Symbia Labs creates at the frontier of human and AI cooperation.
Your team can build software with AI faster than ever, and you've seen what happens next: the project stalls because no one has a stake in it, the people it's for don't benefit, and the business never sees a return. We start with who it's for and what it's worth, then take the idea from first sketch to shipped software and show our work: what we built, what we said it would do and what it did.
What we do
Bring us an idea at the whiteboard or a build that has stalled halfway. We do the strategy, design, build and deployment ourselves, so nothing gets lost in a handoff between firms.
Strategy and architecture
A platform choice in AI, IoT or cloud commits a team for years. Our advice comes from more than a decade of industrial IoT and data work at Splunk, InfluxData and HiveMQ. Engagements cover architecture review, build-versus-buy, vendor evaluation and roadmap planning.
Who's behind it →Prototypes and production software
We design and build enterprise, cloud, desktop, mobile and embedded software, from the first working prototype to the shipped product. That includes equipment with no published interface: we work out how the device communicates and build software that controls it, confirming every change with the device itself.
See our work →Building with AI agents
AI agents write code quickly, and someone still has to stand behind it. We set up teams to build software with agents and train them to use it. Symbia, our open source tool, keeps a signed record of every command and file change an agent makes through it, so a reviewer can check the work later without access to the machine that ran it.
About Symbia →Apps we build to stay at the frontier
We build our own apps to close gaps that hurt users and the independent developers who build for them. Some of the apps are on sale and some are still in development. Every one is also an exercise that keeps our skills at the frontier, and a working example of what we can build for you: controlling hardware with no published interface, getting through Apple's App Store review, and letting Claude or another AI agent run the app alongside you through MCP, the open standard for connecting agents to software. None of them asks for an account or sends anything to us.
| App | Works with | What it does | For AI agents | Price and status |
|---|---|---|---|---|
 Bias |
TASCAM Model 12, 16, 24 and 2400 | Runs the Audio Unit plugins you own on every channel, live or on playback, and prints the result to the SD card. Markers and takes follow the mixer's recorder | Reads the meters and works the inserts, songs, markers and takes | $9.99 on the Mac App Store |
 Yoke |
Universal Audio OX Amp Top Box | Puts every OX on your network on one screen, on the Mac, iPad and iPhone, and reads every change back from the unit | Runs the whole rig, with the same readback you get | $4.99, in App Review |
 Room |
AV receiver or interface, LiDAR iPhone or iPad Pro, Apple TV | Shows where each speaker sits, which way it points and how the bass changes across the room, measured with music you know | None | $9.99, in App Review |
 Rig |
MOTU MIDI Express XT, Nektar Pacer and Panorama T4, EHX Super Switcher, Korg ARP 2600 M, TASCAM Model 2400 | Keeps the MIDI routes you draw and puts your sample rate back when another app changes it. Replaces the MIDI Studio and Audio Devices windows in Audio MIDI Setup | The same work you do in the window, through 42 tools | Beta |
 Play |
Your Audio Unit instruments | Your agent writes a song's parts and finds a sound for each one among the instruments you own. Every part plays on its own output, ready to record | Writes, voices and performs the song, through 36 tools | In development |
White papers
We publish the measurements behind our products, including where they fall short. Each figure is checked against an outside reference, such as a top chess engine or the published math for an audio filter, so you can see exactly how we test.
Ruse: audio plugins built from blocks, with gates
In music software, a plugin that crashes takes the whole recording session down, and runaway feedback can blast the speakers at full volume. Ruse is our framework for audio plugins that an AI agent writes and no person reads, so it has to catch those faults itself. Every plugin is built from tested signal-processing blocks and must pass seven automated checks before it can be packaged. The paper also lists what doesn't hold up, down to a filter that misses its documented response by 3 dB.
- 7Gates every plugin passes before packaging
- 281 of 281Tests that don't need macOS, passing on x86_64 Linux
- +382 dBFSRunaway feedback the gates caught in the case study
Mate: a legible chess engine that tutors
The strongest chess engines score positions with neural networks, so their numbers are accurate and hard to explain. Mate, our chess engine, builds every score from five ideas a club player already knows, such as material, loose pieces and king exposure, and explains its play in plain sentences. Fixed templates write those sentences with no language model involved, and an independent checker tests each one against the board.
- 1870 ± 80Rating on Stockfish's UCI_Elo scale at 10 s + 0.1 s
- 0 of 922False claims in its sentences about the opponent's moves
- 129 of 136Games won against people on Lichess
Room: measuring the listening room
Home theater room correction measures sound at the listening seat, so it can't say what to move or where acoustic treatment belongs. Room, our measurement app, maps the room's shape along with its sound. The depth sensor on an iPhone or iPad Pro records the room as you walk, while an Apple TV plays music through your own system. The paper places each speaker to about an inch and its aim to within ±10°.
- about 1 inSpeaker position
- ±10°Direction each speaker points
- about 2 msPhone and Mac against the Apple TV's clock
Bias Tape: a multitrack tape machine model
Recording software that imitates analog tape usually runs a separate tape machine on every track, each drifting on its own. Bias Tape, part of our Bias app for TASCAM's Model mixers, treats the whole session as one tape machine. The tracks drift together and bleed into their neighbors the way they do on a real multitrack recorder, in real time on every channel.
- 0.2 dBFrom the target level for 3% third-harmonic distortion, three tape formulations
- 0.62%Of real time per mono channel, M-series laptop
- 3Predictions that broke, all reported
Symbia, Ruse and Mate
Three projects we publish under the MIT license, free to use in your own work.
Symbia
When an agent tells you the tests passed, Symbia lets you check. Every command and file change it makes through Symbia goes into a signed, tamper-evident record, with predictions written down before each test and results after. Anyone you share the record with can verify it on their own computer.
Ruse
Describe an audio effect and your AI agent builds it with Ruse as a plugin for music software such as Logic. Nobody may ever read the agent's code, so Ruse won't package a plugin until it passes seven automated checks for faults like runaway feedback. Developers can sell closed-source plugins built with it, with no license fees.
Mate
Mate is a chess engine that explains itself. Every score breaks down into five things a club player already understands, like loose pieces and king exposure, and a separate checker tests everything it says against the board. Playing as Symbia on Lichess, a popular online chess site, it has won 129 of 136 games against people.
Shipped in October
| Date | Shipped | What it is |
|---|---|---|
| Oct 29 | Mate, open source | The chess engine from the Mate paper, with its automated checks and the bot that plays on Lichess. MIT license |
| Oct 22 | Bias on the Mac App Store | Our Mac app for TASCAM's Model mixers, which runs a musician's own audio plugins on every channel. One-time purchase, $9.99 |
| Oct 19 | Ruse 0.2.0 | Our framework for building audio plugins with an AI agent, now with fifteen tested building blocks and seven automated checks before packaging |
| Oct 17 | Six Ruse example plugins | Six finished example plugins built with Ruse, ready to install |
Who's behind it
Brian Gilmore founded Symbia Labs after more than a decade of putting new technology to work in real operations, as Chief Technology Advocate at Splunk, Director of IoT and Emerging Tech at InfluxData, and VP of Community and Advocacy at HiveMQ. His US patents in edge analytics and asset-tree monitoring are cited by 133 later patent publications from 60 companies, Amazon, Cisco, Hewlett Packard Enterprise, IBM, Huawei, Ericsson, ABB, Siemens, Honeywell and Intel among them, by Google Patents' listings on October 29, 2026.
- Splunk2013–2021
- InfluxData2021–2022
- HiveMQ2023–2024
- Machine learning in edge analytics, US 10,460,25576 citing publications
- Analytics for edge devices, US 10,536,35131 citing publications
- Control interface for asset tree monitoring, US 10,503,78427 citing publications
- Companies whose later patents cite these families60
Tell us what you're building.
An idea that needs a prototype, a system that has to ship or a platform decision to make: write and tell us what you need.