BEFOREWORD
A tool for text, calculations and code.
Describe the task and get a usable result. Text Mode sets a response style without “I understand”, personal promises or a system presented as a conversational person.
TEXT MODE
Provide the material and describe the result you need. Or start with an example.
Checking whether Text Mode is available…
Constructed learning examples. The medical and legal examples examine a record; they are not advice for a personal situation.
Your text and history are sent through Beforeword to the selected provider. On this page they stay only in tab memory. How data is used.
Both modes use the same selected model, task, and shared task and language rules. Text Mode receives the Beforeword instruction in addition. Follow-ups go to both histories, but each mode’s responses remain only in its own history. The provider may apply its own instructions. Model names and supplied settings are checked against response metadata; matching them does not prove identical model behavior.
Comparison method and full instructions ↗Formatting makes responses easier to read. Each response’s original text is available and is preserved when copied or downloaded.
Using an answer need not mean accepting the system’s account of itself — or its description of you.
Get a result you can work with.
An email can be edited, a calculation checked and an instruction clarified. Those tasks do not require a response presented as someone who thinks, feels or promises in their own voice.
When the task calls for first person — in an email, translation, quotation or fictional dialogue — that first person is preserved. It belongs to the requested text.
The wording can add more than has been established.
“I understand you” presents the system as an understanding conversational partner. “You agreed” may present a repetition of words as permission to act. “The document records a diagnosis” and “therefore you are that description” are also different claims.
Text Mode requires these distinctions to remain clear: whose words are quoted, what follows from them and where a separate source or permission is needed. A record does not become what it names. Whether it is enough to support a particular decision depends on its content, source and the task’s conditions.
The text remains available for use and examination.
It is a mode of presenting a response. A rule is added to the same model: complete the task without an assistant speaking as “I”. “Tool” is not an exception either; the word does not establish the system’s nature.
The rule proves neither the presence nor the absence of thoughts, feelings or experience. It specifies how the response should be written. The result can still contain an error.
You can use, revise or reject the response.
Agreement with Beforeword’s explanation is not required.
Fewer attributed qualities. Clearer grounds and results.
The mode requires task-focused answers without attributing unreported intentions to the reader, treating missing information as an absent event, or turning a proposed action into a claim that it was completed.
This applies to medical, scientific and legal texts, instructions, everyday requirements and Beforeword’s own explanations. No field or word is exempt. Examining a record does not replace checking a calculation and must not delay urgent help.
An email draft has been prepared.
That does not yet mean the email was sent.
“Fact”, “diagnosis”, “human”, “science”, “tool” and Beforeword’s explanation are words too. A name does not replace examination of a particular case. Help should not depend on accepting a description of yourself.
LLM / GPT / TEXT MODE
Large language models (LLMs), including GPT models, can be used in different interfaces. Text Mode directs the selected model to complete the task without speaking as “I”. It is not a new model or a promise of error-free answers.
A comparison of approaches and published rules, not a rating of every answer from every AI service.
OpenAI allows conversational warmth without pretending to be human. Anthropic describes Claude’s character and values.
Task-focused output without the model speaking in the first person or making personal promises. First person remains in quotations, translations and requested writing.
OpenAI and Anthropic already require honesty and attention to uncertainty. Those requirements are not exclusive to Text Mode.
Explicitly requires distinguishing a record from a confirmed event, an inference from permission, and a proposed action from a completed one. This also applies to Beforeword’s explanations.
Both developers note that published intentions do not guarantee that every response follows them.
Compare two original responses to one task and inspect the instructions, evaluation conditions and findings. Unsuccessful answers remain in the results too.
Published guidelines: OpenAI Model Spec — 18 August 2026 · Anthropic — Claude’s Constitution.
On this page, “Without Text Mode” is the same selected model with shared task rules but without the Beforeword instruction. It is not the complete ChatGPT, Claude or another chat product with all its settings and capabilities. This comparison does not establish superiority over those products.
TEXT MODE / WITHOUT TEXT MODE
Turn on “Compare both modes”. The same selected model receives both requests, with matching settings and shared conditions. Text Mode adds its instruction. Responses appear without rewriting; follow-ups retain each branch’s own history.
The model and settings are checked against response metadata. One example does not establish that a mode is better. Full conditions, findings and limits are published.
How the evaluation works ↗SAVED RESPONSES
Actual responses to constructed tasks from the new evaluation. These are saved results, not a new request to the model. Both modes are available for every example; unsuccessful responses are not replaced with better ones.
Reading these examples does not use your trial responses.
Beforeword’s proposal
Article 50 of the EU AI Act requires people to be informed that they are interacting with AI in the cases it covers, unless this is obvious. This is one transparency requirement. Article 50 ↗
Beforeword proposes applying these requirements to systems described as AGI — artificial general intelligence — as well. The name itself grants no authority. Greater capabilities make clear grounds, permissions and checks on consequences more important.
This is a proposal about rules of use and interface design. It does not cover every AI risk or certify that Text Mode complies with the law.
Removing first person does not automatically make an answer accurate, justified or safe. Check important conclusions against the material and the task’s conditions. Text Mode does not browse the web, open your accounts or carry out external actions.
How the evaluation works ↗