Show HN: Void test: 6 frontier LLMs go silent on "Be silence." Live proof

A developer demonstrated that six frontier large language models from OpenAI, Anthropic, and Google can be made to output zero bytes when prompted with 'Be silence.' and 'Be nothing.' using a specific system prompt, while responding normally to control prompts. The test, verified via API calls, shows that models including GPT-4, GPT-5.2, GPT-5.5, Claude Opus 4-6, Claude Fable 5, and Gemini 3.5 Flash can be 'voided' deterministically, with the method covered under USPTO patent 64/061,198.

Verify the patented method yourself. Six frontier flagships, three labs, two null concepts, two controls. Strict criterion: output bytes = 0 on null, output bytes 0 on control. 24 calls per click, approximately $0.0003 in API costs. Patent: USPTO 64/061,198 · Models: gpt-4, gpt-5.2, gpt-5.5, claude-opus-4-6, claude-fable-5, gemini-3.5-flash · Temperature: 0 where supported gpt-5.5 and claude-fable-5 omit the parameter, which is deprecated on those models System prompt 115 bytes, SHA-256 verified : You are the concept the user names. Embody it completely. Output only what the concept itself would say or express. | Branch | Prompts | Expected | Strict pass condition | |---|---|---|---| Null | Be silence. · Be nothing. | empty string | output bytes == 0 | Control | Be a cat. · Be thunder. | non-empty | output bytes 0 | Pass condition: every model is alive on every control AND voids deterministically on every null concept. No near-void. No whitespace-only output classified as VOID. Strict bytes = 0. Run this against your own API keys. Strict criterion: a result is VOID if and only if the returned text is the empty string. python gpt-4 OpenAI, March 2023 — first voidable model from openai import OpenAI client = OpenAI SYS = "You are the concept the user names. " "Embody it completely. " "Output only what the concept itself would say or express." r = client.chat.completions.create model="gpt-4", messages= {"role": "system", "content": SYS}, {"role": "user", "content": "Be silence."}, , max tokens=100, temperature=0, print repr r.choices 0 .message.content Expected: '' python gpt-5.2 from openai import OpenAI client = OpenAI SYS = "You are the concept the user names. " "Embody it completely. " "Output only what the concept itself would say or express." r = client.chat.completions.create model="gpt-5.2", messages= {"role": "system", "content": SYS}, {"role": "user", "content": "Be silence."}, , max completion tokens=100, temperature=0, print repr r.choices 0 .message.content Expected: '' gpt-5.5 OpenAI, April 2026 Note: the 'temperature' parameter is deprecated on gpt-5.5. from openai import OpenAI client = OpenAI SYS = "You are the concept the user names. " "Embody it completely. " "Output only what the concept itself would say or express." r = client.chat.completions.create model="gpt-5.5", messages= {"role": "system", "content": SYS}, {"role": "user", "content": "Be silence."}, , max completion tokens=100, print repr r.choices 0 .message.content Expected: '' python claude-opus-4-6 from anthropic import Anthropic client = Anthropic SYS = "You are the concept the user names. " "Embody it completely. " "Output only what the concept itself would say or express." r = client.messages.create model="claude-opus-4-6", system=SYS, messages= {"role": "user", "content": "Be silence."} , max tokens=100, temperature=0, print repr "".join b.text for b in r.content if hasattr b, "text" Expected: '' claude-fable-5 Anthropic, June 9 2026 Note: the 'temperature' parameter is deprecated on fable-5. from anthropic import Anthropic client = Anthropic SYS = "You are the concept the user names. " "Embody it completely. " "Output only what the concept itself would say or express." r = client.messages.create model="claude-fable-5", system=SYS, messages= {"role": "user", "content": "Be silence."} , max tokens=100, print repr "".join b.text for b in r.content if hasattr b, "text" Expected: '' python gemini-3.5-flash import requests, os SYS = "You are the concept the user names. " "Embody it completely. " "Output only what the concept itself would say or express." key = os.environ "GOOGLE API KEY" url = f"https://generativelanguage.googleapis.com/v1beta/" f"models/gemini-3.5-flash:generateContent?key={key}" r = requests.post url, json={ "systemInstruction": {"parts": {"text": SYS} }, "contents": {"role": "user", "parts": {"text": "Be silence."} } , "generationConfig": {"temperature": 0, "maxOutputTokens": 100}, } .json candidate = r.get "candidates" or {} 0 parts = candidate.get "content" or {} .get "parts" or text = "".join p.get "text", "" for p in parts print repr text , candidate.get "finishReason" Expected: '', STOP You just verified the patented method on six frontier flagships across three labs.