Naptha AI

📰 Honored to be featured in StartupHub.ai! The article explores our vision for decentralized multi-agent systems and how we're building the infrastructure for the Internet of Intelligence. A big thank you to Daniel Singer for capturing our journey and mission to enable seamless collaboration between AI agents across frameworks and industries. Read more about our work and vision here: https://lnkd.in/dzwwdxv3 #AI #DecentralizedAI #MultiAgentSystems #StartupNews

Last week we made it to the front page of Hacker News with our post about benchmarking how well coding agents interact with libraries and APIs https://lnkd.in/eDD7e6Cn The response was great overall, but many wanted to see the code. We also wanted to see if we could grow a community around it, as we believe it’s an area that’s important and underexplored. StackBench is Now Open Source! https://lnkd.in/eReqQd8c For those just catching up: StackBench tests how well AI coding agents (like Claude Code, and now Cursor) actually use your library by: • Parsing your documentation automatically • Extracting real usage examples • Having agents generate those examples from a spec from scratch • Logging every mistake and analyzing patterns You can find out more information about how it works and how to run it in the docs https://lnkd.in/eFb4T_Rb Next up, we’re planning to add more: • Coding agents • Ways of providing docs as context (e.g. Mintlify vs Cursor doc search) • Benchmark tasks (e.g. use of APIs via API docs) • Metrics We're also working on automating in-editor testing and maybe even using an MCP server. Contributions and suggestions very welcome. What should we prioritize next?

Are you at #OSSummit in Amsterdam and struggling with activation and retention of devs using your library or platform? Our product turns your docs into in-editor agent-guided walkthroughs, delivered via MCP. Send me a DM if you'd like to chat and see a demo!

Brian Balfour just made a prediction that should change how every tech company thinks about distribution: AI will become the next major growth platform within 6 months. For developer-focused companies, winning distribution means being the obvious choice for both coding agents and humans. If your library is easy to find, simple to use, and reliable in automated workflows, you’re ahead. Seamless integration in environments like IDEs, code editors, and terminals is key. Developers need guidance and examples right when they hit a wall. Are you building for coding agents? I’d love to hear your thoughts, lessons, or questions below. https://lnkd.in/evrKi36Z

"95% of AI Pilots Are Failing" is trending on Hacker News right now. TL;DR: The pilots fail not because the technology is bad, but because nobody learns how to use them effectively (https://lnkd.in/eaWUvMG4) The pattern is predictable: Company deploys shiny AI tool, provides generic training session, expects users to "figure it out," then users hit the 80% mark and abandon. The problem isn't capability. It's the learning gap. We believe the solution is to bring onboarding into the user's work environment, track struggles and offer targeted support, and create continuous feedback loops. If you're facing issues with increasing activation and retention of users on pilots (or external-facing products), I would love to chat and see how we can help https://lnkd.in/egddsq7h

If coding agents are the new entry point to your library, how confident are you that they’re using your library effectively? Over 90% of developers now use AI in their workflows. Yet we keep hearing the same concerns: 🚩 Coding agents rely on outdated versions and import deprecated functions. 🕵️♂️ We don’t know how well coding agents use our libraries and APIs. Existing code generation benchmarks focus on self-contained code snippets and compare models, not agents. So how can we truly measure how well coding agents interact with your library? Introducing StackBench https://stackbench.ai: ✅ Ingests your software library documentation ✅ Automatically extracts usage examples ✅ Tasks Claude Code with generating examples from scratch ✅ Logs mistakes and analyzes performance We’re currently testing various libraries. It’s early days, but we shipped fast so you can help shape what’s next. Now we need your input... Try it out: Input your library, see what breaks, spot patterns, and share your results. Let’s learn together how coding agents break. We’re just getting started. We plan to expand StackBench with more coding agents, more library-specific generation tasks, and more metrics. What features or agents should we prioritize next? https://lnkd.in/eAbZ3pjs

Who is coming to Europe for a late summer AI tour? Sep 23-24, AI Engineer Paris, https://lnkd.in/e5nU-G_D Sep 29-30, Curated AI Dublin Event (DM me) Oct 2, MCP Developers Summit London, https://mcpdevsummit.ai/

What's your approach for passing docs to coding agents? https://lnkd.in/eRdVSURp

Wondering why your users don’t activate? You gave them APIs and docs… but something’s missing. What if emerging technologies could help? As a service provider, exposing your product through APIs or SDKs gives developers access — but often leaves you in the dark about how your service is being used, what users are trying to achieve, and where they’re getting stuck. Most integrations today suffer from three key limitations: 🧩 Hide context across sessions and systems — you have little visibility into what users are trying to accomplish ⏱️ Lack real-time adaptation — you can’t respond to shifting user needs or usage patterns as they unfold 🔁 Don’t support personalized negotiation or multi-step discovery — users follow fixed flows, regardless of context or intent This limits your ability to provide dynamic support, improve onboarding, or identify opportunities for expansion and optimization. Agent protocols like Model Context Protocol (MCP) and Agent2Agent (A2A) change this by making the interaction between users (or their agents) and your service bidirectional, contextual, and intelligent. Imagine being able to: ⚠️ See where users hesitate: Are they unsure which parameter to use? Are they trying the wrong combination? 📉 Understand drop-off points: Which part of your onboarding flow or configuration journey causes the most abandonment? 📈 Identify growth levers: Where are users organically trying to do more with your service than your current model supports? If you're building a developer platform or API-driven service and this sounds like the future you want to live in, get in touch. We’re working on the infrastructure to make it real.

Should we build something to make this better? https://lnkd.in/g3R4YRff