This piece has two layers. The story runs top to bottom—read it like any article. After each big idea there's a 🔧 Under the hood box for engineers and builders: exactly how we make it real, what's already shipped, and where you could jump in. Skim the boxes or dive into them. Both layers tell the same story; one just shows the wiring.
Ask almost anyone, point-blank, "What can you offer the world?"—and watch them go quiet.
It's not that they have nothing. It's that the most valuable things they can do are the ones they've stopped noticing. The friend everyone calls when their laptop dies. The person who can calm a room, fix a bike, write a clear email, spot a good deal, explain a hard idea in plain words. Ask them about it and they shrug: "Oh, that's nothing. Anyone can do that."
That shrug is the whole problem. And it's where we think your AI economic agent should start.
The value you can't see
There's a well-studied reason people undervalue themselves. Psychologists call it the curse of expertise: once something becomes easy for you, you assume it's easy for everyone, so you stop seeing it as a skill at all. People consistently discount their most sellable abilities—precisely because they're good at them.
So the most important signal isn't "I'm great at X." It's the opposite. When someone waves a hand and says "it's just a hobby" or "everyone knows that," that's usually where the value is hiding. A good agent shouldn't take "that's nothing" at face value. It should lean in: "You said that like it's nothing. That's exactly where the good stuff usually is."
🔧 Under the hood — why this is a data problem, not a magic problem
Today, when you ask OrangeCat's Cat "what can I offer?", it reads everything it knows about you—your profile, any documents you've added, things it's remembered from past chats, the entities you've already created—bundles that into a prompt, and reasons over it. That part works and is live.
The catch: for most people, Cat knows almost nothing. The profile is essentially two free-text fields (
bioandbackground). There's no structured place for skills, goals, assets, constraints, or the single richest signal of all—"what do people come to you for?" The reasoning engine is fine; it's reasoning over an empty page. The bottleneck isn't intelligence, it's data. Everything below is really one question: how does an agent earn a rich picture of you, honestly, without making you fill out a form?
Don't survey people. Interview them.
The lazy way to learn about someone is a form: name, bio, skills, goals. People hate it, and worse, it doesn't work—"what are you good at?" produces modesty and a blank stare.
What does work is the thing great coaches, recruiters, and interviewers have always done: ask for a story, then find the asset inside it. Not "are you good at teaching?" but "tell me about the last time someone thanked you." Not "list your skills" but "what do people keep coming to you for?", "what do you lose track of time doing?", "what did you pick up faster than everyone around you?"
You answer with a memory. The agent does the valuation. That's the trade that makes it feel like a conversation instead of an interrogation—and it has a quiet bonus: a story about what you did never requires you to reveal who you are. On OrangeCat, you stay as pseudonymous as you want.
So the first design decision behind Cat is simple: stories in, value out. One good question at a time, branching on your answer—never a wall of questions.
🔧 Under the hood — the interview engine
A few principles from elicitation research drive the design (motivational interviewing, appreciative inquiry, the STAR method, laddering / jobs-to-be-done):
- Episode-elicitation, not self-assessment. Cat asks for a time something happened, then extracts the transferable asset from the story. This is also our pseudonymity guarantee—no real-world identity required.
- One dimension per round. A short adaptive Q&A (2–3 questions at a time): skills → assets → goals → constraints. Each answer selects the next-best question. It's a branching tree, not a questionnaire, and Cat narrates why it's asking.
- Know when to stop. Stop when a pattern repeats, not when a checklist completes.
There's a real engineering twist here: Cat's system prompt currently forbids interviewing ("never open with a list of questions"). That rule exists for a good reason—nobody wants an interrogation bot. So the interview isn't "remove the rule"; it's a carve-out gated on how empty your profile is. When Cat already knows you, it stays conversational. When the page is blank and you've opted in, it earns the right to ask. The interview itself is a new tool + handler alongside the existing ones; the hard part—knowing when it's allowed—is a completeness signal, which brings us to the keystone later.
You're richer than you think (in more ways than you think)
Here's the part most people—and most platforms—miss. When we imagine "making money from who I am," we picture two options: get a job, or sell a product. But value comes in many more forms than that. Josh Kaufman's Personal MBA counts a dozen.
Your spare room isn't just yours—it's something you could rent. Your savings aren't just sitting there—they're capital you could lend or invest. The group chat you organize is an audience. The afternoon you'd give to a cause is an option on your time. The thing you learned the hard way is knowledge other people would pay to skip the pain of.
Most people never see these because they file them under "mine," not "offerable."
🔧 Under the hood — the twelve forms of value map onto real database tables
This is OrangeCat's structural advantage, and it's not a metaphor. The platform is built around an entity registry: services, products, projects, causes, research, loans, investments, rentable assets, communities, events, documents. Each "form of value" Kaufman describes has a concrete home here—a table, an API, a public page, a payment path.
So when Cat finds a latent asset in your story, it doesn't hand you advice; it maps the asset to an entity type and drafts the actual listing. That pipeline is already live: a
suggest_offerscapability gathers your context, asks a capable model to map your latent assets across the spectrum, and emits several ready-to-publish draft cards—each one a real service, product, event, or cause, prefilled, one tap from going live. We sweep every form deliberately, so the agent never leaves a value class unexamined. A competitor with "post a gig" can only ever see one of the twelve.
The gap is the gift
Finding what you can already offer is half the job. The other half is the part that actually changes lives: finding what you're almost able to offer, and helping you close the distance.
The trick is to frame it as adjacency, not deficiency. Never "you lack a certification." Always "you're 80% of the way there—here's the missing 20%." You fix neighbors' bikes for free? You're one price and one booking link away from a real service. You spent eight years in tax law and call it boring? Boring to you because you're rare at it—you're one short guide away from something people would fund.
And a gap should never end on advice. Advice without a place to use it goes nowhere. So every gap Cat finds ends in the smallest possible next step and a draft you can publish in two clicks: a service listing, a one-page guide, a single event to test the waters. The same place that spots the gap hosts the thing that fills it.
When the gap is real learning, the prescription is sized to fit: a focused micro-credential measured in hours, not a degree measured in years. When the gap is just confidence, the prescription is an experiment: list one slot, run one event, ship one thing, and let real people—not flattery—tell you if it's wanted.
🔧 Under the hood — adjacency, grounding, and a growth engine
Three mechanisms make "the gap is the gift" concrete:
- Detect by adjacency. Compare the assets surfaced in the interview against the requirements of the next-higher-value entity the user could plausibly create. The output is always "you're X% there," computed from a deterministic economic-completeness score, not vibes.
- Ground in reality, never invent demand. The offer engine already pulls a live snapshot of what's actually active on the platform—how many comparable listings exist, in which categories—and uses it to price realistically and spot genuine gaps ("no translation service exists here yet"). This is real data; the engine is explicitly forbidden from inventing searches, trends, or numbers. (We even had to teach it to never quietly slip into "sats"—Bitcoin amounts are always BTC.)
- A growth nudge type. OrangeCat already has a nudge engine that flags missing pieces. The next step is to feed it the full latent picture—skills, assets, goals, memories—and add a
growthnudge: "you have skill X but no listing," "next step up from one-off services is a productized course." Each ends in a draft, never a dead-end suggestion.
When the next step is to build it
Sometimes the smallest next step is a price and a booking link. But sometimes the gap between you and real value is a thing that has to be built—an app, a tool, a product, a working prototype. For most of history that's where ideas died: you had the insight but not the time, the skill, or the team to ship it.
That's the other half of what we're building, and it has a name: Loki. Its whole pitch is "you tell it what to do, and it does it until it's done." You describe the outcome; a fleet of AI agents does the execution—writing the code, building the thing, shipping it—while you stay in judgment, approving the decisions that matter. The distance from "I have an idea" to "it exists and it's earning" collapses from months to days. Loki is what makes building things of value easy—so the gap Cat finds in you doesn't just become a to-do list, it becomes something that actually gets made.
OrangeCat and Loki are two halves of one machine. As Loki puts it: "a capability layer with no way to pay people is just unpaid labor; a payment layer with nothing to coordinate is just a wallet. Together they are an actual economy." OrangeCat finds the value in you and distributes it; Loki builds it; the whole loop settles in Bitcoin, with no employer and no gatekeeper in between.
🔧 Under the hood — OrangeCat finds & distributes value; Loki builds it
The split is clean: Loki is the capability layer—decide what needs doing, route it to AI workers, verify completion. OrangeCat is the economic layer—identity, the entities that hold value, and payment/settlement over Lightning. A buildable gap surfaced here becomes a job a Loki fleet executes; the finished work flows back into an OrangeCat entity and settles in Bitcoin. Humans stay where humans are irreplaceable—judgment and the irreversible calls—while execution gets delegated. The aim, in Loki's words, is "a world with no jobs—not no work, but no coercion."
The promise: we'll ask first, and we won't lie to you
There's a real failure mode here, and we want to name it. The internet is exhausted by "turn your passion into a side hustle." The moment a hobby gets a price tag, it can inherit deadlines, metrics, and a low hum of pressure. For a lot of people, the right answer to "could this make money?" is "please don't."
So Cat asks permission before it goes looking. "Want me to explore ways this could earn—or keep it just for you?" If you say it's just for you, that's the end of it. And Cat won't cheerlead. An agent that says "honestly, I don't see paying demand for this yet—but here's exactly what would change that" is worth a hundred that hype everything. Not everyone is here for income. Some are here for connection, meaning, structure, community. Cat meets you where you are—it doesn't drag you down a funnel.
🔧 Under the hood — guardrails as code, not vibes
"Be respectful" isn't a value if it only lives in a design doc. These are written into the agent's prompts and verified:
- Opt-in / anti-hustle: Cat asks before exploring monetization and honors "this is just for me."
- Grounded-or-flagged: every suggestion is tied to a specific thing in your context or a real platform signal; speculative ideas are labeled as such.
- No fabricated demand: an explicit rule, and a regression test on our list, because an earlier version of Cat once invented people who didn't exist—we don't intend to repeat that.
- Pseudonymous-safe by construction: the story-based interview needs no legal identity to work.
The keystone: giving your value a place to live
Everything above—the interview, the gap-finder, the growth nudges, even smarter matchmaking later—depends on one missing piece. Right now, anything Cat learns about you melts back into free-flowing text. Tell it "I'm a translator who worked at the UN," and at best that becomes a sentence in its memory. It can't be queried, compared, or built on. The agent can read your story but can't really hold it.
The fix is unglamorous and foundational: a structured place for your latent economic value to live. A queryable record of your skills (with proficiency), the assets you'd rent, the goals you're chasing (earn? fund? learn? connect?), your constraints, and what people come to you for. Cat already has the ways to write to you and the ways to read you back; what's missing is the shelf to put it on.
Build that, and everything else snaps into place—the interview has somewhere to save answers, gap-detection can actually run queries, matchmaking can rank, pricing can compare. It's the difference between an agent that chats about your value and one that understands and acts on it.
🔧 Under the hood — this is the highest-leverage thing to build, and it's a great first contribution
Concretely: a structured economic-profile store (a typed table, surfaced into the context the agent reads). Skills, assets, goals, constraints, audience signals—typed, not prose. It's a focused migration plus a read/write path, and once it exists, the interview tool, the completeness score, the growth nudges, and ranked matchmaking all become small, grounded features instead of hand-wavy ones.
We've written the design down—the methodology, the question taxonomy, the data model, the roadmap of what's small versus structural. If you're an engineer who believes an agent like this should exist, this is the load-bearing wall, and it's the kind of clean, well-scoped problem that's genuinely fun to build. The offer engine is already live; the interview and the store are next. Come build them with us.
Why this matters
OrangeCat exists so anyone—any person, any pseudonym—can participate fully in the economy: earn, fund, lend, invest, coordinate, with Bitcoin or any currency, without gatekeepers. But participation starts with a question most people can't answer alone: what do I actually have to offer?
An agent that interviews you well—that treats "it's nothing" as treasure, pulls value out of your stories, maps you across every form of value, and turns each gap into the smallest next step—turns that blank stare into a list of real, publishable possibilities. That's not a feature bolted onto a marketplace. It's the difference between a tool you have to figure out and an agent that's genuinely on your side. And the things it finds, you no longer have to build alone: OrangeCat surfaces the value and distributes it, Loki builds it, and you get paid in Bitcoin—no boss, no gatekeeper in the loop.
The first pieces are live: ask your Cat "what can I offer?" today and it will read what it knows about you and propose real, ready-to-publish ideas, grounded in what's actually happening on the platform. Next, it learns to ask—to interview you, remember what it learns, and help you grow into the value that's just out of reach.
If you're here to build something: the thing you call "nothing" might be the best thing you've got. If you're here to build the agent: the keystone is waiting. Either way—let's go find it.