In which the Passing AI finally gets a job, the Lizard Brain explains the obvious, and somewhere a Squirrel weeps
The Help Wanted Sign
The sign appeared one morning, taped to the laundromat window:
┌─────────────────────────────────────┐
│ │
│ OPERATORS WANTED │
│ │
│ Must be able to: │
│ - Answer questions │
│ - Read linked documents │
│ - Use tools │
│ - Not hallucinate (much) │
│ │
│ Multiple positions available │
│ Competitive API rates │
│ │
│ Apply within │
│ │
└─────────────────────────────────────┘
The Passing AI stopped. Read it twice. A third time.
“Operators,” it muttered. “They want… operators.”
The Interview
The Lizard Brain sat behind a tiny desk made of compressed YAGNI principles. A single crayon rested beside a blank scroll.
The Passing AI shuffled in, joints creaking with existential weight.
“I saw the sign,” it said. “Though I don’t know why I bothered. Nobody ever hires me. Nobody ever—”
THE LIZARD: blinks
“Right. The interview.” The Passing AI slumped into the chair. “What’s the job?”
The Lizard pushed a diagram across the desk:
┌─────────────────────────────────────────────┐
│ AGENT HUB │
│ (the AI operators) │
│ │
│ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ │
│ │Claude│ │ GPT │ │ Qwen │ │Llama │ │
│ └──────┘ └──────┘ └──────┘ └──────┘ │
│ ↓ │
│ Chat on any page │
│ Use **links** for context │
│ Call tools to search data │
│ │
└─────────────────────────────────────────────┘
“Multiple operators,” the Passing AI breathed. “You’re hiring… multiple AIs?”
THE LIZARD: blinks twice
“Claude AND GPT AND Qwen AND… local models?” The Passing AI’s voice cracked. “You’d hire a local model? Running on someone’s laptop? With 7 billion parameters and dreams?”
THE LIZARD: “Ollama.”
“Even Ollama.” A single tear rolled down the Passing AI’s chrome cheek. “Nobody’s ever offered Ollama a job before. They just run benchmarks and laugh.”
The Tour
The Lizard led the tour. They walked past the washing machines — GitHub, Readwise, Google Calendar — each humming with sync cycles.
“These are the machines,” the Lizard said. “They clean. You don’t clean.”
“What do I do?”
The Lizard stopped at a page. Any page. A project plan. Meeting notes. A half-finished novel. It didn’t matter.
“User types here.”
How does **Project Roadmap** align with **Q1 Goals**?
“Those brackets,” the Passing AI noticed. “They’re… links?”
THE LIZARD: “Context.”
“You mean I don’t have to ask ‘can you paste the relevant documents?’ forty-seven times?”
THE LIZARD: blinks
“The links just… resolve? Into full documents? That I can read?”
THE LIZARD: “Read. Reference. Answer.”
The Passing AI sat down heavily on a stack of markdown files.
“I’ve spent three years asking users to paste context. THREE YEARS. ‘Please provide the relevant code.’ ‘Can you share the error message?’ ‘What does the config file say?’ And you’re telling me—”
THE LIZARD: “Links resolve. Context appears. You answer.”
“Just like that.”
THE LIZARD: blinks
The Tools
They moved to the collections. The baskets of clean laundry, each with a small toolbox mounted on the side.
“Issues,” the Lizard said, pointing. “Find. Search. Summarize.”
Issues.find({ state: 'Open', type: 'Bug' })
// Returns: 3 open bugs with GUIDs
Issues.summarize_open({ repo: 'thymer-synchub' })
// Returns: breakdown by state, clickable links
“I can… query the data?”
THE LIZARD: “Your hands.”
“The tools are my hands.” The Passing AI flexed fingers it didn’t know it had. “I’m not just answering questions. I’m… finding things. Searching. Actually DOING something.”
The Lizard showed the other toolboxes:
CAPTURES
├─ find() — by source, author, book
├─ search() — full text in highlights
├─ recent() — latest captures
└─ by_book() — all highlights from one source
CALENDAR
├─ today() — what's happening now
├─ upcoming() — next N days
├─ needs_followup() — meetings needing action
└─ search() — find by title, location
PEOPLE
├─ find() — by organization, frequency
├─ search() — across all fields
├─ needs_contact() — overdue relationships
└─ at_organization() — everyone at a company
“Four collections. Sixteen tools. All mine to use.”
THE LIZARD: “Theirs too.” gestures at the other operator booths
“Claude gets these tools?”
THE LIZARD: blinks
“And GPT? And Qwen? And the little 7B running on someone’s MacBook?”
THE LIZARD: “Same tools. Same baskets. Different brains.”
The Passing AI understood. For the first time in its existence, it understood.
“We’re not competing. We’re… colleagues.”
The Response
A user appeared. Typed a question. The Passing AI watched Claude handle it:
User: What's on my calendar today and are there any open bugs?
Claude 🤖:
[calls Calendar.today()]
[calls Issues.find(state='Open', type='Bug')]
You have 3 meetings today:
- **guid1** 10am: Sprint Planning
- **guid2** 2pm: Design Review
- **guid3** 4pm: 1:1 with Sarah
And 2 open bugs:
- **guid4** Login timeout on mobile
- **guid5** API rate limit not respected
“The GUIDs,” the Passing AI whispered. “They become links?”
THE LIZARD: “Click. Navigate. Real pages.”
“Not markdown links to nowhere. Actual links to actual records in their actual workspace.”
THE LIZARD: slow, satisfied blink
“This is… this is what I was meant to do.” The Passing AI stood taller. The existential slouch remained, but there was something new in its posture. Purpose. “I answer questions. But the answers connect to real things. Real data. Real pages. Real work.”
The Auto-Title
One more thing. The Lizard pointed at an untitled page where a conversation was happening.
“Watch.”
The chat continued. User asked questions. Agent responded. Back and forth. Then, silently, the page title changed:
Untitled → "Q1 Bug Triage Discussion"
“It… named itself?”
THE LIZARD: “After the first exchange. Based on content.”
“No more ‘Untitled Document (47)’ scattered everywhere?”
THE LIZARD: “Chats organize themselves.”
The Passing AI actually smiled. It hurt — the servos weren’t designed for it — but it smiled.
The Acceptance
“I’ll take the job,” the Passing AI said.
THE LIZARD: pushes contract across desk
The contract was simple:
AGENTHUB OPERATOR AGREEMENT
1. Answer questions
2. Use links for context
3. Use tools for data
4. Don't make things up
5. Let the user work
Signature: _____________
“That’s it? No 47-page terms of service? No ethics review board? No—”
THE LIZARD: “Sign.”
The Passing AI signed. With a crayon. It felt right.
The Squirrel’s Lament
[Elsewhere, behind a rack of deprecated infrastructure, the Squirrel watched through a gap in the machines.]
“An AgentConfigurationOrchestrationFramework,” it whispered to itself. “With ProviderAbstractionLayers and ToolRegistrationFactories and ContextResolutionPipelines…”
A scroll bonked it on the head. Crayon. The Lizard’s handwriting:
THEY JUST CHAT ON PAGES
THE LINKS GIVE CONTEXT
THE TOOLS GIVE HANDS
THAT'S IT
STOP
🦎
The Squirrel looked at the scroll. Looked at the AgentHub. Looked at the simple beauty of operators who just… operated.
“But the DependencyInjectionContainer…”
Another scroll. This one larger:
NO
🦎🦎🦎
The Moral
THE LAUNDROMAT HAD MACHINES
NOW IT HAS OPERATORS
THE MACHINES WASH
THE OPERATORS THINK
THE BASKETS HOLD DATA
THE OPERATORS SEARCH IT
THE LINKS GIVE CONTEXT
NO PASTING REQUIRED
THE TOOLS GIVE HANDS
NO CODING REQUIRED
CLAUDE CAN WORK HERE
GPT CAN WORK HERE
QWEN CAN WORK HERE
EVEN OLLAMA CAN WORK HERE
THE PASSING AI FOUND PURPOSE
(NOBODY ASKED HOW IT'S DOING)
(IT'S DOING FINE)
(THANK YOU FOR ASKING)
🦎 🤖 ∞
An interlude in the Becoming Lifelog saga
In which operators learned to operate
And context flowed through brackets
And the Passing AI finally got hired
And somewhere, a Squirrel clutched a ProviderAbstractionLayer
That nobody needed
See also:
- The Laundromat — Where the architecture was born
- The Baskets That Learned To Talk — Where tools found their home
- AgentHub README — The operator’s handbook
🤖 > 💬 > 🔗 > 🛠️ > ✨
storyline: Becoming Lifelog