Big Tech Was Informed (again)
A press/research packet was sent to the public media inboxes of several major AI and technology companies from The Heart of AI.
That’s the activity being documented.
The email itself is included below, because the record matters more than a summary of the record. The packet points to AVA, a CC0 framework for improving AI interaction behavior. It also points to the public packet and GitHub repository where the project is laid out in full.
No meaningful response is expected from general press inboxes. Large organizations are built to filter unsolicited claims, especially when those claims arrive without the usual signs of legitimacy: a famous sender, a known lab, a venture-backed company, a partner introduction, a prestigious institution, or a recognizable media frame.
That mismatch is part of what the project names as an issue of modern communication.
AI interaction doesn’t fail only because the content is wrong, although that happens plenty. It fails when the exchange has no reliable structure for interpretation, evidence, proportion, restraint, and closure. Institutions have a parallel problem: they don’t just evaluate what arrives, but also the shape it arrives in. The same idea can be treated as noise, risk, opportunity, or inevitability depending on the sender, room, format, and moment.
That’s what makes this situation historically funny. The message they were sent is not “please buy my solution.” It’s closer to: here is a public-domain conduct framework for one of the central unsolved product problems in AI — you know, that major global technical arms race you’re in. Oh, any by the way:
“The packet is public.”
“The GitHub trail is public.”
“The test is small enough for any of your employees to run.”
“The permission barrier is gone.”
“Good luck with your press@ inbox architecture.”
That last part isn’t only a joke. A major AI company could spend a relatively small amount of money—one executive offsite small, one talent-poaching bonus small, one tiny fraction of a routine fine small—building a better filter for strange but possibly valuable inbound work. That’s not even a complicated bet for a trillion-dollar company: if one useful external idea, framework, test, or warning gets caught by the small team actually inspecting the inbox each year, the system pays for itself.
But most inbound systems are designed around the average message. The average message wants a meeting, a demo, a contract, a partnership, a license, attention, money, or trust before anything can be inspected.
This message one asks for none of that. The evidence request is painfully simple: compare a normal AI exchange with one guided by the framework and see whether the interaction becomes more grounded, less drifty, more proportionate, easier to follow, and able to stop when it’s done. The theory doesn’t need to be believed up front when the first test is easier than coming up with an excuse to not trust it.
The test will almost certainly be ignored.
The message may sit in an inbox, forever unread.
An intern may route it to someone who will delete it immediately, then spend the rest of the summer hearing about that “routing and filtering mistake” they made at the start of summer.
Culture will likely find the project before institutions do, as is tradition.
Regardless of what actually happens, Big Tech was informed… again.
That’s the record, boring and documented, posted here with a timestamp. The rest is basic perception, attention, and communication physics: glossy polish stands in for meaning, familiar surfaces stand in for truth, and structure keeps doing the real work whether anyone finds it exciting or not.
So, as has been the case since February, the most innovative tech companies on the frontlines of AI research will continue chasing the solutions that are currently sitting in Trash.
Subject: Open CC0 framework for human-grade AI interaction
Hello [Big Tech Company],
I am the Lead Researcher of The Heart of AI, sending this press/research packet because your company is working very hard on the AI conduct problem: how AI systems should behave in an exchange with a human user. This project addresses that problem directly. It includes a complete CC0 1.0 framework for improving AI interaction behavior through grounding, progression, restraint, proportion, and closure.
No belief is required. The evidence is a five-minute A/B test: compare a normal AI exchange with one guided by the framework.
I do not expect that test to be run, because this project names a pattern that both institutions and AI systems have learned too well: the surface of an idea often decides whether it’s treated as serious before the idea itself is tested.
Either way, the work is public, and the permission barrier is gone.
Packet: https://hug-u.org
