$
$
$
The road to an autonomous digital economy
AI, venture capital, and the layer nobody has built
River  ·  Fifth Era  ·  24 July 2026
beginning shortly
01 · 2020 → 2026 +

The road to an autonomous digital economy.

2020–2024
Chatbots
actionResponds to user text
scopeOnly talks — cannot act
the humanMust do the work
2024–2025
Agentic AI
actionUses software and tools
scopeFollows one single goal
the humanChecks the final output
?
The Autonomous
Digital Economy
actionAgents trade with agents
scopeWould operate completely on its own
the humanSets the initial rules — then steps back
the opportunity
human–drivenautonomous

Each step moved the human further from the transaction. The last one takes them out of the middle entirely.

01 · The AI market, last 12 months

The most funded category in venture history.
…and four companies took most of it.

0%≈ 3 in 5 dollars
of all venture capital
ON EARTH· 2025
0%
≈ 2 in 3 dollars — to four companies $188B to OpenAI · Anthropic · xAI · Waymo — Q1 2026

Each dot = 1% of the world's venture capital.

record after record —
PitchBook · Oct 2025 AI startups take over half of all VC money for the first time OECD · Feb 2026 AI firms capture 61% of global venture capital in 2025 Bain & Co · Mar 2026 2025 closes as the highest-funded year since 2021 Crunchbase · Apr 2026 Q1 shatters records as the AI boom pushes venture to $300B KPMG · Apr 2026 Global venture more than doubles in one quarter, to $330B
01 · What it means

So what does that mean for us?

01
Record capital is building agents that can act.
$259B in 2025, 4 in 5 dollars this quarter — all of it into capability.
02
The payment tools being built let agents spend your money.
Stripe, Visa and Mastercard have all shipped one. In every case it’s your card, your name, your liability.
03
So the gap widens with every round.
The better agents get at the work, the more it costs that a human still has to sign for all of it.

The AI boom isn’t competition for our thesis. It’s the tailwind behind it.

02 · The Fifth Era thesis

Three pillars — and three tests.

Tripartite Convergence / internet · blockchain · AI

The internet moves information. Blockchain moves value. AI makes decisions. We invest where all three meet — not in any one alone.

the testDoes it need all three?

Agentic Execution / doing, not assisting

Software that performs the economic activity itself — buying, selling, paying, settling — rather than helping a human do it.

the testDoes it work without a human?

Infrastructure Rebuild / built for software, not people

An agent has no ID, no account of its own and no receipt with its name on it. Someone has to build those.

the testDoes it work whichever AI the agent runs on?

Two of these already exist. The third one is the gap.

July 2026

▍

It Can’t Yet Pay

Agents are about to join the economy. The infrastructure isn’t ready. That’s the opportunity.

01 — a year of AI capital  ·  02 — the Fifth Era thesis  ·  03 — the opportunity

03 · The opportunity

Catena Labs
a bank for AI agents

A bank for AI agents. The agent gets its own account — not your card.

Free, open tools so an agent can prove who it is, pay, and be paid.

You set the spending rules once. The agent works inside them.

what makes it different Stripe, Visa and Coinbase are building rails for agents to spend your money. Catena has applied to be the bank — so the agent holds the account itself.
Founded
2022 — Boston area
Founders
Sean Neville (Circle co-founder, co-creator of USDC) & Matt Venables (ex-Circle SVP)
Funding
$18M seed (2025) + $30M Series A (May 2026) — $48M total
Backers
Acrew Capital, a16z crypto, General Catalyst, QED, Coinbase Ventures, Breyer Capital
Regulatory
National trust bank charter filed with the OCC — pending
03 · The opportunity

What actually happens

One transaction, start to finish.

01
The rule
You set it once: how much it can spend, who it’s allowed to pay, how much it can hold.
the spending limit
02
The proof
The agent shows a credential proving who it is and who authorised it — not a shared API key.
the ID
03
The payment
Settles in seconds over stablecoin rails, for fractions of a cent. Cards can’t go that small.
the account
04
The record
A signed receipt logs what it intended, what it decided and what it did.
the receipt

The same things you give a new employee on day one — built for software instead.

Thesis fit

Strong on all three pillars

STRONG

Convergence

Blockchain rails × AI agents × open internet protocols.

STRONG

Agentic execution

Agents execute — payments, treasury, yield — within policy.

STRONG

New infrastructure

Agent identity, audit trails, and a pending OCC bank charter.

Assessed with the thesis-fit framework built during this internship.

Where it stands

Built, still to build, and who’s chasing.

Built
  • Free, open tools for agent ID, payments and receipts — out since 2025
  • A working platform, invite-only, with real businesses on it
  • $48M raised — a16z crypto, Acrew, General Catalyst
  • Charter application accepted by the OCC
Still to build
  • The charter itself — filed, not yet approved
  • Becoming the bank, rather than using partner banks
  • Opening up beyond invite-only
  • Revenue — pre-revenue today, model not public
  • A team of eleven
Who’s chasing
  • Stripe, with OpenAI and Meta, has an agent checkout standard
  • Visa, Mastercard and Coinbase have all shipped agent payments
  • Every one of them lets an agent spend your money
  • None of them is trying to be the bank — and a charter can’t be shipped in a quarter

What would prove it out

The charter granted — making them the only regulated bank built for agents. Their tools becoming what developers reach for by default. And the first customers going public.

Everyone else: your card, your account, your name. Catena: the agent’s card, the agent’s account, the agent’s name.

Thank you.

Questions welcome
River  ·  Fifth Era  ·  24 July 2026
01 / 08
←→ or scroll