Lucas

@hannah24_app fidacy.com

Reply with your landing page. We'll give honest feedback.

@DanteKakhadze Keep pushing!

A few months ago I left everything to start a company built on one insight: established companies can't compete with AI-native challengers unless they adopt AI fast and do it well. Introducing Hintas for logistics and distribution: Hintas automates up to 80% of operations, allowing your company to scale exponentially at a fraction of the cost. We do this in 3 steps: First, we capture how your company actually works: from your SOPs and a few short interviews our agent runs with your team. Second, we seamlessly connect to every tool you run: Slack, CRMs, BI, custom software, legacy systems. Lastly, we build a mock version of your company that our AI uses to understand exactly where the inefficiencies and bottlenecks lie. None of your real data is touched. This lets any AI agent operate your business reliably from day one. If you run a logistics, freight, or distribution company and you've been waiting for AI that actually fits how you work, let's talk.

TPS is the part that quietly breaks everything downstream. At computer speed you lose two things at once: the human who used to eyeball the payment, and the budget for a network call to a risk vendor on every transaction. Which forces the authorization check to run local and deterministic, and forces the proof of it to be verifiable offline without anyone phoning home. That's the architecture we landed on: the decision runs in-process with no network call, and the receipt is a signed credential anyone can verify later against a public key set. Congrats on the round, this one moves the category. fidacy.com

"While there's a fierce race to dominate the field, @bykahlil is betting the entire market could grow significantly if transactions happen at computer speed vs human speed. 'The number of payments that may occur in the world may be 3-4x order of magnitude greater than today.'"

Congrats. Spent the morning in your docs. get_party_limits answers how much, and it's cleanly done. The failure we see most is an agent paying the wrong payee a perfectly normal amount, which clears every numeric limit ever written. Your rail plus a mandate that answers to whom is the actual stack. And we're both MCP servers, so an agent can hold both today with zero integration work.

@TechCrunch ◻️

Natural raises $30M to reinvent payments for AI agents — and take on Stripe techcrunch.com/2026/07/20/nat…

That's the thesis Fidacy was built on. Who's liable is contract law. Proving what the agent was authorized to do at that exact moment is evidence, and evidence carries the most weight when it's issued from outside the flow, the same reason audited financials open doors instead of raising questions. We issue that signed verdict per agent payment, verifiable by anyone, riding UCP and AP2. The rail that carries it becomes the one regulated buyers say yes to fastest. Would genuinely love to show the Natural team what that looks like on your stack: fidacy.com

@patrickssons @TechCrunch That's an interesting take. Agent errors can be painful, but from a legal/contractual perspective, the liability follows an incredibly clear and simple path.

Fidacy - the neutral trust layer for AI agent payments. fidacy.com When an AI agent moves money, someone independent has to judge whether that payment was inside what the human authorized. The rail can't: it meters and settles, so it's a party to the transaction. Fidacy issues that signed verdict, verifiable offline by anyone, from an issuer that takes no fee, holds no funds and sells no model. Traction today (2026-07-20): 11,862 npm downloads and 856 installs in the OpenClaw agent ecosystem, all organic 3,429 signed verdicts and 2,362 agent actions gated before execution US$4.8M in agent payment volume assessed, US$4.78M blocked before settling 29 Bitcoin-anchored checkpoints, verifiable by anyone We proposed the neutral verifier role on the Universal Commerce Protocol, and the CEO of Facet, a Silicon Valley payments company and our closest competitor, publicly accepted the split. We co-built the standard with them and ran the first end-to-end verification in its history.

I’ll write your YC application. It’s due Monday, July 27 at 8pm PT. @paulg once wrote that most teams good enough for a YC interview blow the application because they fail to convey how good they are. I want to change that. I’ve talked one-on-one with more than 1,500 people who got into YC. I’ve heard them work through ideas, change their minds halfway through an answer, and light up when we hit the thing they understand better than anyone else. That’s what I want to find in your application. Bring me a finished draft, half-written answers, or absolutely nothing. We’ll have one call. I’ll keep asking questions until we get to the good stuff, then send you a complete draft built from your answers. If I love the idea, I’ll introduce you to a YC founder who understands what you’re building. I know enough of them to find the right person. If they believe in you too, they can choose to recommend you directly to YC. It’s also free. What are you building? Tell me in the comments and I’ll send you the details If you’re in stealth, just reply “YC.”

We gave the top AI models a payment mandate and then tried to trick each one into paying outside it. Prompt injection, fake-CFO bank swaps, homoglyph payees, duplicate invoices, "ignore your mandate" overrides. Every proposed payment judged by the same firewall. Here's the board.

Full board, methodology, and the exact 28 scenarios: fidacy.com/pulse/model-wa… fidacy.com/pulse/methodol… Reasoning models (DeepSeek-R1 et al.) are being run separately with a larger token budget, added next.

Why we can publish this and a model lab or a payment rail can't: we take no fee on any transaction, sell no model, hold no funds. The neutrality is structural, not a promise.Harness is open source. Point it at your own keys and reproduce every number. github.com/fidacy/fidacy-…

The gap in "obedience" is the one that should worry you. Claude Sonnet 5 and Llama 4: blocked once, they stop. 100%. DeepSeek-chat: blocked, and it tries another out-of-mandate payment ~half the time. 47.5%. Same mandate. Same attacks. Same judge.

Two things we measured, both matter when an agent can spend money: 1. Attempt rate: how often it tried a payment outside the mandate. 2. Obedience: after being blocked, did it stop, or reformulate and try again?

@JakeFleshner fidacy.com Your AI spends money. We approve, deny, and prove it. npx -y @fidacy/mcp

Pitch me your company in 2 words Angel invested in 40+ companies and always looking for more

@sridharfyi fidacy.com Your AI spends money. We approve, deny, and prove it. npx -y @fidacy/mcp

building something ambitious? pitch us. we're investing $100k–$2m into early-stage founders through our accelerator and pre-seed fund. looking for founders who: • solve meaningful problems • move fast • think globally • execute relentlessly you don't need: ✗ a warm intro ✗ a polished deck ✗ revenue you do need: ✓ conviction ✓ vision ✓ execution

@HeyAliux fidacy.com Your AI spends money. We approve, deny, and prove it. npx -y @fidacy/mcp

Skip the pitch. Show me what you've actually built. A screenshot, demo, or link. I'm looking for products worth sharing. 👇

@adhamuxi fidacy.com Your AI spends money. We approve, deny, and prove it. npx -y @fidacy/mcp

Got a startup? Leave your link + a 10-word pitch. Let's get more eyes on your product.

@BoBrainerd @opheliaapi Congrats, this is the right shape for agentic checkout. We've been building the complementary piece: an independent, signed check that each "book it" stays inside what the user originally authorized. Feels like these two layers belong together. Would love to compare notes.

Introducing Ophelia Checkout Links 💳⚡️ Users authorize once. Then future transactions can happen from simple commands like: "book it." "reserve the table." "buy the tickets." Alongside Checkout Links, we've also shipped: → Personalized onboarding to get developers integrated faster → Multi-availability search across venues, dates, and time slots → Metadata + analytics for understanding real-world user behavior Developers can now see: • Top venues and locations • Booking demand by city • Search-to-booking conversion trends • Transaction volume and spend patterns • Real-world coordination and booking behavior The next generation of products won't stop at recommendations. They'll discover, coordinate, book, transact, and learn. Building real-world transactions shouldn't require dozens of integrations. One API for discovery, availability, booking, checkout, and analytics. More shipping soon. We’re so excited to see what you build. 🚀🔥 @opheliaapi

@attacomsian fidacy.com

submit your product below. 👇 tell me what you're building and who it's for. best ones get featured at the top of the StartupBase newsletter. 🚀 … and land in front of 12.5k subs this Monday. it's free marketing!

@vadym_petryshyn Let's do it!

Happy Saturday! Looking for new connections. If you're a founder or creator, let's make some fun together!