Philippe Botteri

We at Foundation Capital been closely tracking the shift from software-as-a-service to services-as-software for over a year. In my latest with Jaya Gupta, we distill a year of learnings into three high-level trends: ➡️ Features aren’t your moat anymore - differentiation comes from how deeply and effecitvely you integrate your product into the customer’s world. ➡️ The traditional pre-sales vs. post-sales divide is gone - customers expect a working solution during the sales process. Selling and implementation have fused into a single, continuous effort. ➡️ Pay for outcomes, not seats - pricing is moving from per-seat or per-API charges toward alignment with business outcomes (you pay when the AI actually delivers value). There's lots of complexity in pulling this off, but it's where the strongest AI startups are heading. Why do these shifts matter? Because beyond the $200B SaaS market lies a $4.6T pool of spending on labor and services that this wave of AI companies is poised to absorb. Our full piece here: https://lnkd.in/gyBKXcit

272

27 Comments

🚀 Excited to share how we cracked a major challenge at Coherence: automatically generating production-ready API integrations at scale. Our newest product unlocks insights into data for your customers using AI, but to do this, the AI needs access to your product data. When every customer has unique APIs with 50+ endpoints, complex auth schemes, and custom business logic, manually writing integrations becomes impossible. Our solution? We built an AI system that uses Claude Code CLI and Daytona's secure sandboxes to automatically generate Model Context Protocol (MCP) servers - turning weeks of integration work into minutes. The magic happens when you combine the right tools: Daytona provides isolated, ephemeral environments for secure code generation, while Anthropic Claude Code intelligently analyzes API specs and iteratively builds, tests, and validates the integration code. This isn't just about saving time - it's about making AI agents truly accessible to any application, regardless of API complexity. We're now helping companies add intelligent agents to their products in under an hour. Curious about the technical details? Check out our deep dive on how AI is building the infrastructure for AI: https://lnkd.in/ey7uBb2a

74

5 Comments

So proud to be sharing our first distributed inference job in our team's channel today for the first time 😍 ByteNite will be launching its flagship serverless computing product to the public next week, bringing distributed high-throughput computing into infrastructure-free pipelines. Can't be more excited to witness the power of the technology that we've created in these years help solve real use cases. 🙇🏼 -6 !! Check out the stable diffusion tutorial: https://lnkd.in/gBt4iuqP #distributedcomputing #aiinference #serverless

53

4 Comments

Speaking at STAC Summit New York on the Data Streaming panel, “Adapting messaging platforms for new data demands.” I’ll share lessons from our work at AlgoX2: ◦Higher-frequency, smaller market-data messages: scaling fan-out and avoiding backpressure. ◦Fast, efficient persistence of trading-engine data to storage without impacting production. ◦What’s next: tighter storage + analytics integration, hardware acceleration and kernel-bypass, balancing latency with durability. #STACSummit

28