Vijay Patha

Something worth sharing with Product folks: https://lnkd.in/grQhi-qY Brian Balfour article on Reforge covers multiple layers of what the product work with AI will really mean for us with great examples that summarizes the complexity of the software development new setup. Attaching my favorite here: Customers expectations on products relying on AI.

20

5 Comments

MPP just launched. Stripe , Tempo, and Visa all in. Production grade agentic payments across every major rail. The session primitive is actually clever. Authorize a budget upfront, stream payments without per-transaction friction. That’s the right call for a payment rail. But session authorization answers one question: can this agent spend? Intent governance answers a different one: should this agent be buying this, right now, with these parameters? Permission and meaning are not the same question. They never were. MPP solves the first one well. The second one doesn’t have a name yet. It will.

10

4 Comments

New Replit + Claude Workflow = Speed + Quality I tried something different yesterday. After hitting some friction with Replit’s Agent, I decided to pivot. I started over with Claude (not Claude Code) and built my client's Healthcare app prototype screen by screen. Once the UI was dialed in, I had Replit stitch it all together in a new project. The result? The process worked perfectly. Claude follows instructions with precision in the right context. The workflow felt faster and more cost-effective—within a few hours of guiding Claude to refine the design to my vision, we had five interactive screens ready to be transformed into an interactive prototype within Replit. But this isn’t "AI magic." Building a SaaS startup's prototype in a day only happened because of the groundwork laid over the previous week: 1. Stakeholder Alignment: In-person session with the clinical team, CEO, and VC advisor to define the Target Operating Model. 2. Compliant Controls: Consulting on why Human-in-the-Loop (HITL) checkpoints are required to mitigate risk in Agentic AI applications. 3. Strategic Augmentation: Helping the team distinguish where AI can automate vs. where human oversight is essential for Responsible AI implementation. 4. The "Secret Secure Sauce": Performing a preliminary RCSA (Risk & Control Self-Assessment) and establishing EBCE (Evidence-Based Control Evaluation) criteria before creating the product design. I know my start-up friends roll their eyes at me when I talk about risk controls and change management; they think all those years "at the bank" clouded my innovative spirit. I think I'm more innovative due to that experience. Imagine the innovative skills you hone working with all that governance while still finding ways to move fast and deliver enterprise-grade quality. The truth is, you can have your cake and eat it too… it just takes a little more time in the bakery. And it's essential for regulated industries: high-fidelity interactive designs require more than just a prompt; they require building the governance right into the workflows, so risk is reduced before people pick up the process. In design and development, AI carries the heavy load, but the human must direct the process through the application of one's own agency and a lens of structured control and performance standards. I love Replit, but when the Agent doesn't seem to be connecting with me, I'm going to pivot to this modular approach: using Claude to produce specific .tsx files and then using a detailed "master prompt" to stitch them together in Replit. Back at it again today for the clinician side of this app, using this same approach for high speed and high fidelity. #BuildInPublic #AIWorkflow #ResponsibleAI #RiskManagement #ProductStrategy #Replit #ClaudeAI

6

PayPal just pushed a major update. PayPal World isn’t just aggregating wallets. It’s abstracting national payment rails (UPI, WeChat Pay, etc.) into a unified interface dissolving borders without rewriting behaviors. Now imagine this stitched layer with PYUSD embedded programmable, composable, settlement-native. Not just faster payments, but intelligent money flows across legacy and crypto-native networks.

3

1 Comment

In the early days of a startup, what people engage with isn't the product. It’s how the founders see the future. Early in YC, Eli said something that stayed with me. When he hired great engineers, he never shared just one doc. One was requirements. The other was expectations. At the time, it felt like a hiring detail. Only later did I realize it was something deeper. Most early-stage companies communicate what they're building. Very few communicate how they think. As we’re writing a company handbook at Hyprnote, we decided to publish it openly on our website. Not because it’s "best practice", but because in the early days, authentic founder writing does more than any polished landing page. Before traction, before growth charts, companies were built on manifestos. On personal beliefs about how the world should work. That's what people connected to. Open source makes our code public by default. This is us doing the same for our thinking. The handbook isn't a rulebook. It's a snapshot of how we define good work, ownership, and trust - right now. Curious how others see this: what's a belief or expectation at your company that exists, but has never been written down publicly?

35

1 Comment

Speed without structure is fragile. In the early stages, most teams optimize for what feels immediately causal: build the product, acquire users, push growth. That bias shapes every decision. Resources flow to what moves the visible metrics fastest. The problem isn’t neglect…. it’s sequencing. Some systems don’t create speed upfront. They prevent it from breaking later. If i would do it all again, i would accelerate hiring HR and legal professionals to the first 10 employees. Short-term wins quietly accumulate long-term constraints. Talent investments outside of core product delivery convert speed from a burst into something sustainable. They let you keep acting decisively without compounding invisible risk. Real speed isn’t early acceleration. It’s how long you can keep moving without breaking.

17

1 Comment

Another counterintuitive post - No, AI isn't causing the layoffs you'd expect. Where AI’s impact is showing up is more subtle, and arguably more important. Startups that received seed funding in 2024–2025 are raising capital with smaller teams than they did just a few years ago. The average team at seed funding has four fewer people today than in 2019. The implication is clear: AI is allowing founders to do more with less. Another question lingers: when companies cite AI for layoffs, is it strategy or spin? It’s probably a bit of both. DM me- Always happy to chat about the report and dive into any questions! Check out the latest report at the link here: https://lnkd.in/gpuqrHwN This is put together from SVB's proprietary data, PitchBook, and SVB Analysis. #tech #svb #venturecapital #vc #ai

16