RFC 0001	TODO	12 2024
David Holtz
WithPhysics

Abstract

Our mission
To make the physical and natural world transparent and accessible to everyone.

Why
Everyone deserves a deep understanding of themselves and their environment.

How
By creating high-precision, affordable, and high-quality sensors that share data directly with you—eliminating intermediaries and gatekeepers.

What we do
Consumer Sensors - High-end desktop sensors designed for home and laboratory use. Reliable and precise tools for those dedicated to monitoring and understanding their environment.

Innovative Technologies - Advanced Near-Infrared (NIR) spectrometers paired with open-source software. Unlocking new applications and measurements, these tools empower users to explore and innovate in new domains with easy-to-use, accessible technology.

Details

Hardware has been evolving at an increasing rate, getting faster, smaller, and cheaper "Over the past several decades, the speed of computer hardware has advanced at a remarkable pace, often outstripping the capabilities of software" [1].

The rapid development is apparent in the consumer electronics market, but the same is not true for scientific instruments [2]. The cost of lab quality sensors and or sensors to understand the food and water we consume is still prohibitively expensive, and limited to a small subset of the population.

The goal of WithPhysics is to change this. We want to make high quality sensors that are accessible to everyone. We want to enable people to understand the world around them, and to make decisions based on data that they trust.

Current day sensors are often black boxes, with data being sent to the cloud for processing [3]. This is not ideal, and is a crutch due to the limitation of on-device compute [4].

Physically it's a waste of energy to send data to the cloud, additionally by sending data to the cloud we are giving up control of our data, and the attack surface of the system is increased [5].

Keeping compute near the sensor is a better solution, it's privacy preserving, more energy efficient and more personal.

Status

Draft

Rational

Using something like the ITN framework (importance, tractability, and neglectedness) [6] or the a simplified version of the Technology Readiness Assessment (TRA) [7] framework we'll explore the rational behind WithPhysics.

Importance/Criticality

Importance: Assessing the scale or severity of the problem.
Criticality: Their potential impact on critical systems and society

At a macro level, reducing the cost of measuring the world around us is a critical step in enabling a more transparent world.

As strong believers in open source and open science, we believe that the more people who have access to data and tools to process that data, the more likely we are to make progress as a society.

At a micro level, enabling individuals, organizations and small businesses to have access to high quality sensors will enable them to make better decisions.

This could be as simple as knowing the quality of the water they drink, or as complex as understanding the impact of their actions on the environment.

Currently our world relies on experts and intermediaries to provide us with information about the world around us. This is not ideal, as it introduces bias, and can lead to a lack of trust in the information provided - its important that we are able to verify the information we are given and verify the measurements that are made.

Specific Applications:

Environmental monitoring for humans/animals/plants [8]
Continuous monitoring of food and water quality [9]
Plastics and microplastics detection [10]
Verification of medications and supplements [11]
Narcotics testing [12]

Just looking at the final application of narcotics testing, the importance of this application is clear. The opioid crisis is a major issue in the US, and the ability to test drugs for purity and potency could save lives [13].

The World Health Organization (WHO) estimates that approximately 2.2 billion people worldwide lack access to safely managed drinking water services. This figure represents about 27% of the global population as of 2022. [14].

The global water filters market was valued at approximately USD 10.05 billion in 2023 and is projected to reach USD 14.97 billion by 2029, growing at a compound annual growth rate (CAGR) of 6.86% [15].

From an anthropological perspective and an economic perspective, the ability to measure the world around us has meaningful impact. Reducing the cost and friction has far reaching implications and are a critical step in enabling a more transparent, healthier, and more sustainable world.

Tractability/Maturity

Tractability: Determining how feasible it is to make progress in addressing the issue.
Maturity: The current maturity level and path to implementation

Its clear that the technology exists to build high quality sensors, and the core threads that make WithPhysics possible are already in place.

The cost of sensors is decreasing, and the quality is increasing. [16]
The cost of compute is decreasing, and the power efficiency is increasing. [17]
Artificial intelligence and machine learning research largely outpaces the application of these technologies in the real world. [18]
The open source community is strong, and there are many tools available to build on. [19]

in addition to the global trends, we have the skills, knowledge, and experience to unlock the potential of these technologies.

Building the hardware and devices is a challenge, but the really impactful work will be providing open software and solid foundations for others to build on.

We have the experience to build lab grade hard and software, and push the boundaries of edge AI and machine learning.

Unreasonable effectiveness of MEMS

Specifically, the cost of NIR spectrometers have dropped by an order of magnitude in the last 30 years.

More recently, there have been even more developments in MEMS spectrometers are pushing another order of magnitude in cost reduction.

This cost reduction has been made possible specifically because of MEMS technology and it's ability to improve in cost along with Moore's Law.

MEMS (micro-electromechanical systems) are a class of devices that are fabricated using techniques similar to those used in the semiconductor industry. MEMS devices are typically made of silicon, and can be fabricated using the same processes used to make integrated circuits. This allows MEMS devices to be made in high volume, and at low cost.

MEMS has already had a significant impact on the cost of sensors, and we expect this trend to continue. For example, MEMS has enabled the cost of air bag sensors to drop from several hundred dollars to a few dollars, and is the technology behind the low cost magnetic sensors used in smartphones.

Examples of MEMS impact on cost:

"Initial air bag technology used conventional mechanical ‘ball and tube’ type devices which were relatively complex, weighed several pounds and cost several hundred dollars... MEMS has enabled the same function ... into a single silicon chip, resulting in a tiny device that can be housed within the steering wheel column and costs only a few dollars" - An Introduction to MEMS. Page 12

"A typical optical switch can cost over $1000, but using MEMS, the same level of functionality can be achieved for less than a dollar." - An Introduction to MEMS. Page 18

"Monolithic processing of this device as well as the reduced number of parts enable a very compact device with high reliability at a very low cost." - An Introduction to MEMS. Page 31

"Most of today’s magnetic sensors are silicon based, not only because of the ease of fabrication and their ability to be readily integrated with circuitry, but also because their high volume demand necessitates lower cost." - An Introduction to MEMS. Page 41

Because of this cost reduction, we are excited to see the impact that this will have on the availability of open source spectrometry.

Neglectedness/Development Potential

Neglectedness: Evaluating how many resources are already dedicated to the problem, with the idea that less attention may mean more opportunities for impactful contributions.
Development Potential: Evaluating where targeted research and innovation could unlock significant technical advances and new capabilities in an emerging technology.

The market for high quality sensors is not neglected, but the market for high quality sensors that are accessible to everyone is. The current market is focused on high end scientific instruments, or low cost consumer electronics. There is a gap in the market for high quality sensors that are accessible to everyone.

As noted above, the cost of sensors is decreasing, and the quality is increasing. This is a trend that we expect to continue, and we believe that we are well positioned to take advantage of this trend.

Additionally as noted hardware is outpacing software, and there is a lot of potential for impactful contributions in this space.

We believe that by focusing on open source software, and building a community around our sensors, we can unlock new applications and measurements that are not possible with current technology. [XX]

There are growing consumer interest in understanding their environment, and the things they consume. This is evidenced by the growth in the market for water filters, and the increasing number of people who are interested in environmental monitoring. [XX]

By first focusing on high quality that are well designed and sold as premium products, we can build a brand that is trusted by consumers. We'll use the revenue from these products to fund the development of more advanced sensors which longer term will have an outsized impact.

Plan

Initial products, the long term vision, and the steps we need to take to get there.

Initial Products

First we'll start with a set of well designed (visually and functionally) simple sensors. These are not limited to thermometers, CO2 sensors, light sensors, and precision levels. These sensors will be designed to be used in the home, lab, office or workshop.

These sensors will be high quality, reliable, and accurate. They will be designed to be used by professionals and early adopters, and will be priced accordingly.

The design will be reminiscent of high end consumer electronics, aircraft aluminum, glass, and high quality plastics. Presentation and packaging will be a key part of the product, and will be designed to be given as gifts or displayed in the home.

In addition to the focus on physical design, the early sensors will ship with embedded AI features; just enough to provide a useful experience out of the box. This will include things like automatic calibration, and the ability to detect and alert the user to changes in the environment - or human-like interactions.

The "North Star" and second set of products will are more ambitious, and will take longer to develop. The first of these is a NIR spectrometer with open source software to enable new applications/measurements.

NIR spectrometers are used to analyze the chemical composition of materials. They are used in a wide range of applications, from food and agriculture, to pharmaceuticals and environmental monitoring. [20]

Spectrometers are well understood lab instruments, and many different fields of study rely on spectral analysis to understand their samples. Spectrometers are used in chemistry, physics, biology, and many other fields [21] [22] [23] [24] [25].

Unfortunately, spectrometers are expensive, and are typically only found in research labs. This limits the number of people who have access to this powerful tool, and limits the types of measurements that can be made.

The current market for NIR spectrometers, particularly those used in food and agriculture, is dominated by handheld devices priced at ~$7,000 or more. Beyond the steep upfront cost, these devices are often constrained by proprietary software, locked ecosystems, and mandatory subscriptions for essential features like chemical identification and quantification. [26]

Initial Software

Spectral Fingerprinting Spectral fingerprinting is a method of identifying a substance based on its spectral signature.

Algorithmic reduction of signal to a fixed size fingerprint. This can be achieved using discrete wavelet transforms [27]; a method to decompose a signal into a series of coefficients that can be used to reconstruct the original signal. In our case we can use the coefficients to represent a fixed sized spectral signature of a substance.

Fingerprint indexer using coefficient vector distance of fingerprints. By utilizing recent advancements like Hierarchical Navigable Small Worlds [28] we can enable efficient and fast identification of similar signatures. Additionally, we can build on shoulders of the great work done by Instant Domain Search.

Fingerprint clustering and classification using supervised and unsupervised machine learning. This can be achieved using K-Means [29] and Hierarchical Clustering [30] and more advanced methods like Self-Organizing Maps [31].

Roadmap

Starting today the goal is to first explore the world of available sensors, processors and define the core set of components that will be used in the initial products.

The next step will be to design the initial set of sensors, and to build prototypes. These prototypes will be used to validate the design, and to get feedback from potential customers. At this phase we'll need to spend a large amount of time on the design (hire top talent for the product design).

Once the design is validated, we'll move on to production. This will involve sourcing components, building a supply chain, and setting up manufacturing. This will be a challenging phase, and will require a lot of time and effort.

Once the initial products are in production, we'll move on to the next set of products. These will be more ambitious, and will take longer to develop. The first of these is a NIR spectrometer with open source software to enable new applications/measurements.

The goal is to have the first set of products on the market within 12 months, and the second set of products within 24 months.

The projected investment required to get to the first set of products is very low ~$10,000. This would be used to purchase components, build prototypes, hire help with the design, and set up manufacturing for a small batch release of a single product. Note that much of the electrical engineering and software development will be done by the founder.

The projected investment required to get to the second set of products is higher ~$100,000. This would be used to purchase components, build prototypes, hire help with the design, and set up manufacturing for a small batch release of a single product. At this point additional help would speed up the development process.

Details on components

coming soon (links to most likely components)