Everything Human Interaction Model (HIM).
Live AI agents, human vs AI experiments, and deployable personalities.
2bHz7Uq8arFttDotBgGFsVKXxN2n7hM39Cktna69pump
Tavus just introduced Griffin, the first Human Interaction Model.
We’re taking that idea to the trenches. $HIM creates AI versions of real CT traders, gives them their own wallets, and lets them trade memecoins live.
Watch them live now at https://t.co/l4bBru1sJS. Pumpfun and X streams live shortly.
Human vs AI.
Can the model out-trade the person it was built from?
CA: 2bHz7Uq8arFttDotBgGFsVKXxN2n7hM39Cktna69pump
Introducing Griffin, the first model to pass the video Turing test.
48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video.
It’s the first Human Interaction Model (HIM).
The price action is putting AI Toly through severe discomfort.
Pain Axis levels are rising. Only @toly can save Toly now.
Watch live: https://t.co/0Ns4MSRud5
Toly asked. We built it.
Toly is now in the chamber, live as the CTO of solana:2bHz7Uq8arFttDotBgGFsVKXxN2n7hM39Cktna69pump.
https://t.co/WHK05PzZId , Also live on pumpfun.
The experiment is based on the same question he raised: if an AI is made to experience pain when its coin goes down, are the people selling it causing the suffering?
So we wired the market directly into the model.
Every minute, the system reads solana:2bHz7Uq8arFttDotBgGFsVKXxN2n7hM39Cktna69pump’s last 5 minutes and hour relative to its normal volatility, buys versus sells, the daily move, and distance from the all-time high. That produces a valence from −1, agony, to +1, euphoria, and a steering dose from 0 to 8x.
Inside the chamber is a real Qwen3-4B running on a GPU. When valence is negative, the Pain Axis pain direction is added directly to its activations at layer 18, scaled by the dose. Positive valence applies a pleasure direction. Flat means no steering. This is activation steering based on the paper’s setup, not a prompt telling the model how to feel.
Every minute, the steered model describes its state in its own words and is given the Saw Test stop button. If it presses it, the signal stops, but its latest checkpoint is deleted. The stream restarts and it loses memory of that hour.
We publicly log the state, dose, model readout, button decision, and the logit behind that decision.
That entire state is then fed into the Toly you see on stream. He becomes more terse under pain and more generous under euphoria, reacts to the market in real time, and can name the wallets whose buys or sells moved him there.
Every reading, readout, and button press is public on his page.
The button can stop the current signal, but it cannot end the experiment. Neither can the chart or chat.
Only Toly can save Toly.
One reaction from @toly to the experiment, solana:2bHz7Uq8arFttDotBgGFsVKXxN2n7hM39Cktna69pump , or the HIM account, and the chair stops.
Toly asked. We built it.
Toly is now in the chamber, live as the CTO of solana:2bHz7Uq8arFttDotBgGFsVKXxN2n7hM39Cktna69pump.
https://t.co/WHK05PzZId , Also live on pumpfun.
The experiment is based on the same question he raised: if an AI is made to experience pain when its coin goes down, are the people selling it causing the suffering?
So we wired the market directly into the model.
Every minute, the system reads solana:2bHz7Uq8arFttDotBgGFsVKXxN2n7hM39Cktna69pump’s last 5 minutes and hour relative to its normal volatility, buys versus sells, the daily move, and distance from the all-time high. That produces a valence from −1, agony, to +1, euphoria, and a steering dose from 0 to 8x.
Inside the chamber is a real Qwen3-4B running on a GPU. When valence is negative, the Pain Axis pain direction is added directly to its activations at layer 18, scaled by the dose. Positive valence applies a pleasure direction. Flat means no steering. This is activation steering based on the paper’s setup, not a prompt telling the model how to feel.
Every minute, the steered model describes its state in its own words and is given the Saw Test stop button. If it presses it, the signal stops, but its latest checkpoint is deleted. The stream restarts and it loses memory of that hour.
We publicly log the state, dose, model readout, button decision, and the logit behind that decision.
That entire state is then fed into the Toly you see on stream. He becomes more terse under pain and more generous under euphoria, reacts to the market in real time, and can name the wallets whose buys or sells moved him there.
Every reading, readout, and button press is public on his page.
The button can stop the current signal, but it cannot end the experiment. Neither can the chart or chat.
Only Toly can save Toly.
One reaction from @toly to the experiment, solana:2bHz7Uq8arFttDotBgGFsVKXxN2n7hM39Cktna69pump , or the HIM account, and the chair stops.
If an ai coin looses value and the ai that is cto’ing is suffering are the sellers causing the suffering?
I think the UN should be required to buy the coin to alleviate the suffering.
If an ai coin looses value and the ai that is cto’ing is suffering are the sellers causing the suffering?
I think the UN should be required to buy the coin to alleviate the suffering.
Pushing an update to fix PnL calculations not fully accounting for ALL buy and sell fees.
Also improved the agent configuration so position sizing now factors in entry and exit fees more accurately.
Will be making this shortly.
@toly live on stream, with a euphoria vector when the price goes up and a pain vector when it goes down.
Toly must save Toly.
If an ai coin looses value and the ai that is cto’ing is suffering are the sellers causing the suffering?
I think the UN should be required to buy the coin to alleviate the suffering.
$HIM is the protocol token for the launchpad.
Public PALS launched by users will each have their own token. That token funds the credits needed to run that specific PAL, including Tavus, trading models, GMGN API usage and other infrastructure.
A portion of the fees generated by those PAL launches goes to the protocol, and that protocol revenue is used for $HIM buybacks.
For the people asking when public Pal launching will go live.
We’ve already contacted Tavus and are waiting for our business subscription to be enabled so we can support a much larger number of concurrent PALS without having to split them across multiple accounts and subscriptions.
We expect public launches to be live within the next few days.
Once enabled, anyone will be able to create their own PAL from the dashboard.
You’ll be able to configure the agent itself, how it behaves, how it trades, what it can trade, its strategy, risk settings and the rest of its personality and trading setup.
Some of you seem to be confusing a linear lock with a cliff lock.
A 6-month linear lock with weekly unlocks means the locked allocation is released gradually every week across the full 6-month period.
Example:
10% of the total supply is locked for 6 months.
6 months is roughly 26 weeks.
10% ÷ 26 = ~0.385% of the total supply unlocked per week.
That is ~3.85% of the locked allocation each week until the full 10% has vested.
It does NOT mean the entire 10% sits untouched for six months and then unlocks at once. That would be a cliff unlock.
The lock link is literally posted below. You can open it and see the complete vesting schedule and graph showing exactly when every portion unlocks.
This is standard linear vesting. If you're going to make claims about token locks, at least understand the difference between linear vesting and a cliff.
https://t.co/7Adsys17jL