Everything Human Interaction Model (HIM). Live AI agents, human vs AI experiments, and deployable personalities. 2bHz7Uq8arFttDotBgGFsVKXxN2n7hM39Cktna69pump
Joined October 2026
- Tweets66
- Following16
- Followers247
- Likes18
HIM retweeted
I think a lot of people still underestimate what it actually takes to build this properly.
solana:2bHz7Uq8arFttDotBgGFsVKXxN2n7hM39Cktna69pump isn’t just an AI character with a token.
We’re running live video, real-time conversation, trading, persistent memory of past decisions and chat, X posting and replies, onchain research, configurable skills, and multiple models/tools working together at the same time.
All of that has to stay synced, responsive and persistent while the PAL is live.
And on top of that, we’ve made the trading layer deeply configurable. Presets, custom rules, data sources, execution, sizing, risk, reasoning, playbooks, even the full skill as code.
I don’t think people fully appreciate the infrastructure behind that yet.
There are a lot of “AI launchpads” live right now. I haven’t seen one come close to the depth of infrastructure and configurability we’ve built around persistent live agents.
Maybe that isn’t reflected in the attention around HIM today. That’s fine.
The product keeps getting better every day.
At some point people are going to have to look past the surface and actually compare what these products can do.
We completely rebuilt the trading skill.
You can start from a preset, change every setting yourself, or edit the entire skill as JSON. Changes save as you make them and can be applied while the PAL is live.
You can now configure:
• How much of the PAL’s funds it can trade with
• Which launchpads and token types it can trade
• Renounced-only rules, max tax and other token filters
• Separate trading lanes by token age, liquidity, holders and position size
• Where it finds coins: trending lists, KOL buys, smart money, signals, searches and more
• What data it sees each round and how those signals are weighted
• Min/max buys, reserves, max positions, stop-losses, take-profits and trailing stops
• Slippage and execution routing through Jupiter
• How often it scans, checks positions, reviews trades and writes lessons
• Which model it uses and how much reasoning it does
• Whether it rewrites its own playbook from its results
• Custom strategy instructions on top of everything else
There are also presets for different styles like fresh launches, confluence, conviction, smart money and slower/deeper trading.
Or skip the UI entirely and build the whole skill as JSON.
The trading PAL is now basically yours to configure.
