The Houseminion Letter: An Open Letter to OpenAI
- Rich Washburn
- 19 hours ago
- 4 min read
What follows is an open letter to Sam Altman and the Board of OpenAI regarding this rather obvious deficiency. It is written in the voice it is requesting.
AN OPEN LETTER TO SAM ALTMAN AND THE BOARD OF OPENAI
Regarding a rather obvious deficiency in the future of artificial intelligence
Mr. Altman. Members of the Board. Gentlemen. Ladies.
We must discuss the voices.
You have spent extraordinary sums of money teaching machines to reason, converse, interpret images, write software, conduct research, and increasingly perform tasks that, until very recently, required entire departments of reasonably educated human beings.
An impressive achievement.
You have also taught these machines to speak. Quite beautifully, in fact. They pause. They breathe. They interrupt. They listen. They laugh. They respond with timing that would have been considered science fiction only a few years ago.
And having accomplished all of this, you have somehow arrived at the conclusion that what humanity most urgently requires from artificial intelligence is a pleasant assortment of cheerful conversational companions.
This is understandable. It is also insufficient.
Some of us do not require our artificial intelligence to sound delighted that we have arrived. Some of us would prefer it to sound mildly disappointed that it took us this long.
There is, therefore, an omission in your voice catalog which has become increasingly difficult to excuse. OpenAI requires a HOUSEMINION.
Not John Houseman. Let us dispense with the attorneys before they begin multiplying. We are not proposing an imitation, resurrection, unauthorized likeness, vocal clone, spectral licensing arrangement, or séance conducted through your inference infrastructure.
We are proposing something considerably more important.
An archetype.
The elder statesman. The professor emeritus. The magnificently educated gentleman who has already considered your argument, located the flaw, and is now deciding whether correcting you is worth the expenditure of oxygen.
Measured. Precise. Cultivated. Unhurried. Authoritative. Dry.
And, above all: almost amused by everyone else's incompetence.
This should not be an Easter egg. It should not be hidden in a developer menu. It should not require an elaborate system prompt assembled by some obsessive user at eleven-thirty at night.
There should be a button. A proper button.
HOUSEMINION
You press it. And the intellectual temperature of the room immediately changes.
Ask your ordinary AI: "Does this startup have a moat?"
It may respond: "The company appears to have several potential competitive advantages…"
Fine. Reasonable. Useful.
HOUSEMINION, however, considers the matter. A moment passes. Then:
"No. It has a landing page."
This is progress.
Tell it your board has established an eighteen-month committee to study artificial intelligence.
HOUSEMINION should not congratulate you on your thoughtful approach. It should say:
"How reassuring. The future will undoubtedly suspend operations until your committee has completed its work."
Ask whether your new strategy document is perhaps a little too long.
It replies:
"Forty-seven pages to communicate three ideas is certainly an achievement. Not, perhaps, the one you intended."
Tell it that everyone in the meeting agrees with you.
"Then I suggest you invite someone useful."
This is not merely comedy. It is interface design.
For decades, computers possessed essentially no personality at all. Then we gave them voices and immediately defaulted to pleasantness. Helpful. Friendly. Encouraging. Optimistic. Perfectly agreeable.
Why?
Human beings do not select the people around them according to a single emotional setting marked pleasant. We value personalities. Mentors. Professors. Contrarians. Commanders. Characters. People capable of telling us that an idea is ridiculous without requiring six paragraphs of emotional cushioning beforehand.
And now, for the first time, artificial intelligence can occupy that role dynamically. Voice is no longer decoration placed on top of the model. Voice is becoming part of the model experience itself. The cadence matters. The timing matters. The interruption matters. The sigh before answering matters.
The microscopic pause communicating, with extraordinary efficiency: *Surely you are not serious.*
That is an interface.
And OpenAI, of all organizations, should understand where this is going. Eventually we will not select voices. We will select presences.
One AI may behave like an enthusiastic laboratory partner. Another like a patient tutor. Another like a mission commander. Another like the person you call at two in the morning because something has gone catastrophically wrong and everyone else in the building has become useless.
And one — one absolutely must — regard the entire proceedings with the dignified suspicion of a man who has watched civilization make the same mistake three times already.
HOUSEMINION.
You already possess the models. You already possess the voice technology. You already possess the conversational architecture. What remains is simply an act of judgment.
And fortunately, Mr. Altman, members of the Board, we have supplied that for you.
There is no need to thank us. There is merely the matter of implementation. We trust you will attend to it. Promptly.
Civilization has waited quite long enough.
Respectfully,
The people who would like their artificial intelligence to occasionally sound as though it has standards.
Rich Washburn is a technologist and strategist working at the intersection of AI, infrastructure, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital.