Theory Ventures has more engineers than investors. Here's why that matters for dev tools.

Theory Ventures has more engineers than investors. Here's why that matters for dev tools.

BackerLeader posted 4 min read

Why Theory Ventures Built a Developer Relations Team—And What It Means for Developers

Most venture capital firms treat developers as an afterthought. They chase enterprise deals, court C-suite executives, and make investment decisions based on spreadsheets and market projections. Theory Ventures does the opposite.

The $680M AI and data fund runs more software engineers than investors—four technologists to three investors. And they're not just for show. These engineers build agents, run hackathons, and debug tool calls alongside the developers at their portfolio companies.

"We believe a deep technical understanding helps us invest better," says Tomasz Tunguz, Theory's founder and a veteran of enterprise software investing. "We're playing around with the technology and using it in deployment. That gives us a sense for where the market is going."

This isn't marketing. Theory's head of AI was hired with a specific mandate: reimagine a venture capital firm running 1,000 agents every day. The team deploys MCPs, uses GEPA and DSP, and hosts quarterly internal hackathons. They live in the same environment you do.

Developers Don't Pull Punches

Tunguz knows developer feedback differs from traditional enterprise customer feedback. "Brutally honest," he says. "Developers pull no punches."

You can't sell developers on buzzwords or polished demos. They want to know which keyboard shortcuts you use and whether you've wrestled with Python environment problems. They spot fake expertise instantly.

"Being able to walk the walk is really important," Tunguz explains. "If we aim to invest in developer tooling and AI systems, we want to learn alongside developers implementing internally, just the way they implement internally."

Theory measures their developer relations success through engagement metrics—events hosted, positive feedback, community involvement. Like GitHub stars, these indicators sit a degree or two away from financial performance. But they matter because attention has become the scarcest resource in tech.

"With AI synthesizing everything, the most important thing to differentiate and attract attention is being genuine," Tunguz says.

What Developers Are Actually Struggling With

Theory's portfolio companies give them direct insight into what's actually hard about building with AI. Tool calling tops the list.

"Figuring out why an AI decides to call which tool when and why it doesn't consistently select to use that tool; that's one of the most challenging tasks developers face," Tunguz says.

The architecture question compounds the problem. Should a three-part tool run as a hardcoded script, or should the AI make intermediate decisions? When accuracy drops to 75-80% across three steps, end-to-end reliability falls below 60%.

Code generation works brilliantly now. Tunguz can define a task, let an agent work for two or three hours, and get successful results. The whole job has shifted to architecture. "It's not a line of code that I'll be writing maybe ever again," he says.

The second big shift: developers don't leave their IDE anymore. No more tab-switching to Stack Overflow. Everything happens in a single flow state, with context-aware help right in the terminal.

How Investment Criteria Is Changing

AI moves faster than any market Tunguz has seen. "You wake up and the world has changed. There's some new academic paper, some new architecture, some Chinese company that has distilled an American model."

The fastest way to lose money in venture capital used to be investing in an IDE. Nobody except IntelliJ ever made money on one. Now agentic coding companies are the fastest-growing investments Theory tracks.

"The IDE may be less valuable than it once was because most of the software is being automated anyway," Tunguz observes. Cloud Code recognized this early by focusing on the terminal instead of panels and intermediate state views.

The biggest pattern Tunguz sees in successful AI startups: AI provisioning the product directly. When Databricks acquired Neon, 75% of new Neon instances were being generated by AI. An application fires up a Postgres instance on Superbase or Neon automatically. No human involved.

Developers Are More Important Than Ever

AI hasn't diminished the developer's role in enterprise buying decisions. It's amplified it.

Contract sizes are increasing. Budget decisions now span both software and labor costs. "Should we hire this AI agent instead of hiring 10 new software engineers?" Tunguz asks. "The developer is the person best equipped to determine whether the agent can perform that level of work."

Theory invests in tools that grow from communities. Heroku grew from Rails. Supabase from TypeScript. Understanding the momentum behind these communities helps predict how broadly a tool will be adopted.

The next 18 months will bring opportunities in AI site reliability engineering, security automation and review, and education for software engineers and marketing engineers. But the biggest opportunity: staff-level software engineers that are actually AI.

"We're not there yet," Tunguz admits. "Junior developer work from an AI? You can get there. Staff level or senior architects? Still a ways to go."

Theory Ventures proves that a VC firm can walk the walk with developers. They're building with the same tools, hitting the same problems, and staying brutally honest about what works and what doesn't.

That's the kind of investor developers deserve.

If you read this far, tweet to the author to show them you care. Tweet a Thanks

Really thoughtful piece showing how Theory Ventures flips the usual VC model by putting real developers at the centre. Feels like a refreshing change when investors actually code alongside their portfolio teams.

More Posts

AI Stack for Solo Developers: Build an Entire Startup with These Tools

Ijeoma JahsWay - Aug 11

One small company built successful AI with five software engineers—here's what big companies miss.

Tom Smith - Sep 22

AI platform & Gaia RAG help devs build secure apps with quantum-proof encryption & protection.

Tom Smith - Jun 21

The DevRel Advantage: How Developer Relations Drives Business Growth

excellus - Mar 17

TwitterAPI.io: Simplified Twitter API Integration for Developers

kaitoInfra - May 8
chevron_left