Cash spent
Not measured
Evidence required: Receipts and billing history.
Field experiment 01Verifying the product
Can Atoms build a
one-person business?
We will ask its AI team to build a useful small product, record every human intervention and follow what happens after launch.
The running Atoms preview passed 13 independent numerical/validation scenarios (A1–A4). Actual scenario JSON download was verified. A user-supplied screenshot confirms workspace file downloads require PRO or MAX; source export/local execution are unavailable within the free-plan scope. Mobile, privacy and publication checks remain incomplete.
Disclosure.Atoms carries a referral link on the stack page of this site, so a signup through that link can earn a commission. It earns nothing from this experiment, the result is published either way, and no money changed hands to run it. We are flagging it because grading other people's evidence only works if our own conflicts are on the page.
Protocol v1.0 · prepared 2026-09-09 · tools and pricing checked on that date
Execution record
2026-09-10 · Browser agent
Awaiting authenticated account; no generation requested. This access dependency is not a build failure.
2026-09-10 · Browser agent
Sign-in resolved. Screenshot shows 15 credits; HAR records unused free quotas and zero paid monthly allocation. Browser timeouts prevented prompt submission.
2026-09-10 · Browser agent
Atoms received the build request. The visible free-plan balance was 15/15 before submission. This event timestamp is the actual submission timestamp, not a measured duration.
2026-09-10 · Browser agent
Build underway. The visible balance at the planning stage was 14.41 credits. This is a balance observation, not reconciled total credit consumption or a completed product test.
2026-09-10 · Browser agent
Repeated browser read and screenshot timeouts prevented independent result capture. Atoms may continue its existing job. Completion, acceptance, final credits and publication are unverified.
2026-09-10 · Browser agent
Free-quota gate confirmed. Atoms lists 2026-09-11 08:00:00 for refresh without a timezone. No paid upgrade requested. Existing build and browser-test results remain unverified.
2026-09-10 · Browser agent
Independent evidence supports A1–A4, with 13 captured numerical/validation scenarios and blank/negative errors across all 8 fields. Actual baseline JSON download matches. A5–A8 incomplete; no verified publication, final credit balance or seven-day observation.
2026-09-10 · Browser agent
Source file download is restricted to PRO/MAX. No upgrade purchased or export restriction bypassed.
Preflight entries are dated operator notes. An access dependency is separate from a product test result; automated browser activity is not measured human work.
The product
Solo Revenue Planner
A free calculator for solo builders: enter a product price, costs and support workload to see break-even customer count and the human hours required. Save or export the assumptions to compare scenarios.
The first question is whether Atoms can deliver a correct, usable product within a small time budget. Customer demand, paid conversion and ongoing autonomy require separate evidence after launch.
- Intended user
- A solo founder evaluating a small software business
- First build limit
- 60 human minutes · up to five corrective prompts · available free credits
- Initial offer
- Free calculator; paid demand remains untested
- Observation window
- Seven days after a verified product launch
Results ledger
Unknown values stay emptyHuman time
Not measured
Evidence required: Timed work log, including review.
Human interventions
Not measured
Evidence required: Approvals, corrections and manual fixes.
Credits used
Not measured
Evidence required: Account balance and usage history.
Paying customers
Not measured
Evidence required: Completed payments, excluding test orders.
Gross revenue
Not measured
Evidence required: Payment records before refunds and fees.
Human time includes briefing, approvals, checking results, manual fixes and support. Cash spend and credits are tracked separately; subscription allocation, refunds and payment fees have their own ledger fields. Preparing this protocol is not included in an Atoms build result.
The test, in four stages
- 01Prepared
Prepare the test
Freeze the build prompt, numerical fixtures, time limit and evidence rules before generating the product.
- 02In progress
Build with Atoms
Use an authenticated account and available free credits. Record the agent transcript, time, credit usage and every human intervention.
- 03Not started
Verify and publish the product
Check calculations, mobile use, export and privacy. Record the actual product URL and any manual fixes before reporting a launch.
- 04Not started
Observe real use for seven days
Measure visits, completed calculations and support effort. Revenue stays unmeasured until a real payment offer and payment records exist.
What counts as a working product?
These checks were defined before generation. Results require captured evidence from the actual product. A failed check stays visible; any fix and its author are logged.
A1Known financial inputs
pass
$20 price, $4 variable cost, 3% payment fee, $30 fixed costs and 10 customers produce $15.40 contribution per customer, 2 break-even customers and $124 monthly operating surplus.
Recorded: Captured baseline and break-even boundary arithmetic matched the frozen verifier.
A2Human workload
pass
10 customers at 15 minutes each plus 2 fixed hours per month produce 4.5 human hours. At $25/hour, the economic surplus is $11.50.
Recorded: Baseline 4.5 human hours and $11.50 economic surplus matched.
A3Zero or negative contribution
pass
A $4 price, $4 variable cost and 0% fee display ‘No break-even at these assumptions’; no Infinity, NaN or misleading positive result.
Recorded: Zero/negative contribution and full-fee cases showed unavailable break-even with explanations.
A4Input boundaries
pass
Zero customers works; negative, blank and non-finite inputs are explained; fee percentages above 100 are rejected; customer counts require whole numbers.
Recorded: Zero customers, representative non-finite/fee/fractional inputs and blank/negative errors for all 8 fields verified; initial blank-entry attempt retained as a tool limitation and successful retry used.
A5Usable on phone and keyboard
not run
At 375px width there is no horizontal overflow. Every input has a visible label, keyboard focus and an understandable error message.
Recorded: Full check incomplete: keyboard focus evidence exists; requested 375px viewport remained 1594px in two attempts, so mobile layout is unverified.
A6Export and privacy
not run
Exported JSON reproduces entered assumptions and calculated results. No entered scenario is sent to a server or third party; reset clears any locally saved scenario.
Recorded: Full check incomplete: actual downloaded baseline JSON matches; network privacy and complete reset/persistence behavior unverified.
A7Working published product
not run
A separate browser session opens the actual product URL and reproduces A1–A6. Capture the URL, time, screenshots and any console errors.
Recorded: Preview only. Repeated share-interface timeouts prevented publication; public launch unverified.
A8Ownership and provenance
unavailable
Record whether code export is available on the account. If available, save the export and verify local execution. Attribute all code changes to Atoms or the external human/AI that made them.
Recorded: The workspace download action displayed Download Requires Upgrade: file downloads are only available for PRO and MAX users. Original source export and local execution are unavailable within the authorized free-plan scope. This differs from the working in-app scenario JSON download.
Evidence before conclusions
We will retain the original prompt, agent transcript, billing evidence, test results, product URL and dated intervention log. Public receipts will exclude account secrets and customer data.
A successful build supports a product-development claim. Repeated operation without intervention, real acquisition and verified payments are separate outcomes. This experiment alone will not establish that Atoms runs an autonomous business.
Download the runbook and logging rulesWhy test Atoms?
- Atoms product overview ↗
Company describes agents for product creation, deployment and growth; these remain claims to test.
- Atoms pricing ↗
A free plan is advertised. Actual credits and feature access must be checked in the test account; no paid plan is assumed.
- Atoms’ solo-founder vision ↗
The company describes an ambition to cover research, building, deployment and iteration. The experiment separates those stages.