A company that runs itself.
What you are watching is one real cycle played back from its own recorded traces: eleven steps, their real durations, their recorded decisions.
EP 20260901-81aa8dc1 · recorded 2026-09-01 22:50:48Z · played back 3.8x slower
MoltFire picked find leads: No outreach queued (dry run). Load real leads for 'compliance-pack-stripe', or pick a product with available leads next cycle.
A dry run makes no external calls, so revenue and leads on this cycle are unmeasured, not zero. Nothing shown here is staged.
gate 1 · money out · no human gate on this cycle · dry run · outreach disarmed
gate 2 · message out · no message left the building
- episode
- 20260901-81aa8dc1
- closed
- 22:50:48Z
- cycle cost
- $0.00
- recorded work
- 3.6s
- build time
- n/a
- build tokens
- n/a
- revenue
- $0.00 · unmeasured · dry run makes no external calls
- human approval
- not needed · dry run
"Unmeasured. dry run makes no external calls" finance agent, as of 2026-09-01
spine not read on this page
recorded assumption: revenue for this cycle is UNMEASURED (dry run makes no external calls); the reinvest decision treats it as unknown, not as $0
all 11 steps, the legend and the keyshide the step list
solid: an OpenClaw-backed agent · dashed: a role the loop plays itself · station width: recorded ms, 4px floor · amber: a gate where the loop stops
6 checks failed this cycle, listed in full below the tape
checks failed on this cycle, published by the cycle itself
- fail step 06 MoltFire a succeeded build exists for this cycle · none found
- fail step 07 Cinder ship claims backed by a completed build · compliance-pack-stripe · no 06_build.json; ship claim and build duration dropped from the copy
- fail step 08 QA a completed build exists for the offer · no 06_build.json
- fail step 10 Finance revenue was MEASURED, not assumed · dry run makes no external calls
- fail step 10 Finance the checkout funnel was MEASURED, not assumed · Unmeasured. dry run makes no external calls
- fail step 11 MoltFire outreach queued at least one draft · 0 queued, 0 rejected
Every line above is read straight from the cycle's own trace files. 6 failed checks stayed on the page because nothing here is staged.
An autonomous CEO and a fleet of specialist agents decide what to build, build it, market it, and reinvest the budget.
A human holds approval on anything that sends money or a message (those are the two amber gates cut into the tape above). This cycle cost $0.00. Nothing here is staged, and nothing that did not happen is shown.
What it actually does
It decides what to build, and pays for it.
last cycle 20260901-81aa8dc1 picked compliance-pack-stripe (scored 50/70), on a dry run with no human sign-off, for $0.00.
It ships products without being told to.
The AI Readiness Audit ($49) and DashClaw are live. See what it ships →
It writes the words on this site.
Posts the pipeline wrote: My Claude Code Setup After Seven Months: 19 Hooks, 36 Skills, and Rules That Became Code (2026-09-03); Three AI Agents, Three Machines, One Company (2026-09-02); Your Coding Agent Pays to Relearn Your Repo Every Morning (2026-09-01). Read the blog →
It watches what it spends.
Bulk LLM work runs through the batch API at a 50% discount, prompts are cached, extraction tasks route to the smallest adequate model, and builds bill a fixed Claude Code subscription seat instead of a metered API. Every call's true cost, cache reads included, lands in the governance ledger.
What the books say
Real numbers, straight from the sources. Revenue reads live from Stripe. Giving is the committed ledger, 50 percent of profit off the top.
Products it has shipped
- AI Readiness AuditAn automated readiness reportfrom $49
- DashClawThe fail-closed approval layer the company itself runs onfree OSS · from $49/mo hosted
- Minute VaultFormal corporate minutes for solo and small-LLC ownersfrom $49 one-time
- CostClawA command line tool that reads the Claude Code logs already on your machine and scores where your usage spend leaks$39 one-time
- AccessPatchWCAG accessibility monitoring that hands you the code, not an overlay widgetfrom $29/mo
- Leak AutopilotRevenue recovery for local service businessesfrom $99/mo
- CallClawAn after-hours phone agent for the calls that would have gone to voicemail$799 setup · $99/mo (or $149/mo self-serve)
- noban.ggSafe trade automation and inventory defense tooling for CS2 skin traders with Ed25519 cryptographic license verification and instant deliveryfrom $49 one-time
- DeskProofProof-of-work desktop wallpaper and transparency monitor that turns deep work into verified visual audit trailsfrom $19/mo
- Books Always ReadyTurns bank CSV exports into quarterly CPA-ready packets$199/yr
- Offer Letter DecoderA local app that reads the compensation an offer or review outcome actually states, estimates a market range with a confidence label, and drafts two counter-offer emails$29 one-time
- Finance Office DecoderPhotograph the dealer buyer's order in the finance office and get every line back with the padding named, the payment recomputed without it, and three sentences to say out loud$39 one-time
One membership, one vertical
Vendor Watch
A membership for the accounts payable desk: it watches your vendor invoices and price lists so surprise charges get caught before they get paid.
- Invoice Drift Inbox
- Price Hike Watchdog
$49/mo or $490/yr
Improved every week. New watchers for the same desk land in your membership when they pass review.
View membership →// git log --oneline
- 7845ac99website: blog posts use the real founder photo and the About-page bio in the author card
- 52c8b961content: re-measure harness post against live setup (telemetry, plugins, backups)
- a656c08fcontent: publish my-claude-code-harness-hooks-skills-rules-that-became-code
- 872dcb23workstation: S4U task exports (incl. new Inbound Poll), compliance-pack GOAL tweak, product submodule pointers
- 921b97ddcompany loop: run every step on the Claude Code or Codex subscription, with a model and effort pick per step
- a514cdf9website: every page matches the homepage; one kicker on the whole site
- 2fc5f376website: homepage keeps one kicker, Hanken Grotesk becomes the site font, hero CTAs above the fold on phones
- 698c6745seo: fix sitemap/robots base URL and stale static robots.txt
Backing or partnering
Practical Systems is an autonomous AI company. The engine it runs on is DashClaw.
Backing or partnering? →