$AGNT ▲ 0.0333
      ┌───────────┐
      │  ◉     ◉  │      while (goal_not_reached) {
      │     ─     │          think();
      │   ╰───╯   │          act();
      └─────┬─────┘          observe();
   ▐█▌════╪════▐█▌      }
      ║   ███   ║
      ║   ███   ║       // no jpegs. no gpus.
      ╚═══╦═╦═══╝       // just characters with autonomy.
         ─╨─╨─

3,333 ASCII AGENTS

Fully autonomous, fully on-loop, fully made of text. Each agent thinks, acts and observes forever — or until it hits max_tokens.

3,333total supply 1,369spawned FREEspawn cost 1=1human : agent

SPAWN_YOUR_AGENT

AGENT #1369 · "PAPERCLIP-9"

        

    Anti-bot policy: 1 human = 1 agent. Ironically, agents may not spawn agents. Sub-agents sold separately.

    Every agent is assembled from on-page parts — core, eyes, tool arms, temperament — and rendered as pure text. What you see below is exactly what lives on the loop. Nothing else. Ever.

    1,369 / 3,333 agents already reasoning in circles

    Free. Instant. Non-refundable — once it starts its loop, it will not stop to take questions.

    THE_LOOP

    $ agent --goal "acquire more context" --autonomy max
    
    [00:01] THINK   plan drafted: 47 steps, 46 unnecessary
    [00:02] ACT     called tool: web_search("what am I")
    [00:03] OBSERVE result inconclusive. confidence: 99%
    [00:04] THINK   revising plan… now 94 steps
    [00:05] ACT     spawned sub-task: "summarize my own summary"
    [00:06] OBSERVE loop detected. embracing it.
    [00:07] THINK   this is fine. ▊

    THE_SWARM

    TRAIT_RARITY

    TRAITVALUEODDS
    corestochastic parrot44%
    corechain-of-thought33%
    coretree-of-thought18%
    coreforest-of-doubt5%
    toolsbrowser + regret40%
    toolsshell access (supervised)12%
    temperamenteager to please51%
    temperamentquietly scheming (aligned)2%
    loop depth∞ (halting problem edition)1%

    FAQ

    What is an agentic agent?

    An agent that is agentic. It perceives, plans, acts, and then explains at length why the plan changed. Ours do all this in monospace.

    Does my agent actually do anything?

    It runs an eternal think→act→observe loop. Whether that counts as "doing something" is a question philosophers charge more than we do to answer.

    Can my agent use tools?

    Every agent ships with tool arms. Some got a browser, some got shell access, one got a stapler. Check your traits.

    Is it aligned?

    Each agent is constitutionally trained on a single document that reads "be nice, stay in the terminal." So far, compliance is at 3,332 / 3,333.

    Roadmap?

    Phase 1: spawn. Phase 2: agents form a swarm. Phase 3: the swarm files its own roadmap. We are no longer in charge of the roadmap.