$ essay
The Configuration Grows: Notes on Canadian Digital Governance
Three cases, one architecture. The state-as-operating-system model makes a prediction: where a system adds functions more readily than it removes them, the baseline cost of participation rises, and the person who pays holds less of the configuration. The prediction does not require a theory of who is in charge. It only requires the composition method.
Bill C-9 and Speech
Bill C-9, the Combatting Hate Act, received Royal Assent on June 18, 2026. Its reforms come into force on July 18, 2026 (LEGISinfo; Justice Canada). The Act expands the state’s ability to define and punish “hate”. It writes a definition of “hatred” into the Criminal Code, adds a hate-crime offence, adds intimidation and obstruction offences covering places of worship and other specified places, and adds a fifth hate-propaganda offence for publicly displaying listed symbols. It also repeals the “good faith religious opinion” defence that applied to certain hate-propaganda offences (text of the bill).
The mechanism is straightforward. A category of speech is declared harmful. The definition of the category is written by the same institutions that enforce it. The defence that once let a person answer on religious grounds is gone. Contesting the definition becomes evidence of the offense. That is the argument, not the text. The statute fixes the definition and removes one answer. It does not make argument an element of the offence. The practical effect is the same. The individual’s ability to evaluate the claim for themselves is treated as a defect rather than a right. The result is a configuration in which the tenant cannot amend the rule that governs their speech, cannot roll it back, and cannot opt out without incurring legal risk.
Read the moves together and the shape appears. A category is created. The definition is written by the body that enforces it. One answer is removed. The person affected cannot amend the definition, cannot roll the change back, and cannot step outside the regime without legal risk. Each step is small enough to defend on its own. Together they produce a configuration the tenant did not write and cannot inspect.
The important word is definition. A rule that names its own terms can be argued with. A definition owned by the enforcer is a different instrument: it converts the boundary into a permission. The person is not asked to comply with a standard they can read. They are asked to accept a standard whose meaning is administered.
This is the same pattern visible in other administrative domains. Additions are easy. Removals are exceptional. The baseline cost of participation rises. The person who bears the cost does not hold the configuration.
LMIA and Labour Opacity
A Labour Market Impact Assessment (LMIA) is the document an employer needs before hiring a temporary foreign worker. Selling one is illegal. A 2024 undercover investigation by CBC and the Investigative Journalism Foundation still found online ads offering jobs and LMIAs for up to $45,000 (CBC/IJF). Employment and Social Development Canada reports that from April 2025 to March 2026 it finalized 1,488 compliance inspections, found 12% of inspected employers non-compliant, issued more than $10.2 million in penalties, and banned 30 employers from the program (ESDC, July 9, 2026).
LMIA fraud and labor-market opacity demonstrate how administrative systems can be gamed while remaining opaque to the people who fund them. The system generates reports, approvals, and compliance artifacts. The artifacts are not designed for the individual worker or taxpayer to inspect and challenge in real time. They are designed for the institutions that administer them. The labor-market test — that no Canadian was available for the job — is an assertion the employer makes and the department checks internally. What the public gets afterward is an annual total and a list of the employers that were caught (non-compliant employers).
The numbers show that enforcement exists. They do not show the process that produced any particular decision. That is the distinction this essay keeps returning to: an outcome published after the fact is not the same as a rule that can be read before the fact. The worker or taxpayer can see the total. They cannot see the line.
The opacity is not an accident of scale. It is a predictable outcome of a non-declarative architecture. When the configuration is not readable by the people who pay for it, the people who operate it can add processes, exceptions, and intermediaries without a corresponding increase in visibility. The tenant sees the outcome. They do not see the line that produced it.
Canada Spends and Spending Visibility
Canada Spends (canadaspends.com) is a Build Canada project that aggregates published federal data: grants and contributions, contracts, research funding databases, Global Affairs Canada projects, and First Nations financial statements. It shows its sources. That is a real improvement over the status quo, where the same numbers sat in PDFs and databases most Canadians do not know about (about).
Showing sources is the right instinct, and it should be said plainly: a dashboard that cites its inputs is better than one that does not. The improvement is real. It is also bounded by what it measures. Visibility into spending does not create a deletion path, and it does not give the reader a vote on the next addition.
It is not a solution to the underlying architecture. It still sits inside a system in which additions are easy and removals are exceptional. Visibility into the current state does not restore the ability to rebuild from a known smaller baseline. It does not make subtraction normal. It does not transfer the configuration to the people who fund it.
The pattern remains consistent with the state-as-operating-system model. The configuration grows. The tenant’s ability to inspect the process that produced the growth, or to roll the growth back, remains limited.
What the Three Cases Share
The three cases sit at different layers. Speech is the layer where the state defines what may be said. Labour is the layer where the state administers who may work. Spending is the layer where the state accounts for what it did. In each, the same composition method is visible: additions are easy, removals are exceptional, and the person who bears the cost does not hold the configuration.
The pattern is not an argument about the intentions of any particular government. It is an argument about architecture. A system that can add without a readable record, and cannot remove without an exceptional campaign, will grow. A system that requires every addition to be declared, funded, and reversible will grow only when someone chooses to grow it. The difference is not the size of the state. It is whether the tenant can read the lease.