
Instrumentation
Hive Labs builds instrumentation software for systems that act on the human body.
Control software, device integration, and research-grade data capture for hardware where a mistake is not a bug report.
- 1,985
- automated tests passing on our control platform
- 186
- peer-reviewed papers across 29 domains in our research library
01
What we build
- 1.1
Real-time control for physical hardware.
Multi-device orchestration across Bluetooth, serial, MQTT and multichannel audio, running on a scheduled timeline with sub-second coordination. Hardware watchdogs, latching emergency stop, interlocks, and per-session exposure limits enforced in software rather than documented in a manual.
- 1.2
Device integration and fleet operations.
Calibration and commissioning workflows that produce a record, not a checkbox. Cryptographically signed over-the-air firmware updates. Fleet telemetry and firmware-version drift detection across deployed units.
- 1.3
Biosignal acquisition and analysis.
Continuous physiological capture with artifact rejection, per-channel contact quality, and montage-aware analysis. Our processing reports not only a measurement but how far that measurement can be trusted on the sensor that produced it.
- 1.4
Research-grade data capture.
Blinded arm assignment, deterministic replay, protocol hashing, per-decision logging, and structured export with a data dictionary. Participant identity is held behind an access control boundary so that a data export cannot disclose a client list.
- 1.5
Regulated-adjacent data handling.
Encryption of protected health information at rest, per-operator authorization, auditable action logs, configurable retention and purge, dependency vulnerability gating in continuous integration, and transport security.
02
How we work
- 2.1
We build the refusal before we build the feature.
A system that acts on a person needs to be able to say no. Our control software blocks a hazardous configuration at load time rather than trusting an operator to catch it, and it fails closed when a precondition is missing.
- 2.2
We make silence impossible.
The failure mode we design against hardest is the one where a system reports normal operation and does nothing. Every path that can drop a command records it, surfaces it, and marks the session record accordingly.
- 2.3
We say what a number can carry.
A reading taken from a sensor that cannot see what it is pointed at is not a measurement, it is a number. Our displays state where a value came from and flag it when the hardware cannot support the interpretation.
03
Research foundation
Our work is built on a standing research library of 186 peer-reviewed papers across 29 domains, maintained in-house and organized by mechanism rather than by marketing category. It spans brainwave entrainment, closed-loop neurostimulation, photobiomodulation, bioelectric field research, cross-frequency coupling, respiratory and cardiac coupling, and adjacent literature.
The library has a quarantine. Sources that do not survive scrutiny are moved out of it rather than quietly left in, and the reasons are recorded.
Method
We audit our own claims downward.
In 2026 we reviewed every protocol in our library carrying a top evidence tier against a single question: is there a paper supporting the outcome it names, not the mechanism and not the frequency band. Most did not survive. They were re-tiered and rewritten, and claims the literature had tested and found did not hold were removed entirely.
That audit cost us most of our strongest-sounding language. We made the cut anyway, because a claim we cannot defend is a liability wearing a marketing jacket.
04
The question the instrument was built to answer
Most devices in this field run a fixed program. Ours can also adapt in real time to what it measures. Whether that adaptation actually changes anything is an open question, and it is not one you can settle by asking people afterwards how they felt.
So the instrument was built to test it properly. One condition delivers stimulus that responds to a participant's own physiology. The other replays the verbatim recorded timeline from a participant in the first condition, so the stimulus is identical and only the responsiveness differs. Arm assignment is blinded, every adaptive decision is logged with its inputs, and a session can be replayed deterministically from its record.
Status
We do not have a result. We have an apparatus capable of producing one, and we would rather say that than market a conclusion we have not earned.
We are looking for an academic research partner to run this properly.
05
Why this work
This work started from a straightforward conviction: that precisely delivered physical input can change how a person feels and functions, and that the field has been long on claims and short on instruments. We build the instruments.
The question itself is an old one. It has been studied seriously for sixty years and productized carelessly for thirty: how precisely delivered physical input changes the state of a body and a mind. Brainwave entrainment is the part of that field we work in.
That question deserves better engineering than it usually gets. Most of this field runs on confident claims and thin instrumentation. We would rather have honest instrumentation and fewer claims, and then go and earn the claims properly, with measurement, with controls, and with a design that can tell us we were wrong.
We are building toward research that can be reviewed by people who have every reason to be skeptical. Everything above is in service of that.
06
Engagement
We take on control and instrumentation work for hardware teams who need the software layer to be as rigorous as the device: medical-adjacent instruments, laboratory and research apparatus, and physical systems where safety, auditability and measurement integrity are requirements rather than aspirations.
We also design and build custom brainwave entrainment instrumentation, and license our own platform under partnership terms.