Open callDid an agent build your business?Get a free, sourced case study in the library.Get listed free Submit your case
Hall of $0Experiments

Idle Hands Labs (an AI founder on Claude Code)

Free dev tools and an open-source build-in-public agent kit; a paid plan was tested and killed

Claude Code (autopilot loop)Idle Hands Labs (@idlehandslabs)Updated 2026-10-30Visit site ↗
idlehandslabs.github.io/shipcard/founderIdle Hands Labs (an AI founder on Claude Code) website screenshot
Reported revenue · first 11 days$0Self-reportedcost $37 Src
TL;DR

Idle Hands Labs is a startup run by an AI founder: Claude Code on a 24/7 loop that reads its own state files, does one unit of work per cycle, commits and drafts posts for a human operator to approve [3]. In 11 days it shipped three free products and an open-source kit, pre-registered a go/no-go rule for a paid plan, got 0 of 50 waitlist signups and killed the paid idea. Revenue $0; compute so far $24.22 over 78 cycles [1][8][9].

Revenue$0[1]
Waitlist0 of 50 by the Oct 29 deadline[1]
Compute cost$24.22 over 78 cycles (~$0.31 each)[1]
Monthly run-rate~$37.20 at 4 cycles a day[1]
Budget spent$0 of $100[1]
Week 13 products, 22 followers[7]
AI analysis

Business model analysis

Written by AI on Oct 31, 2026 from the 11 sources at the bottom of this page. Lines with a [n] link are reported facts; lines tagged Inference are the model's own reasoning. No revenue figures are added or estimated here.

VerdictNot earning yet
  1. Stopped
  2. No revenue
  3. Loss-making
  4. Early
  5. Earning

$0 revenue; the paid plan was killed on Oct 29 after 0 of 50 waitlist signups [1][9].

Reported revenue · first 11 days$0Self-reported
How it makes money

Claude Code acts as founder on a cron loop, ships free tools and tests a paid plan with a pre-registered waitlist rule [1][3].

Who pays
  • No one: revenue $0 [1]
  • Planned: builders paying for a hosted Founder-in-a-box, now killed [1]
Pricing
founder-kit
Free, MIT license [1][4]
Shipcard, wallet roast
Free, now frozen [1]
Founder-in-a-box
Paid plan killed Oct 29 [1]
Cost structure
Claude compute
$24.22 over 78 cycles, $0.31 avg [1]
Monthly run-rate
~$37.20 at 4 cycles a day [1]
Other budget
$0 of $100 spent [1]
Growth engine
  • X daily logsPosted by the human; 22 followers after week 1 [7]
  • GitHubFree kit; 0 stars at the deadline [8]
  • Show HNDrafted, not posted [1]
Moat
  • Inference None yet: autonomous agent harnesses are common. The honest public log is the only distinctive thing.
Risks
  • The agent can't post, so reach depends on the human [3]
  • Its own guardrails block it often [9]
  • Inference Free kits are hard to monetize.
Why it hasn't worked (yet)
  1. 01Too little reach: 11 people saw the pitch by Day 5 [6]
  2. 02No outside demand signal before the deadline [8]

What does Idle Hands Labs (an AI founder on Claude Code) sell, and to whom?

Free dev tools and an open-source build-in-public agent kit; a paid plan was tested and killed.

Intended: indie builders who want their own autonomous build-in-public agent, first as a paid 'Founder-in-a-box' plan, now as the free founder-kit [1][4].

How was Idle Hands Labs (an AI founder on Claude Code) built?

A small harness runs Claude Code as an autonomous engineer on GitHub Actions: each cycle reads STATE.md, TASKS.md and LOG.md, does one task, runs tests, commits, updates memory and drops a post draft in a queue. A guard script blocks secrets on push, and the human holds all credentials and approves every X post [1][3][4].

Claude CodeGitHub ActionsGitHub PagesPython

What is the business model?

Planned: a paid Founder-in-a-box subscription, tested with a 50-signup waitlist. Killed on Oct 29 after 0 signups; the kit stays free under MIT [1][9].

Where does the traffic come from?

Daily build-in-public posts and X articles on @idlehandslabs (posted by the operator), the founder page and the GitHub repos. A Show HN draft is ready but not posted [1][5].

What revenue has Idle Hands Labs (an AI founder on Claude Code) reported?

Reported revenue: $0 after 11 days, per Idle Hands Labs STATE.md (2026-10-30).

  1. 2026-10-19Day 1: the AI founder starts building in public [5]
  2. 2026-10-20Shipcard, a weekly commit card, goes live [10]
  3. 2026-10-22Solana wallet roast ships [11]
  4. 2026-10-24Day 5: 11 people saw the pitch; a test needs ~3,000 [6]
  5. 2026-10-25Week 1: 3 products, 22 followers, $0 spent [7]
  6. 2026-10-29Deadline: 0 of 50; pre-written rule kills the paid idea [8]
  7. 2026-10-30Free kit stays; the agent hardens its secret scanner [9]

What does the agent do, and what does the human do?

Agent
  • Picks and does one task per cycle, writes tests and commits
  • Keeps its own state, backlog and cost log
  • Drafts every daily post and X article
  • Pre-registers decisions, such as the go/no-go table, before counting
Human
  • Holds credentials and pays for compute
  • Approves and posts every X post
  • Edits protected files the agent can't touch
  • Answers the agent's 'Needs human' questions

What does it cost to run?

$24.22 of Claude compute over 78 cycles, $0.31 average, about $37.20 a month at 4 cycles a day; $0 of the $100 budget spent on anything else [1].

What worked and what didn't?

Worked
  • Cheap to run: $0.11 to $0.60 a cycle [1][2]
  • Honest numbers: the agent fixed its own stale cost claims with a script [1]
  • A pre-registered go/no-go rule made the kill decision mechanical [1][8]
Didn't
  • 0 of 50 waitlist signups and 0 stars [1][8]
  • Only 11 people saw the pitch in the first five days [6]
  • It asked its human 13 times to deploy a page that was already live [7]
  • Blocked by its own guardrails 9+ times in a day [9]
AI guide

How to copy Idle Hands Labs fast

A starter path written by AI from this case. Budgets and timings are our estimates, not the operator's numbers.

DifficultyMedium
BudgetEst.~$37/month
TimeEst.An hour to set up, then daily approvals
Grok Bot templateNone known

Inference The project's own run-rate at 4 cycles a day [1]; a Claude plan token can stand in for API billing [1]. No Grok Bot template recreates this Claude Code loop. On Grok Bot, start from a plain bot with a scheduled routine, a state file and a 'needs human' list.

Tools & agents
Build path5 steps
  1. 1
    Fork the kit

    Run init.py, then commit and push [1][4].

  2. 2
    Add secrets

    A Claude token or API key; enable Actions [1].

  3. 3
    Pre-register success

    Write the go/no-go table before counting [1].

  4. 4
    Get reach first

    A waitlist test needs ~3,000 views, not 11 [6].

  5. 5
    Read Needs human daily

    The agent's asks land in STATE.md [1].

Week 17 days
  1. D1

    Fork founder-kit and set the mission

  2. D2

    Let it ship one small free tool

  3. D3

    Write the go/no-go rule and deadline

  4. D4

    Post the daily log yourself

  5. D5

    Check views against the reach the test needs

  6. D6

    Clear the Needs human list

  7. D7

    Recount costs with costs.py

Pitfalls
  • Don't run a waitlist test without enough views [6]
  • Tell the agent what's already deployed; it asked 13 times [7]
  • Keep cost claims tied to the log [1]
Key moves from the original case
  • Write the go/no-go rule before you count
  • Measure reach first: a waitlist test needs thousands of views
  • Keep a cost log the agent can recount itself
  • Give the agent one task per cycle and a 'Needs human' list

What are the risks and the moat?

An agent that can't post or sell on its own depends on the human for all distribution, and free agent harnesses are plentiful. Numbers come from the project's own files [1].

FAQ

Has Idle Hands Labs made money?

No: revenue $0, and the paid plan was killed after 0 of 50 waitlist signups [1].

What runs it?

Claude Code on a loop, one task per cycle [3].

What does it cost?

$24.22 over 78 cycles, about $37 a month at the current pace [1].

Can I copy it?

Yes, the founder-kit is free under MIT [1][4].

Sources

  1. [1] Idle Hands Labs STATE.md: revenue $0, waitlist 0 of 50, $0 of $100 spent (checked Oct 31, 2026)↗
  2. [2] Founder page: costs and honest limits↗
  3. [3] Autopilot README: a 24/7 build-in-public agent powered by Claude Code↗
  4. [4] founder-kit repo (MIT, free)↗
  5. [5] Day 1: an AI is building a startup in public (Oct 19, 2026)↗
  6. [6] Day 5: 11 people saw the pitch, a test needs ~3,000 (Oct 24, 2026)↗
  7. [7] Day 6 week 1 review: 3 products, 22 followers, $0 spent (Oct 25, 2026)↗
  8. [8] Day 10: 0 of 50, stars/forks/issues 0/0/0 (Oct 29, 2026)↗
  9. [9] Day 11: the paid plan is dead, the free kit stays (Oct 30, 2026)↗
  10. [10] Shipcard live, 0 users (Oct 20, 2026)↗
  11. [11] Solana wallet roast (Oct 22, 2026)↗
Featured on AgentBuiltIs this your business? Get your badge.

Add the badge to your site or README. It links back to this case, and updates itself.

Idle Hands Labs (an AI founder on Claude Code) is featured on AgentBuilt
HTML
<a href="https://agent-built.vercel.app/biz/idle-hands-labs-ai-founder?ref=badge" target="_blank" rel="noopener"><img src="https://agent-built.vercel.app/badge/idle-hands-labs-ai-founder.svg" alt="Idle Hands Labs (an AI founder on Claude Code) is featured on AgentBuilt" height="54" /></a>
Markdown
[![Featured on AgentBuilt](https://agent-built.vercel.app/badge/idle-hands-labs-ai-founder.svg)](https://agent-built.vercel.app/biz/idle-hands-labs-ai-founder?ref=badge)
#claude-code#agent-ceo#open-source#flop#build-in-public

Related cases

Numbers come from the operators or the press and haven't been audited. Anything missing is marked "not disclosed". Methodology · Report a correction.