Background Agents: The Sandbox Symposium
Background Agents: The AI Tinkerers Sandbox Symposium

A live, competitive research hackathon where engineers put code sandboxes head-to-head.
Why
Between Ramp’s background agent writeup, StrongDM’s write-only software factory, gas town, ralph wiggum and a ton more, background coding agents and the infrastructure to build them has captured the current zeitgeist. Daytona.io raised a 24m series A, half the presentations at our last OpenClaw hackathon were around building secure remote compute for OpenClaw, the list goes on and on and on.
Background agents aren’t going anywhere, there’s 10 new sandbox infra startups every week, and founders and builders everywhere want to know: what’s best for my use case? What are the tradeoffs?
Here’s your chance to:
- Cut through the hype
- Get your hands dirty with as many sandboxes as you can
- Try them out for all your use cases
- Come to your own conclusions
- Share your learnings with other builders!
If you ever heard of the old school Plug Test (Wikipedia) we’re doing that and more!
The Event
The Sandbox Symposium is a hackathon with a twist: participants don’t build apps -- they build research. Teams evaluate as many developer sandbox platforms as they can, then present their methodology, findings, and recommendations to a live audience.
Together we’ll run coding agents, dev servers, web UIs, batch workloads, and much more against all the coding agent sandboxes on the market.
This is the first event of its kind. The sandbox ecosystem is exploding with options, and developers are starving for honest, hands-on comparisons. We’re giving them a stage to do that work, giving sandbox providers a front-row seat to real feedback.
How It Works
We have some incredible sponsors from the best sandbox providers in the world coming to support this event!
Kickoff – Each sponsor gets a 5-minute demo slot to introduce their platform to participants.
The Showdown – Participants spend the day evaluating sandboxes across multiple dimensions, using sponsor tables for setup help, debugging, and feedback.
Use as many sandboxes and providers as you want, this is not limited to just sponsor tech.
The Rules – Planning up front is allowed, but all research data and supporting code must be written/excuted during the hackathon.
Presentations – Teams present their research: what they tested, how they tested it, and what they found.
Judging – No judging, no prizes, just builders learning from each other
| Criterion | What We’re Looking For |
|---|---|
| Security | How do sandboxes compare on isolation, permissions, and trust boundaries? |
| Performance | Speed, latency, resource efficiency across platforms |
| Portability | Interoperability, usability for different coding agents |
| DX / AX | Developer experience and agent experience -- ease of use for humans and AI |
Bonus: breadth of research (number of sandboxes tested).
While there will be sponsors present and its a great opportunity to engage with their tech teams, you are welcome to explore any sandbox tech you think might fit your use case!
What you don’t get:
- A rigged competition. Vendors cannot submit research or compete. The findings are intended to be independent.
Sponsors
Blaxel

Blaxel is the Perpetual Sandbox Platform - it keeps infinite secure sandboxes on automatic standby so your background agents never wait for cold starts. With 25ms resume times, full memory and filesystem state preservation across suspend/resume cycles, and support for 50,000+ concurrent sandboxes, Blaxel is built from the ground up for the background agent era. Idle sandboxes cost zero. Active sandboxes colocate your agents, MCP servers, and compute on the same backbone for sub-millisecond latency. Come by the Blaxel table to see how perpetual sandboxes change the way you think about agent infrastructure.
Daytona

Daytona gives every agent a computer. Their programmatic, stateful sandboxes launch in under 90ms and support forking into parallel branches, mid-execution snapshots, and full computing environments with CPU, memory, storage, and GPU — composable on demand. Fresh off a Series A led by FirstMark Capital (with strategic backing from Datadog and Figma Ventures), Daytona is already powering agents at
LangChain, Turing, Writer, and SambaNova. Come to the Daytona table to get hands-on with sandbox forking, snapshotting, and native Git integration.
Ona

Ona is the enterprise platform for AI software engineering agents. Born from years of building cloud development environments, Ona now gives every agent a full, connected cloud environment with your tools, network access, and permissions — not just a sandbox, but a production-adjacent execution context. Agents work autonomously in the background: task in, pull request out. With VPC deployment, kernel-level policy enforcement, scoped credentials, and SOC 2/GDPR certification, Ona is built for teams where security and auditability aren’t optional. Their customers report a 4x increase in development throughput, and inside Ona’s own engineering org, agents co-author 83% of merged pull requests. Come by the Ona table to see how connected environments change what’s possible when agents have real access to your stack.
Venue Sponsor: Tandem

Thanks as always to the wonderful folks at Tandem for hosting our hackathon event. Tandem is an AI-native office leasing marketplace that helps companies find and secure flexible workspace in as little as 14 days. They’re generously providing the space for the Sandbox Symposium. If your team is looking for flexible office space in SF or NYC — month-to-month terms, no multi-year commitments — swing by and say hi.
Interested in Sponsoring?
Reach out to [email protected] to learn more.