI (classification, extraction)	Improving — rides cost deflation	Holds or up
   Per-result vertical agents (legal, clinical, financial ops)	Margin expansion as results improve	Premium to category
   Horizontal wrappers at sub-$30 ARPU	Structurally impaired	Exit or pivot

   ---

   The practical number is this: a thirty percent inference cost reduction on a $70M run-rate drops $20M+ straight to EBITDA, which is a higher-ROI use of operating capital than most GTM line items currently on the same board's roadmap. The portcos that have already elevated inference ownership to a VP line with quarterly cost-per-query targets will separate from the ones still treating it as a platform problem.

   Action items

   • Issue portfolio-wide data call: cost per inference, gross margin trajectory, reasoning-mode exposure, and optimization roadmap from every AI portco above $5M ARR — due in 14 days
   • Add inference economics diligence module to every new AI deal: fully-loaded COGS per query, optimization stack audit, reasoning % of workload, margin bridge to 70%+
   • Reprice or mandate pivot conversations for any portfolio company selling horizontal AI at sub-$30 ARPU without distribution moat or commerce monetization
   • Source vertical AI deals in regulated/high-ARPU workflows (legal, healthcare, finance, supply chain) where $200-2,000/seat absorbs inference cost and per-result pricing is natural

   Sources:🔳 Turing Post · Aaron Holmes · AI Weekly · Engineer's Codex · TLDR

3. 03

   AI Agent Load Shock Opens Three Displacement Windows Simultaneously

   GitHub Is Breaking Under the Weight

   GitHub posted 85.51% uptime across the last ninety days, dropped to 86% in May, and managed a data integrity incident touching 2,092 pull requests, a six-hour Elasticsearch outage, and a critical security vulnerability inside a single week. The CTO's explanation is, usefully, also the investment thesis: AI agents are driving 3.5x load growth in two years, a curve that previously took fifteen years to accumulate. Internal planning went from "10x capacity" in October 2025 to "30x capacity" in February 2026. Four months.

   The part that moves capital: Vercel, Linear, Railway, Resend, and Sentry are absorbing the same demand curve without breaking. That is architectural alpha showing up as reliability, and it is the cleanest justification on offer for the AI-native infrastructure cohort's premium multiples.

   AI coding agents just compressed the developer infrastructure upgrade cycle from 10 years to 18 months. Every stateful developer service — code hosting, CI/CD, artifact registries, observability — faces the same shock.

   Demand Breaking Hyperscaler Containment

   The same load shock is showing up at data center scale, which is the less surprising half of the story. Hut 8 signed a $9.8B lease with an investment-grade counterparty for a Texas AI data center, the largest single-day stock catalyst any crypto miner has produced. Anthropic's 300MW SpaceX deal, which we covered previously, was the crack. Hut 8 is the flood. Microsoft is quietly shelving clean-energy targets to accelerate the buildout, which tells you what the buildout is worth to them relative to the pledges. The miner-to-AI pivot is financed.

   Three Parallel Displacement Windows

   Window	Incumbent at Risk	Challenger Opportunity	Entry Timing
   Code hosting/git forges	GitHub (MSFT)	AI-native git forges, Forgejo, Graphite	Series A prices in 12-18 months
   Observability	Datadog (OpenAI leaving is the tell)	Axiom, Grafana Cloud, ClickHouse-based	Re-enter conversations now
   Compute providers	Traditional DC REITs	Crypto miners with >100MW, grid interconnect	Public equity — repricing on announcement

   The Crypto Miner Arbitrage

   Screen listed miners for AI-pivot optionality: >100MW capacity, Texas/Midwest interconnects, and investment-grade counterparty relationships. Anchor the comp to Hut 8's $9.8B lease. This is probably wrong in the specifics, but the re-rate happens on announcement rather than execution, and laggards will catch bids within weeks. Separately, Microsoft walking back the clean-energy pledge is the permission slip for everyone else in the industry. Gas turbine supply chains, SMR developers, and behind-the-meter operators get eighteen months of pricing power. The counter-thesis is that rates or a capex reset arrives first. Possible. Not what the filings suggest.

   ---

   One related tell worth tracking: Uber's CFO admitted the company maxed its 2026 AI budget by April and is reallocating from hiring budgets to cover the overrun. Multiply that across every S&P 500 CFO drafting November budgets, and the pull-forward for AI tooling vendors runs through 2027. Hiring lines are funding it.

   Action items

   • Source the AI-native git forge category: Forgejo, Graphite, Sourcegraph evolution, stealth players — target 3-5 meetings in next 90 days before category-defining Series A prices
   • Screen listed crypto miners for AI-pivot optionality (>100MW, Texas/Midwest, investment-grade relationships) and build comp sheet against Hut 8 $9.8B anchor
   • Flag GitHub dependency as portfolio operating risk — require CI/CD redundancy plans from any portco with release processes bottlenecked by GitHub uptime
   • Re-underwrite developer infrastructure positions using 10-30x AI-driven load as the new TAM denominator, not pre-AI usage curves

   Sources:The Pragmatic Engineer · Bloomberg Technology · TLDR · The Information AM

4. 04

   Kinsale Capital: When Three Credible Shorts Converge on a $7B Compounder

   The Setup

   Three short-sellers with actual track records, Steve Eisman, Brad Safalow of PAA Research, and Edwin Dorsey of Bear Cave, have independently converged on Kinsale Capital, ticker KNSL, a seven-billion-dollar specialty insurer whose stock is up 1,560 percent since its 2016 IPO. Eisman platformed Safalow's thesis in July 2025. Dorsey has since published primary-source regulatory complaints that corroborate it. Three credible bears aligning on the same mid-cap with overlapping evidence is the point at which a narrative starts pricing itself in.

   The Core Thesis

   The attack is specific, which is what makes it interesting. Kinsale's industry-leading margins, the bears argue, are not proprietary underwriting but the margin of selling exclusion-heavy policies to unsophisticated small-business owners in the lightly-regulated E&S market and then denying claims when they arrive. The number doing the work is a sixty percent retention rate against a ninety percent P&C industry norm. That is a thirty-point gap on the metric that most reliably separates underwriting from churn, and there is no flattering way to read it.

   The RSG anecdote will travel: a Colorado armed-security firm paid $25,427 for a Kinsale policy that excluded professional liability, assault, battery, and firearms coverage — everything an armed security company would actually want insured.

   Bull vs. Bear Framework

   Dimension	Bull Case	Bear Case
   Source of margin	Proprietary underwriting tech, low expense ratio	Regulatory arbitrage + information asymmetry
   60% retention	Natural SMB churn	Customers leave when they understand what they bought
   Exclusion growth	Prudent underwriting discipline	Product quality degradation
   Appropriate multiple	Premium compounder (30x+)	Services treadmill (15x)

   The Playbook

   The likely path here is not a fraud reveal. Or rather, the more interesting version does not require one. It is multiple compression, which is less dramatic and more expensive. Compounder premiums require the story to stay clean, and three bears with a media cycle, paywalled follow-ups dropping on schedule, and a non-trivial probability of NAIC attention add up to enough friction to remove the benefit of the doubt. The counter-thesis deserves airtime: the specialty niche is genuinely good, the combined ratio has been enviable, and short-seller convergence has been wrong on quality names before. This is probably not one of those times. It could be.

   ---

   Venture Read-Through

   If forty percent of Kinsale's book churns annually out of dissatisfaction, that is addressable market for transparent-pricing SMB insurance platforms, and reason to refresh diligence on Next Insurance, Coalition, Embroker, and adjacent plays targeting E&S risks with better customer economics.

   Action items

   • Pull NAIC complaint-index data for KNSL against RLI, Markel, WRB, Palomar, and Skyward — if Kinsale runs meaningfully above peers, that's confirmable alpha ahead of sell-side
   • If holding KNSL (direct or via SMID-cap quality/financials factor exposure), stress-test position at 15x earnings; trim to half-size if thesis relies on 'best-in-class underwriting' narrative
   • For event-driven books: build defined-risk short via OTM puts 3-6 months out, sized for the Bear Cave paywalled follow-up and next earnings print as catalysts
   • Scout long opportunities among Kinsale's disgruntled customer base — transparent-pricing SMB insurance and E&S-adjacent fintechs (Next, Coalition, Embroker category)

   Sources:Edwin Dorsey from The Bear Cave