In development
microWASM
Sandboxes and micro VMs built on WebAssembly, with durable state. The goal is fast, portable execution across machines and environments.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- MS-A2
Projects / Work in progress
What’s on the workbench.
Open source software, model research, hardware experiments, and the infrastructure that supports them. Hardware listed here is shared across projects; it does not indicate reserved capacity or available allocations.
Updated September 15, 2026.
Open source software
6 projects
In development
threaded
Combines microWASM and celld to multiplex agent deployments. Designed to scale from a single CPU machine to full cloud clusters.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- MS-A2
In development
fxClaw
An agent system in the vein of OpenClaw, Hermes, PicoClaw, and NanoClaw, built in Zig. It hosts agents natively through microWASM, with fx.sh as the default foundation.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- MS-A2
Ongoing development
freeinferencing.com
Work on freeinferencing.com and its related open source projects.
Visit freeinferencing.com ↗- Project lead
- Shannon Duncan ↗
- Hardware & resources
- MS-A2
In development
ZD Web Server
Updating the ZD harness and editor to serve over the web, with a headless way to work with agent projects across the cluster.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- N150
Experimental prototype
esp32-fx
Runs the fx coding agent on native Linux on an ESP32-S3. WAMR runs the fx WebAssembly core, and a Seeed reTerminal E1001 e-paper display shows the agent’s response. Model inference currently uses a remote provider.
Explore the esp32-fx source on GitHub ↗- Project lead
- Shannon Duncan ↗
- Hardware & resources
- N150
- Seeed reTerminal E1001 with ESP32-S3
GPU & model research
2 projects
Active research
Practical Small Model Usage
Getting useful, high-quality results from small models on memory-constrained systems. The research also explores mixtures of models and swarms of task-specific 500M-parameter models.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- DGX Spark
- GPU bench
- Working toward
- Run 2B–30B-parameter models well on systems with 16 GB of memory or less. These are research goals, not published benchmarks.
Active research
DeepSeek Flash 4.1
Work on running DeepSeek Flash 4.1 across a pair of DGX Sparks.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- 2 × DGX Spark
Hardware experiments
4 projects
Prototype testing
Agent Blade
A clusterable board for swarms of agents. Autonomous agents and very small local models have run on ESP32 prototypes, with battery-powered operation intended to continue through internet or power outages. Current experiments include small SoCs, low-power GPUs mounted behind the blades, and older Android phones with 2–4 GB RAM as inference accelerators.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- Microbench
- First prototype: 6 × ESP32-S3 + 3 × ESP32-P4 on one PCB
- Working toward
- Under $100 per blade board and under 200 W, preferably 30–100 W. Cost and power are design targets, not measured results.
Exploratory research
FPGA model inference
Testing FPGA approaches to encoding model weights and tensor operations directly in hardware.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- Microbench / FPGA platforms
- Working toward
- Explore whether a single card could eventually run a 12B-parameter model at around 13,000 tokens per second. This is a long-term ambition, not an achieved result.
In development
An agent on your wrist
An ESP32-S3 watch or wrist-worn system designed to keep a personal autonomous agent with you.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- K15
- ESP32-S3 wearable
- Working toward
- One autonomous agent running around the clock. Runtime and battery life are still being explored.
In development
A tiny pet on your wrist
A Tamagotchi-style watch or wristband game Shannon is building with his daughter. She is designing it and working with Claude Code and an ESP32-S3 touchscreen to bring it to life.
- Project lead
- Shannon’s Daughter
- Hardware & resources
- Microbench
- ESP32-S3 with touchscreen
Cluster infrastructure
4 projects
Setup in progress
Cluster monitoring & HUD
Setting up Prometheus, cluster monitoring, and the cluster HUD, with the N150 as the controller and head unit.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- N150 / controller
- Cluster HUD
In development
GPU cooling shrouds
Building new cooling shrouds for the Lab’s P100, P4, and V100 cards.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- P100, P4, and V100 GPU cards
Planning
Community provisioning
Planning how to give people secure access through dynamic VLANs, virtual machines, and attached hardware. Isolation and provisioning arrangements are still being worked out.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- Cluster network, VM hosts, and attached hardware
- Specific allocations not yet set
Planned community access
LiteLLM gateway
Preparing a gateway for models served from the cluster, including the DeepSeek Flash 4.1 work, so community members can try them using keys issued through compute grants.
- Project lead
- Shannon Duncan ↗
- Hardware & resources
- Cluster model-serving resources
- Gateway host not yet specified
- Working toward
- Grant-based API access for community testing. The gateway and key distribution are not yet available.
Building something of your own?
Share it with the community, ask for feedback, or tell us what you’re learning. A compute application is not required.
Connect with the community →Need hardware for an idea?
For open research and projects, you can request free compute time. Tell us what you want to try and which resources would help.
See how compute requests work →Good company for your curiosity
Bring what you’re working on.
Or what you’re wondering about.
You don’t need a project or Lab access to take part.