The Joule Hypothesis

Tech Lab Report 001: The Joule Hypothesis

Artificial intelligence is usually presented as a capability: what a model can write, reason through, create, analyze, or play. The resource consumed to produce that work is often hidden behind subscriptions, token limits, and provider bills.

BK Tech Labs begins with a different premise: AI capacity should be finite, visible, and measurable. We call the unit of that capacity a Joule.

The Joule Hypothesis is that recording an AI result together with the capacity used to produce it makes the result more useful to evaluate. It lets the lab ask not only what did the system produce?, but also under what conditions, at what resource cost, and was it worth it?

This brief report states the hypothesis, approved accounting baseline, access conditions, and initial test. The reasons for preserving token usage and financial cost while not using either tokens or currency as the common unit are in the detailed report. Energy and cost calculations are in Appendix A.

Research Question

How much useful AI work can a person, an AI system, or a human-AI team produce from a constrained and auditable amount of capacity?

Joules do not measure intelligence, human effort, or the intrinsic value of a result. They measure the AI capacity made available and consumed while producing an observable result.

The Joule

A Joule is the common unit of AI capacity across the Tech Lab ecosystem:

• 1 kJ = 1,000 J
• 1 MJ = 1,000,000 J

The initial policy converts provider-reported AI usage cost into Joules:

$1 of provider-reported AI usage cost
= 1 kWh reference value
= 3.6 MJ of Joule Capacity

The equality between 1 kWh and 3.6 MJ is a physical unit conversion. The connection between $1 of provider cost and 1 kWh is a policy baseline, not a claim that one dollar of AI usage physically consumes one kilowatt-hour of electricity.

The initial debit formula is:

Joules consumed = provider cost in dollars x 3,600,000

The receipt for an experiment preserves the workload and token details, provider prices and actual cost, Joule policy version, and Joules charged. That evidence makes the common balance auditable and lets later reports revise the policy without rewriting historical results.

Hypothesis

Under controlled and documented conditions, recording both an AI result and the Joules consumed to produce it will yield a more useful evaluation than recording the result alone.

The hypothesis makes four predictions:

1. Visibility improves judgment. People make more deliberate choices when capacity and consumption are visible.
2. Constraints improve comparison. Results reveal more when the resource budget is held constant or reported alongside them.
3. One unit connects different activities. Games, conversations, research, and creation can draw from one understandable capacity balance.
4. Capacity can support participation. A real free allowance lets people test the system, while a larger paid allowance supports repeated and deeper work.

Approved Initial Access Conditions

The first experiment uses three capacity conditions. They define the experiment; they are not a claim that every condition is already available in every BK Tech Labs product.

The paid-member condition is a consistent tenfold increase over the free-member condition. Both begin at 60% of maximum capacity and recharge at 4% of maximum capacity per day. A free member receives enough capacity to do real work and understand the system. A paid member can run more trials, sustain longer investigations, and compare more alternatives before waiting for a recharge.

Capacity is the first measurable difference between free and paid membership. It is not the complete meaning of membership. The experiment must determine whether the work and participation enabled by each condition justify its scale.

Initial Method

Every experiment should preserve five parts:

1. Attempt: the question, task, rules, or desired result.
2. Conditions: the application, model, settings, and starting state.
3. Result: the moves, conversation, output, or other observable artifact.
4. Receipt: workload and tokens, provider prices and actual cost, policy version, and Joules charged.
5. Learning: an interpretation kept distinct from the observed result, plus the next question to test.

The first experiments should be deliberately small. Give a guest enough capacity to see the premise work, give a free member enough to produce a useful result, and let a paid member repeat or extend the same work. Hold the task and conditions steady where possible, then compare result quality, variation, and capacity consumed.

BarKade provides the first constrained environment: a classic game preserves the rules, state, decisions, and outcome alongside the resource receipt. ReChat can extend the same method to the development of an idea through conversation. The Tech Lab Brief reports what the experiments reveal.

What Would Disprove It

The hypothesis should be revised or rejected if Joules do not help people understand or manage AI use; if the receipt obscures raw usage or actual cost; if the accounting creates false equivalence; or if larger allowances produce more consumption without better experiments or learning.

The access conditions also fail if 50 kJ cannot demonstrate the premise, 5 MJ cannot produce real value, or 50 MJ does not enable a meaningfully different and economically sustainable class of work.

The purpose of the Tech Lab is not to make unlimited AI feel free. It is to learn what people and machines can accomplish when AI capacity becomes visible, comparable, and open to examination.

Further Detail

• Detailed report — token and financial accounting rationale, literature, experimental-system design, policy economics, expanded predictions, and failure conditions.
• Appendix A: Energy and Cost Calibration — unit conversions, external calibration references, and methodological boundaries.