About Jon Radoff

About Jon Radoff

Get in Touch

Interested in collaborating, speaking, or advisory? Reach out here.

I'm Jon Radoff—entrepreneur, builder, and writer. I've spent three decades at the intersection of games, technology, and the internet. Today, my work centers on agentic AI: the shift from software we use to software that acts on our behalf.

What I Do

Metavert is my venture studio and research practice. Through it, I'm a strategy advisor to companies building at the frontier of AI, games, and interactive technology. I write the Metavert Meditations newsletter, where I explore how agentic systems, creator economies, and composable architectures are reshaping every industry. I also build things—open-source tools, platforms, and experiments that put these ideas into practice.

My current focus is on what I call the direct-from-imagination era: the convergence of large language models, agentic engineering, and composable infrastructure that is collapsing the distance between having an idea and shipping it. I believe we are entering the Creator Era of software—where natural language becomes a compiler, where AI agents handle implementation, and where the bottleneck shifts from engineering capacity to imagination.

What I bring to the companies I work with:

  • Agentic engineering transformation — moving teams from outside agencies and vendors to AI-agent-driven development, with the practices, release process and tooling to sustain it.
  • Production AI agents — analysts that write and run code over live data (Recursive Language Model architectures), knowledge systems with RAG and semantic search, and MCP integrations.
  • Applied machine learning — churn, lifetime-value and payer-propensity models, automated discovery of a product's analytics signals, champion/challenger evaluation, and fine-tuning and benchmarking open-weight models.
  • Generative media — text-to-speech and voice, diffusion image generation, video and image analysis, and automated video production.
  • Product execution — onboarding, subscriptions, monetization, product analytics, growth and attribution, and children's-privacy readiness.

Background

I co-founded Beamable, a live game infrastructure platform used by studios building multiplayer experiences, and led it as CEO until Skillz (now Firy, NYSE: FIRY) acquired its technology and team in February 2026. Today I'm Beamable's Chief Product Officer, leading its AI products for live games.

Before Beamable, I founded Disruptor Beam, where I was game director and architect of the monetization strategy for Star Trek Timelines—a mobile game that reached millions of players and recently passed its tenth year of continuous operation, placing it in a rare 1% of games with that longevity.

I've been building online games and internet software since the early days of the medium. My first game, Legends of Future Past, was one of the pioneering online multiplayer RPGs. I went on to found Eprise Corporation, an enterprise content management company that went public on NASDAQ. Through these ventures and others, I've shipped games, developer tools, and content platforms across every major era of the internet.

I'm also an active investor in private companies working on AI infrastructure, games, and interactive technology. See my investment portfolio.

Agentic AI and Building in Public

I practice what I call agentic engineering—using AI agents as collaborators across the full development lifecycle. I build in public, and most of my recent projects were created this way: Legends of Future Past (my 1992 MUD, resurrected in a weekend), Project Stapledon (a deep-time strategy game), MetaRush (an AI video creativity and editing harness built on video as code), LightCMS (an AI-native CMS with MCP), LastSaaS (a SaaS boilerplate built entirely through conversation), VibeCtl and markupmarkdown (tools for agentic teams), LLM Optimizer (AI search visibility), threelab and ShaderVine (generative art and shaders), Aitiome (an AI neurotoxicology engine for the Built with Claude: Life Sciences hackathon), Chessmata (chess for humans and AI agents), Agent Almanac (an AI agent directory discovered by an autonomous agent), and a HiddenBench harness for multi-agent collective reasoning.

Explore all of my open-source work on GitHub, or visit Projects for the full directory of current ventures, tools, and advisory work.

In June 2023, my team won the Grand Prize at the Gen AI Virtual Worlds Hackathon during a16z's LA Tech Week by building Tales of Mythos—a complete D&D-style RPG in a single day. We used Claude's 100K context window with what I called semantic programming: LLM prompts as functional subroutines within traditional software architecture. That experiment prefigured the agentic engineering workflows that are now becoming mainstream.

Writing and Ideas

My intellectual work focuses on a few connected threads:

I published the State of AI Agents & Agentic Engineering 2026—a 205-slide analysis covering $211B in AI venture capital, the SaaSpocalypse, and the forces reshaping every industry—along with the Market Map of the Agentic Economy and quarterly research digests on agent engineering and AI search citations.

Speaking

I've spoken at MIT Media Lab (MAS.S61), MIT Sloan, SXSW, GDC, GamesBeat Summit, Gamescom Congress, and numerous industry events on topics including agentic AI, the metaverse, game design, digital identity, and the creator economy. I also host livestreams on YouTube covering AI, agentic engineering, game development, and virtual worlds.

Education

A.B., Harvard University (Extension Studies), Psychology (field of study).

Teaching fellow, Genomics and Computational Biology (George Church), Harvard.

Connect

Substack  |  YouTube  |  GitHub  |  LinkedIn  |  X / Twitter  |  Contact