Dexter Krishnan

VAMP = Winter is here! It's been really quiet so no major intel to share. Processors are starting to issue VAMP fines on a select basis. What I see happening is going to be problematic IMO. I've had a couple of MID's get hit with modest VAMP fines. Less than $1K for clients in the $100M/yr range. The VAMP fines caused the merchant's effective processing rate to spike and this led them to reach out to me. The client VAMP ratio is well below the 2.20% and 1,500 count thresholds. So they are technically compliant and shouldn't be fined per TC40. Processor A charged $4 and Processor B charge $8 per TC40 which showed up on their monthly statement. This means that BIN A is above the 0.50% VAMP threshold and BIN B is excessive with an above 0.70% VAMP ratio. The processors are not VAMP compliant but the merchant is. So who eats the VAMP fines? If the merchant is not excessive, 100% of fines should be paid by the BIN owner. The bank or processor made the choice to accept high risk merchants, was not proactive in terminating the bad actors but now they want the merchants to pay for those poor decisions. Why is this problematic? First, if every decent merchant gets hit with VAMP fines, it'll be a customer service nightmare for everyone. Second, merchants won't put up with that. They'll cancel their MID and simply shift their volume elsewhere. Lastly, the processor will lose decent merchant volume and get even more concentrated with the remaining bad merchants. It'll be interesting when things tighten further up in January...

28

6 Comments

Airbnb isn't a better booking site. It's a trust utility. Most competitors have a hard time making strangers trust each other. Airbnb realized they only needed both sides to trust Airbnb. Then enforced that trust through the interface. We have designed for a few FAANG companies. The lesson we learned: monopoly-scale products design for control, not delight. Startups chase delight to steal users. Dominant platforms chase lock-in: engagement over ergonomics, transaction completion over delight. Airbnb's design is protocol. Every element constrains the market to rules Airbnb sets. The interface is beautiful - AirBNB’s CEO Brian Chesky famously obsesses over craft and pixels. Beauty serves consistency, not novelty. Every listing, every interaction follows the same grammar. The design creates inevitability. Map and list views with preserved state reduce search entropy. Calendars show constraints first. Listing-card grammar forces comparability. Total-price display shifts pricing power to the platform. When Airbnb made total price prominent in 2023, then global default in April 2025, 300,000 listings removed or lowered fees. By year-end 2023, 40% charged no cleaning fee. This is how design becomes control when you own the market: 1. Expose hidden costs, force new equilibrium Total-price display was market regulation by UI. Make fee-loading visible in search and hosts reprice or vanish in the rankings. Airbnb reported 300,000 fee reductions. 40% went to zero. 2. Ship features that only work at monopoly density "I'm Flexible" processed 500 million searches within months of launch in 2021. It needs enormous inventory to function. Below critical mass it collapses. That creates a scale-gated moat competitors can't copy. 3. Accumulate trust signals that compound Fifteen years of reviews, responses, and host behavior produce a path-dependent trust graph. New entrants start at zero. Airbnb compounds. 4. Bundle vertically to block horizontal attacks Categories, Split Stays, AirCover (2022) aren't features. They're governance: standardized discovery, itinerary assembly, and guarantees in one flow. Can't compete with just "vacation rentals" when the monopolist handles everything. In April 2025, total-price display became standard worldwide. Interface fiat restructuring an entire category. Airbnb is worth $75 billion today not because pixels are prettier, but because the interface adapts to the market and focuses on customers of both sides. A decision architecture that sets rules and leaves rivals negotiating terms inside its protocol. Playbook: use design to win the market, then use the market to enforce your design. You can control the market when beautiful consistency becomes the protocol everyone else has to follow.

11

2 Comments

I’m an AI skeptic that used AI to code an iOS app. That I’m actually releasing on TestFlight. Some thoughts: 1. You still need to have a good idea. 2. You still need to be clear about the goals and features. 3. You still need to understand what a good user experience is. 4. You still need to be able to articulate what you want clearly. 5. You need to be able to break up the features into clear, communicable prompts. 6. You still need to be able to test it. 7. You still need to be able to debug it. 8. You still need to consider how people will actually use your app, and in what contexts. Because here’s the thing: it’s really good, until something breaks. Then it’s like saying you wrote your speech in Ancient Greek. That’s cool, until you’re asked to actually give the speech and you realize you don’t understand Ancient Greek. 

9

10 Comments

Tech opportunities aren’t disappearing, they’re just reshaping. The future isn’t about finding a job; it’s about owning what you build. AI and blockchain are tearing down old barriers, giving millions of builders the chance to launch, own, and thrive.

5

Red flags when evaluating offshore dev agencies 🚩 🚩 "We have developers in every tech stack" Translation: We'll say yes to anything and figure it out later 🚩 "Rates starting at $15/hour" Translation: You're getting mass-produced code from mass-produced coders 🚩 "We can start Monday" Translation: These devs have been on the bench for months. Ask why. 🚩 "100% satisfaction guaranteed" Translation: We have no legal definition of satisfaction 🚩 "We've worked with Fortune 500 companies" Translation: We did QA for a subsidiary's internal tool once 🚩 "Our developers work YOUR timezone" Translation: We'll burn them out in 3 months and swap them without telling you 🚩 "Full-stack developer who does it all" Translation: Master of nothing, debugger of everything 🚩 "We handle everything so you don't have to manage" Translation: You'll have zero visibility until something breaks 🚩 "Agile certified team" Translation: We do standups. Sometimes. 🚩 "We're not like other agencies" Translation: We are exactly like other agencies — The only question that matters: "Can I talk to 3 clients you've worked with for 12+ months?" Odesa.co What red flags did I miss? Drop them below 👇 #Hiring #OffshoreDevelopment #Startups #Outsourcing #RedFlags #techhiring

8

1 Comment

MCPs are incredible for exploration. They're terrible for production. Everyone's been excited about plugging Claude Code into tools via MCP, and that's valid. The tooling lets you quickly pull live data, test integrations, and prototype systems that would've taken days to even conceptualize. BUT if you try to build a truly scalable system on top of MCPs, it falls apart quick. So far, I've seen the following: ➡️ Data comes back inconsistently formatted: the same request can return different formats depending on context, which breaks everything downstream. ➡️ Sessions are fragile: they don't survive restarts and don't scale. One dropped connection can silently kill a workflow across dozens of accounts. ➡️ No usage controls: there's no built-in way to cap how much an agent is consuming. You're flying blind until something breaks or a bill surprises you. + more I saw all of this while building Kale OS (still in progress but SUPER exciting). We connected MCPs for EmailBison, Instantly.ai, Cloudflare, etc. At first it was great because we could prototype workflows in minutes + test ideas without reading a single line of documentation. That's the real value of MCPs. They give you a landscape understanding of what a tool can do and what's possible. Once we knew what we were building, we had to shift to APIs directly to built it directly. Direct integrations give you control over everything so that you can build your systems your way. This is how you build something anti-fragile aka something that doesn't break at 2am when you're running campaigns across 30 accounts. The framework we use now: 1️⃣ MCP first: to explore, test, and validate that an integration is worth building. Understand the tool's surface area fast. Prototype the workflow. 2️⃣ API second: to build the real thing. Reliable, predictable, scalable. The version that runs in production without someone babysitting it. TLDR; MCPs are an amazing scouting tool, but not a foundation. The teams that figure this out early will build systems that hold up ⚙️ [ operators_diary_7 ] ⚙️

44

20 Comments