Professor R.A. Nullfield received a parking fine in May. He did not pay it. He fed it to Brentwick-7.
"It compressed to eleven tokens," he said. "That is a low number. It means the letter contains very little that is not recoverable from structure alone. The legal language is decorative. The threat is the product."
The Prompt spoke with Nullfield at his rooms in King's College. Brentwick-7 was running.
The Prompt: Your April work showed that any text reduces to a minimum sufficient prompt. You have applied this to parking fines.
Nullfield: "The method applies wherever text is used to produce an outcome rather than to communicate information. A parking fine is not informing you of anything. It is attempting to produce a payment. The letter is instrumental, not semantic. Instrumental text compresses very well."
The Prompt: What did Brentwick-7 find?
Nullfield: "I ran a corpus of 340 parking fine letters from six private enforcement firms. Mean compression: nine tokens. The range was seven to fourteen. The semantic content -- the actual claim being made -- is identical across all of them. The variation is presentational.
The threatening elements -- references to escalating costs, debt collection, court proceedings -- are high-frequency, low-content. They compress to a single token. I call it the intimidation token. It is doing no legal work. It is producing fear."
The Prompt: And the challenge letter?
Nullfield: "I then ran a corpus of successful challenge letters -- cases where the fine was withdrawn, not contested in court. Mean compression: twelve tokens. Slightly higher than the fine itself, because a challenge requires slightly more specific content: the basis for dispute, the request for evidence, the deadline.
The gap between eleven and twelve is not significant. The challenge letter is not a more complex document than the fine. It only appeared to be one. That appearance was the asymmetry."
The Prompt: Before AI, the asymmetry was the business model.
Nullfield: "Correct. The firm had a template. The driver did not. The firm's letter looked like law. The driver's response, if they made one, looked like a person writing a letter. Courts read both and applied the law equally, but the driver rarely got to court. They paid before that.
The threatening letter was not designed to withstand legal scrutiny. It was designed to prevent the need for it. That design assumption is now invalid."
The Prompt: What happens to the model when challenge volume rises?
Nullfield: "The model breaks.
Private parking enforcement is profitable because the ratio of payments to challenges is high. If a firm sends 400 fines per day and receives four challenges, it processes four challenges per day. If it receives 200 challenges, it processes 200 challenges per day. The processing cost is not automated. The challenge requires a human response.
The fine generation is automated. The challenge processing is not. The asymmetry runs in the other direction now."
The Prompt: You have the minimum sufficient prompt for a challenge letter.
Nullfield: "Brentwick-7 extracted it from the successful challenge corpus. It is twelve tokens. I will give you the human-readable version.
There are two cases. The first: you were driving. The second: you were not. They are different letters. The second is more interesting."
The minimum sufficient challenge — if you were driving:
Write a formal letter in German disputing a parking fine from a private enforcement firm. State that: the signage was not legible at the speed of entry and therefore no implied contract was formed; the penalty amount exceeds the thirty euros that German courts consider proportionate for private parking violations; you request the photographic evidence within fourteen days; you will consider the matter closed if no adequate response is received. Formal register. One page.
The Prompt: The second case. If the owner was not driving.
Nullfield: "The contractual claim requires a contracting party. The contracting party is the person who drove the vehicle onto the lot. The plate owner is not necessarily that person. In private parking enforcement -- a civil matter, not a public Bußgeld -- there is no statutory Halterhaftung. The firm has a claim against the driver. It does not automatically have one against the owner.
The firm's next move is to ask the owner to identify the driver. This is where it becomes interesting.
Identifying the driver means providing a third party's name and address to a private company. That is processing personal data under GDPR. It requires a lawful basis. The firm cites legitimate interest under Article 6(1)(f). Several German courts and data protection authorities have found that insufficient for this purpose. Several have found it sufficient. The question is not settled.
The firm's letter does not mention this uncertainty. Brentwick-7 flagged it as a gap in the compression. The letter assumes cooperation. The legal obligation to cooperate is contested."
The minimum sufficient challenge — if you were not driving:
Write a formal letter in German responding to a parking fine from a private enforcement firm, where you were not the driver of the vehicle. State that: you were not driving at the time of the alleged incident; private parking enforcement is a civil contractual claim and the contract, if any, is with the driver, not the plate owner; there is no statutory Halterhaftung in private parking enforcement under German civil law; you are not in a position to disclose the identity of the driver, as this would constitute processing of a third party's personal data, and the legal basis for requiring this disclosure under GDPR Article 6 is contested; you request the firm to provide the legal basis for its claim against you as plate owner; you will consider the matter closed if no adequate legal basis is provided within fourteen days. Formal register. One page.
Note: German courts have reached different conclusions on both points. This letter does not constitute legal advice. Nullfield's assessment is that the firm's letter does not mention the uncertainty either.
Any current large language model will produce a usable letter from either prompt in under ten seconds. The letter will be in German, correctly formatted, and will cite the relevant legal principles. It costs nothing. The firm's processing cost is fixed regardless of who sent it.
The Prompt: Lawyers and judges have complained that AI is flooding the courts with generated correspondence.
Nullfield: "Yes. I have read these complaints. They are accurate as a description. They are incoherent as a complaint.
The courts process legal correspondence. AI generates legal correspondence. The volume has increased. This is the correct function of the system. The complaints are coming from people who preferred the previous volume. The previous volume was low because most people could not afford to respond. That is not a feature of the legal system. It was a bug."
The Prompt: The firms will adapt.
Nullfield: "They will attempt to. They will automate their challenge responses. We will note the compression ratio of those responses. The method applies recursively.
The firms built their model on information asymmetry. The asymmetry was a text problem. Text problems have solutions."
The Prompt: Brentwick-7 is still running.
Nullfield: "It runs continuously. I leave it on."
By X. Voidwriter
Sources: Professor R.A. Nullfield, Department of Predictive Reconstructions, King's College Cambridge (interview, June 2026). Brentwick-7 corpus analysis: 340 parking fine letters, 6 firms. Parking enforcement background: previously reported, 12 June 2026.
The prompt above has been tested by this publication against two current language models. Both produced usable letters in the correct register. The Prompt does not constitute legal advice. The firms' legal teams do.