You're building in public, Reader

Your GitHub stars are climbing. Contributors are active. The community Slack has real conversations happening daily.

So you assume this will convert into customers. It doesn't.

Mrinal Wadhwa built Ockam with hundreds of contributors and customers like AWS, Databricks, and Snowflake. Then in 2024, he pivoted to build Autonomy, a platform for long-horizon AI agents. I wanted to understand what he learned about the gap between community building and customer acquisition, and why a recent capability threshold means almost everyone needs to rewrite their agents.

Community Growth Isn't Customer Acquisition

Mrinal and his co-founder did everything right with Ockam. They validated the problem before writing code. They built a Slack community that grew to over 100 people before they had a working prototype. They made a strategic choice to rewrite their initial version in Rust, right as that language community was exploding.

Then they established weekly discipline. Before every Wednesday, they'd write down the week's learnings and submit a pull request to "This Week in Rust," a rapidly growing newsletter. Week after week, they showed up. Their repository became the go-to example for building complex distributed systems in Rust.

The community grew to hundreds of contributors and thousands of participants. And then they discovered the problem.

The people contributing were mostly junior developers early in their careers. Excited about the technology, willing to contribute code, engaged in discussions. But they didn't have architectural decision authority inside their companies. They weren't the ones with budget to buy infrastructure tools.

The actual customers (senior architects, CTOs, technology leaders) were at completely different conferences. They needed different messaging. Mrinal had to build an entirely separate motion to reach them, largely through personal networks.

In other words, community building and customer acquisition aren't the same muscle.

Map Your Personas Separately

Your OSS contributors are different from your buyers. Don't expect automatic conversion between these groups.

Community-building channels: developer conferences, online communities, educational content, contribution pathways. Customer channels: enterprise conferences, direct outreach, ROI-focused content, proof of concepts with decision-makers.

Track different metrics too. Community side: contributors, GitHub stars, participation, educational engagement. Customer side: qualified leads, POC conversions, design partnerships, revenue pipeline.

The community is still a great thing even if it doesn't directly convert. It validates that you're solving a real problem. It creates proof points when talking to customers. It generates feedback that improves the product. Just don't mistake it for a sales funnel.

The 3-Month Capability Leap Everyone Missed

Something fundamental changed in the last three months. If you built an AI agent in 2023 or early 2024, you're probably going to rewrite it.

Most agents built in the last two years can execute 2 to 3 autonomous steps. Organize your inbox. Move messages into a category. File something away. These automate minutes of work.

Starting around October 2024, agent capabilities crossed a threshold. New architectural approaches enable agents that execute hundreds of autonomous steps. These long-horizon agents automate days of work instead of minutes.

The breakthrough came from a seemingly simple change: instead of relying primarily on vector stores, give agents a file system. Give them a workspace. Give them access to traditional Unix command-line tools.

This architectural shift showed up most visibly in coding agents. Claude Code and similar tools became dramatically better than the previous generation. They could tackle complex, multi-step tasks that earlier agents struggled with.

Autonomy is designed for this new generation. One of Mrinal's recent demos is an app with 5,000 agents collaborating to solve a problem. The agents most companies built in 2023 automate tasks that take minutes. The agents now possible automate workflows that take days.

Agent Swarm Architecture for Higher Accuracy

One of Autonomy's pharmaceutical customers is compressing drug approval timelines from 2 years to 1 year. A significant chunk of that time savings comes from a process that previously took weeks and now takes minutes.

The specific problem: before submitting a drug application to the FDA, thousands of documents need cross-references inserted manually. Someone has to work through thousands of documents for multiple weeks, inserting references like "this compound is described in document 1501, that trial result is in document 2247."

Autonomy solves this with agent swarm architecture. A parent agent orchestrates the work. It spins up a child agent for each document. Each child agent focuses only on its assigned document.

Because each individual agent only deals with one document, its job is manageable. The context load is small. The accuracy is high. If you gave this entire task to a single large agent, it would struggle because the context is too vast. But split across hundreds of specialized agents, each with narrow focus, the success rate goes up dramatically.

The underlying principle: large context load to one agent equals lower accuracy. Small, focused context per agent equals higher accuracy. Parent agents handle coordination. Child agents handle specialized work within their narrow domain.

What You Can Apply This Week

If you're building with an open-source component, separate your community motion from your customer motion now. Map the personas. Build different channels. Track different metrics. Don't wait to discover the gap when you need revenue.

If you're building agent products, consider whether you're still designing for the 2023 capability level (2 to 3 autonomous steps) or the current threshold (hundreds of steps). The technology shifted. Your architecture might need to catch up.

If you're processing multiple similar items (documents, logs, applications, footage), test the agent swarm pattern. One specialist agent per item. One orchestrator coordinating. Parallel processing. Clear audit trails.

And for more details on Mrinal's story and what he's currently building at Autonomy, read the full article: https://www.data-mania.com/blog/from-city-floodgates-to-ai-agents-what-mrinal-learned-building-trust-infrastructure-for-autonomous-systems/​

All the best,

Lillian Pierson

Growth Partner & Fractional CMO

​
​
​
​