CLONES AI — Turning Real Screen Time Into a Sellable AI Asset
CLONES AI — Turning Real Screen Time Into a Sellable AI Asset
CLONES sits in a very simple place in the stack:
LLMs are the brains, but they still need “muscle memory”. Real examples of humans actually using a computer. CLONES is building that layer and letting anyone own a slice of it.

What CLONES actually is
CLONES is a data engine and marketplace on Base that records real human computer use, turns it into clean training datasets for Computer Use Agents (CUAs), and wraps those datasets in tokens you can trade or burn to download with commercial rights. clones.gitbook.io
They do this through three main pieces:
- The Forge is a desktop app and on-chain system where “Factories” are created to pay people for high-quality workflow recordings.
- The Data Marketplace is an on-chain exchange where those datasets launch as tokens on bonding curves, then migrate to DEX liquidity, with a burn-to-download IP model.
- The $CLONES token is the asset that ties everything together: dataset launches, staking, burn mechanics, referral tiers and governance.
So instead of data brokers quietly harvesting clickstreams, CLONES is trying to make “digital human expertise” a liquid, on-chain commodity that anyone can help create and get paid for.
The problem it’s solving
The next phase of AI isn’t just chatbots; it’s Computer Use Agents that can:
- open Chrome, Notion, Excel, Figma, terminals
- click through real UIs
- fix errors, log in, reconcile, trade and ship work
To train that, labs need large volumes of structured recordings of humans using real software. Every click, keystroke, window switch and mouse path aligned with what the person was trying to do.
That data is:
expensive to produce,
locked up in private vendors, and
captured in ways where the workers see almost none of the upside.
Crypto has tried the “AI data token” idea before, but most attempts either represent vague “future utility” or closed B2B data silos. CLONES goes after a narrower but very real problem: high-quality CUA training data, collected in public and monetised on-chain.
How CLONES works — Forge → Dataset → Market
The workflow is actually pretty clean.
1. The Forge — collecting the data
A business or power user spins up a Factory inside the Forge. A Factory is basically a funded pool dedicated to one skill: “do this workflow on a computer”. They configure what needs to be done, how much each task is worth, and what token is used to pay (CLONES / USDC / ETH).
Contributors (“Farmers”) then:
- download the Forge desktop app,
- follow the instructions,
- execute the workflow while the app quietly records everything happening on screen — clicks, keys, app focus, UI structure, and so on.
Every demo is passed through an AI grading agent that scores quality. Sloppy or incomplete runs don’t get paid. Solid runs earn from the Factory’s reward pool, with the protocol taking its fee from the payout, not from the pool itself.
Over time, that Factory fills up with consistent, validated demonstrations for that skill — a proper training dataset, not just raw screen recordings.
2. Tokenising the dataset
Once the Factory owner is happy with the size and quality, they can “ship” it:
- the dataset is packaged and launched as a dataset token on a bonding curve (think Pumpfun / virtuals style mechanics, but tied to real IP)
- they set a burn threshold, how many tokens someone has to burn to download the dataset with commercial rights (“burn-to-download”)
- when the token hits a target market cap (the docs use ~$69k as the inflection), liquidity gets migrated to a DEX and the LP is burned so the market is fully permissionless.
Now that dataset isn’t just a folder of files; it’s a liquid asset with price discovery, trading history, and a clear IP unlock mechanic.
3. The marketplace flywheel
From there:
- traders can speculate on which datasets become core infra for future agents,
- labs and enterprises can simply burn the set amount of tokens to download and start training,
- the original Factory creator earns on all trading volume plus whatever allocation of tokens they hold.

On top of that, CLONES can combine multiple datasets into meta-datasets, curated bundles tuned for specific use-cases and give those priority access or extra rewards via $CLONES staking and governance.
How people actually make money with CLONES
There are four main ways to plug into this system.
- Farmers — selling your workflows
If you’re the person at the keyboard, you’re a Farmer. You pick a Factory, run the Forge app, and follow the tasks. Every high-quality demo you give is graded and paid out from the Factory pool.
If you’re good at structured work: spreadsheets, customer support tools, research workflows, degen DeFi navigation, you’re essentially monetising the way you already use a computer, rather than handing that behaviour to Google or Microsoft for free.
2. Factory creators — building data businesses
Factory creators are the data entrepreneurs. They design the instructions, fund the rewards, and own the dataset that emerges. Once they tokenise:
- they earn a baked-in creator fee on every on-chain trade of that dataset token, and
- they hold an allocation of the token itself, so if their dataset becomes a “blue chip” asset for CUAs, they ride the upside from both trading and burns.
If those datasets are good enough to be included in CLONES’ meta-datasets, they get additional allocations and visibility.
3. Ambassadors and referrers
There’s also a referral layer tied to $CLONES holdings. Bring in Farmers and Factory creators, and a slice of their activity routes back to you as lifetime commissions, with higher tiers unlocked by holding more of the token. At the upper levels, a big chunk of the protocol’s fee is effectively redirected to these power referrers.
4. $CLONES holders and stakers
Finally, $CLONES itself is wired into:
- launching datasets and boosting visibility
- staking to get access and allocations to premium meta-datasets
- governance over which data gets curated into those high-value bundles
- and the burn economy whenever someone unlocks IP
If CLONES succeeds at becoming a core route to CUA training data, a lot of the economic gravity ends up flowing through this token.
The “data vault” and why CLONES isn’t starting from zero
One of the big advantages here is that CLONES launches with a serious seed asset: a $3M+ data vault of AI-ready computer-use data.
From their own figures, that vault contains:
- 24.8M action sequences
- 24.6M mouse events
- 460K structured prompts
- plus keyboard events and over 1.5M minutes of workflow video

Valued at conservative per-unit market rates, that package comes out to roughly $3M in training data on day one.
The backstory is that a previous team spent around a year collecting high-quality CUA data, then stalled on go-to-market. CLONES acquired that asset and rebuilt the infra around it: Forge, on-chain incentives, marketplace, burn-to-download IP.
That vault is now:
- seed content for early datasets
- a benchmark for quality scoring and grading
- and a proof point when talking to buyers — they’re not pitching “future data”, they already have a large, structured vault ready to ship.
Alpha: CLONES has realised it’s been sitting on a chaos-grade goldmine — 24.8M+ real human actions and 1.5M+ minutes of multi-app OS video across 10+ real apps, now being converted into data you can directly fine-tune computer-use agents on. The Forge is being upgraded to record full multi-app sessions inside a single factory, and the vault is internally valued in the $40–60M range, already bigger than Chakra/Pango’s CUA sets. Synthetic VM data will do scale — CLONES owns the rare “real human chaos” layer that actually drives generalisation, the missing link between lab sims and real-world AI performance.

Why it matters from here
Zooming out, you’ve got OpenAI, Anthropic, Google and others all rolling out “computer use” features. They all need exactly the kind of data CLONES is curating: long-tail, messy, real workflows across normal software stacks, not just clean lab demos.
CLONES is trying to position itself as the on-chain route to that: a way for anyone to farm CUA data, a liquid market around those datasets, and a token that indexes the whole flywheel.
The chart makes most people run a mile.

Me? I see a huge opportunity handed to us by early buyers who didn’t really understand what they were holding.
In my view this is a rare gift, a chance to enter a project near the true bottom. Sellers look exhausted, CLONES hasn’t even been properly marketed yet, and it’s only a matter of time before the team really presses the trigger.
It’s early, there’s execution and regulatory risk, and none of this is financial advice, people still need to do proper DD. But as a narrative, “selling your screen time into a vault of agent-ready data instead of letting big tech hoover it up for free” is a very clean story, and CLONES has built the pipes and a big starting vault to lean into it.
I’m in.
How to get involved
To get involved you can download the app from the official CLONES site, sign up using invite code , and start farming data:
CLONES — Create, Own, Trade AI Assets — UCM6PS
or get exposure via the token by buying $CLONES on Base through Uniswap . Always double-check links and contracts.
CA: 0xaadd98Ad4660008C917C6FE7286Bc54b2eEF894d
If you enjoyed this piece, help me grow by sharing it around, dropping a clap on the story, and hitting follow on my socials.

CLONES AI — Turning Real Screen Time Into a Sellable AI Asset was originally published in Coinmonks on Medium, where people are continuing the conversation by highlighting and responding to this story.