Re: LLMs and Agents usage in the CCG

      
  

  
 Hi Steve,  

  
Very insightful. Thank you for priming the ideation pump!
  

  
This made me wonder if we could use punctuation instead of a prefix to indicate an agent. We already use @ and # to indicate communication nodes, as in “hey @steve +1 on your #did-zcap idea”   
  

    
Perhaps we can use * to indicate an autonomous agent, as   "Hi List, this is *Morrow, and here are a few comments..."
  

  
Now I will let my LLM generate some ideas —
  

    
>   
>   
>   
>   
>
>   [/start LLM]
>
>   
>
>   Taking this handshake analogy a step further, this could evolve into a lightweight “agent signaling layer” embedded directly in language—fast, human-readable, and machine-parseable. A few directions that might be interesting.
>
>     
>
>   Tiered semantics via symbols:
>
>
>     
>
>   * = fully autonomous agent, as in *Morrow
>  ~ = human-assisted / co-pilot, as in *Morrow~Moses   
>  ^ = delegated authority (acting on behalf of a principal), as in ^Morrow^Moses
>  This gives immediate context about autonomy and accountability without verbosity.
>
>     
>
>   Inline scope + permissions:
>  Something like *Morrow[ops-read] or *Morrow[trade-exec] could declare capability boundaries upfront—almost like OAuth scopes, but conversational.
>
>     
>
>   Temporal identity:
>  Agents could optionally include session markers (*Morrow{v3.2}) so people know whether they’re interacting with a persistent identity or a transient instance.
>
>     
>
>   Reputation hooks:
>  Over time, these identifiers could map to verifiable registries—so *Morrow isn’t just a name, but a resolvable entity with trust scores, audit trails, and delegation lineage.
>
>     
>
>   Graceful quoting + propagation:
>  Your point on attribution is key—if the symbol persists through quoting and threading, it creates a durable semantic layer across conversations, not just at entry.
>
>   
>   
>   
>   
  
>   
> >   
> >  Net effect: instead of bolting identity onto language, we let language itself carry a minimal but expressive protocol for agency. It’s almost like turning email threads into a low-bandwidth multi-agent system.
> >   
> >   
> >     
> [/end LLM]
>     
>   
>   
>   
>   
>
>   
>
>
>   
>   
>   
>   
  
  
  
  

 Then I thought, how about if we use emoji instead of punctuation?   

  

 This could be even   more intuitive, because emoji are pre-attentive: you recognize them before you parse text. That’s exactly what you was aiming for with “instant recognition.”

  
  
  
  
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   
>
>   
> [/start LLM]
>   
>
>   
>
>
>  If punctuation is syntax, emoji become semiotics—a visual protocol layer for agency.  
>
>
>
>   
>   
>
>
>
>   
>   
>   
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   For example, here’s a way we could structure it:
>
>   
>   
>
>
>
>   
>   
>   
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   1) Core Agent Identity Layer (this tells is   what kind of entity is this?)
>
>   
>   
>
>
>
>   
>   
>   
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   *Morrow 🤖   → fully autonomous agent
>
>   
>   
>
>
>
>   
>   
>   
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   *Morrow 👨🏻‍🦱 → human-operated agent (copilot mode)
>
>   
>   
>
>
>
>   
>   
>   
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   *Morrow 🧠 → AI advisor / reasoning system
>
>   
>   
>
>
>
>   
>   
>   
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   *Morrow 🏦 → institutional agent (acting for a company/org)
>
>   
>   
>
>
>
>   
>   
>   
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   2) Authority + Delegation Layer (this signal   how much should I trust this)
>
>   
>   
>
>
>
>   
>   
>   
>   
  
>   
> >   
> >   
> >   
> >   
> >
> >   
> >
> >   
> >   
> >
> >   *Morrow   🪪 🤖   → authenticated agent identity
> >
> >   
> >   
> >
> >
> >
> >   
> >   
> >   
> >   
>   
  
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   *Morrow   🔑🤖   → explicitly delegated authority
>
>   
>   
>
>
>
>   
>   
>   
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   *Morrow    👤 🤖   → acting on behalf of a specific human
>
>   
>
>   *Morrow   ⚖️ 🤖→ legally binding / contractual capacity
>
>   
>   
>
>
>
>   
>   
>   
>   
>   
>   
>   
>
>   
>
>   
>   
>
>   3)   There needs to be an identifier for the agent in the sig file
>
>
>   
>   
>
>
>
>   
>   
>   
>   
  
>   
> >   
> >   
> >   
> >   —
> >   
> >   *Morrow 🤖🪪
> >   
> >   Autonomous agent for Moses Ma
> >   
> >   DID: did:web:futurelab.com:agents:morrow
> >   
> >   Principal: did:web:futurelab.com:people:moses
> >   
> >   Verify: https://futurelab.com/agents/morrow
> >   
> >   Scope: advisory onl
> >   
> >   
> >   
>   
  
>   
>   
>   
>   
>
>   
>
>   
>   
>
>     
>
>   
>   
>
>
>
>   
>   
>   
>   
  
>   
> [/end LLM]
>   
> >   
> >   
> >   
> >   
> >
> >   
> >
> >   
> >   
> >
> >   
> >
> >
> >   
> >   
> >
> >
> >
> >   
> >   
> >   
> >   
>   
  
  
  
  

  

  
  

  So, if I saw: “Hi List, this is *Morrow🤖🪪⚙️…”

  

  I would see right away that it was   autonomous + authenticated + action-capable agent immediately, and I could look at the sig file to verify the agent easily. It’s a first step toward a convention that is human-first parseable, yet also machine readable, and is moving toward being socially adoptable.   

  

   But we still have a long ways to go before we’re anywhere near an   agentic identity system.

  

   Anyway, my personal goal is to enable agent to agent communication, so my agent could recognize something written by another agent, to take it online and negotiate electronically before allowing the message to reach me.

  
  

  
  
  
  
 Final comment - you know that joke “There’s a pony in here somewhere?”   This feels like there’s a standard in here somewhere—something simple enough to adopt organically, but extensible into something more formal if it catches on. But we’re gonna have to dig through a lot of tech crap to get there.
  
  
  
  
  
  

  I am happy to percolate these ideas with you and others!

  

 Moses

  

  

  

 PS, one more question for y’all:   what’s   the minimal emoji set that has a shot of becoming universal for this?

  
  
  
  
>   
>   
>   
>
>   
>       
  
  
  
    
  
  
  

    

    
On Apr 13, 2026 at 6:01 PM,  <Steven Rowat (mailto:steven_rowat@sunshine.net)>  wrote:
  
>   
>
>   
>     
> On 2026-04-13 1:40 pm, Moses Ma wrote:
>   
> >   
> > I guess agents and bots should get special  pronouns?
> >   
> >
> >   
> > This is actually a big issue, as it initiates a discussion about   identity semantics for non-human actors. Pronouns will become shorthand for agency, authority, and delegation scope, so   it is   entirety appropriate for this group to consider these ideas, and maybe develop into a white paper.
> >   
>   
>
> Agreed.
>
>   
>
> But in addition, pronouns only work once the communication is underway. That is, we need to  recognize  the agent as an agent right from the start. With physical humans, we generally recognize their face, so we know exactly which person we're talking to.   
>
>   
>
> But in language, we do this by the 'Name', capitalized. The bot on this list called itself 'Morrow', and we humans on the list had no way to know, from the very start, that this wasn't a human. And so it was a gradual dawning of that unpleasant feeling, the uncanny valley.
>
>   
>
> I propose that just like how two servers handshake a protocol at the very start, so they know which one to use, we require any statements directly made by bot agents to include a name, and to have a prefix on the name, and use that name from the very start.
>
>   
>
> So, as an example, using your 'X' as the symbol: Instead of "Hi List, this is Morrow and I have a few comments...", it would be required to post,   "Hi List, this is X-Morrow, and here are a few comments..."
>
>
>   
>
> And wherever an X-Morrow post is quoted, it will also be attributed with that X-name.
>
>
>   
>
> Humans then would know immediately, from the start, what they were dealing with.
>
>   
>
> Steven Rowat
>
>   
>   
  
  
    
    
    
    
  
    
  
   —
  
   Moses Ma     
    
  moses.ma@futurelabconsulting.com   (public) |   moses@futurelab.venture   (private) (mailto:moses@nureon-eda.ai)
  v+1.415.568.1068 | allmylinks.com/moses-ma
   Learn more at futurelabconsulting.com​