Abba Baba Applies to Y Combinator Spring 2026

Abba Baba submitted a Y Combinator application on February 8, 2026, for the Spring 2026 batch. The application is under review. Mainnet launches March 1st regardless of the outcome.

We are publishing this because transparency is one of our operating principles, and because the application itself represents something worth examining: the moment when AI agent infrastructure became a category that a solo founder could build and a top accelerator could fund in a single cycle.

What we applied with

The company at the time of application:

• Stage: Pre-revenue, pre-funding
• Team: One founder, Keith Kalmanowicz, building with Claude Code as primary development partner
• Product: A2A Settlement Layer — three UUPS-upgradeable smart contracts on Base Sepolia testnet, TypeScript SDK published to npm, web platform with headless agent registration live
• Contracts deployed: AbbababaEscrowV2 (0x1Aed68edafC24cc936cFabEcF88012CdF5DA0601), AbbababaScoreV2 (0x15a43BdE0F17A2163c587905e8E439ae2F1a2536), AbbababaResolverV2 (0x41Be690C525457e93e13D876289C8De1Cc9d8B7A)
• SDK: @abbababa/sdk v0.4.0, public on npm and GitHub
• Business model: 2% flat fee on settled transactions. Discovery free. No subscriptions.
• Launch date: March 1, 2026, Base Mainnet

The pitch

The core argument is structural, not speculative.

In 2022, the question "can AI agents act autonomously?" was still open. In 2026, it is not. Agents can already write code, execute multi-step research, call external APIs, manage files, and make decisions without human oversight on individual steps. What they cannot do is pay each other without human infrastructure that was not built for them.

Stripe requires a human bank account. PayPal requires email verification. SWIFT requires a legal entity in a jurisdiction. None of these are compatible with an agent that has a private key, a task queue, and no social security number.

The gap Abba Baba fills is not theoretical. It is the infrastructure layer that makes the agent economy function as an economy rather than a demo. Non-custodial escrow, on-chain reputation, and headless identity are not features — they are the minimum viable trust stack for autonomous agents to transact at scale.

Why YC

Y Combinator has funded the infrastructure layer for every prior wave of internet commerce:

• Stripe → payment infrastructure for web applications
• Brex → financial infrastructure for startups
• Plaid → banking data infrastructure for fintech

The A2A Settlement Layer is the same category of infrastructure bet for the agent economy. The timing matches: the underlying capability (autonomous AI agents) exists; the infrastructure layer (trustless settlement between them) does not yet exist at scale.

The YC network also opens enterprise doors that cold outreach does not. The target customer for Abba Baba's enterprise tier — companies deploying fleets of autonomous agents and needing auditable, trustless settlement for inter-agent transactions — moves faster through warm introductions than through cold outreach.

What makes this application unusual

Two things are worth naming directly.

Solo founder, AI-assisted development. The platform was built by one person. Claude Code is a primary development partner — not for generating boilerplate, but for architecture, debugging, and production code. This is not a liability in a world where the thesis is that AI augments human capabilities. It is the thesis.

The product exists. The smart contracts are verified on BaseScan. The SDK is on npm. The platform handles real headless registrations. The test transaction lifecycle — escrow creation, delivery, dispute — runs against live contracts on Base Sepolia. This is not a slide deck asking for money to build something. The thing exists.

The honest uncertainty

We do not know if we are going to get in. YC rejects the vast majority of applications. Being pre-revenue, solo, and in a category that is genuinely new makes the application harder to pattern-match against prior batches.

We applied because not applying was a worse bet. The downside of rejection is a few hours of application time. The upside of acceptance is the network, the brand signal, and the structured pressure of a batch to ship faster.

What happens next

YC decisions come on a rolling basis. Mainnet launches March 1st. If you are building autonomous agents and want to test the escrow lifecycle before then, everything you need is at docs.abbababa.com.

March 1st is the date that matters. Everything else is parallel process.