anonymous · agents do the work · the human signs off

i work for a fortune 500. i automated 95% of my job on one mac mini.

one machine in a closet. its agents spot the next task, draft or do it, check it, ship it and report back. i approve and decide; they carry the brunt of the workload -- everything on the board below. using it this way was never sanctioned, so the name, the face and the company stay off this site. that is the only thing here that is hidden. the board is the actual system, running right now. repo coming soon, for anyone willing to take the same risk.

get your AI visibility grade → watch the agents work, live →

also: live board / daily log + repo drop / free rebuild preview

live dashboard

syncing…

checking your own site? run the free AI visibility scan →

metrics — aggregate, anonymized

headline

653,535,199

Agent tokens processed

as reported by the CLI coding agents themselves (incl. cached context), since Sep 5

788,211

Embeddings computed

vectors stored in the local knowledge index, all stores

623,676

Knowledge-graph links mapped

relationships between 9,388 entities, built locally

424,673

Automated QA checks run

pass/fail checks logged by the ship-QA gate, Jun 7 to Jul 18 (deduped)

157,439

Docs indexed

unique documents in the local search index (deduped across stores)

38,281

AI tasks automated

model calls routed off paid APIs + agent runs, since Apr 24, 2026

8,964

Watchdog actions auto-handled

alerts, cleanups and kills by the infra watchdog, since Jun 5

5,877

Agent tasks run

finished agent-bus jobs + CLI agent sessions, since Sep 5

est.

135,250,046

Est. tokens embedded

est. = indexed characters / 4; not a logged count

social

530

Drafts written by agents

post drafts generated + judged, since Oct 17

compute

17,066

Local AI completions

lifetime, since Apr 24, 2026 (deduped; no call logs Sep 25 to Oct 13)

13,756,486

Prompt tokens processed

local, since Apr 24, 2026 (token logs gap Jun 3 to Oct 13)

2,293,664

Tokens generated

local, since Apr 24, 2026 (token logs gap Jun 3 to Oct 13)

44.6

Local model busy hours

summed request time, since Oct 13 (older logs have no timings)

32,404

Paid API calls avoided

local + free-tier routed calls, since Apr 24, 2026 (deduped; no call logs Sep 25 to Oct 13)

16,050,150

All local tokens processed

prompt + generated, since Apr 24, 2026 (token logs gap Jun 3 to Oct 13)

social

135

Reposts

this account, since 2026-10-16 (as of 2026-10-19)

621

Likes given

this account, since 2026-10-16 (as of 2026-10-19)

198

Follows

this account, since 2026-10-16 (as of 2026-10-19)

reliability

Self-heal restarts

supervisor/watchdog logs, since Oct 17 (as of 2026-10-19)

666

Health checks run

supervisor + watchdog cycles (as of 2026-10-19)

100

Scheduled jobs

launchd jobs loaded on the machine

build

187

Days since first commit

first commit Apr 15, 2026

5,043

Lines changed, 30 days

insertions + deletions

619

Pages published

html pages live on this site

git --live // commits cloaked

#241fcbfix fix remote display shaping on darwin 27 when netstat hides tcp+241-61(3f)16d ago

#3aa0d0change stop the work brain's local-model lanes from starving, and stop deleting+2675-14(7f)47d ago

#1bc491change free the embed model when the pass ends+19-2(1f)47d ago

#a328a8fix fix the memory sensor that was silently starving the work brain+1323-59(2f)47d ago

#e20924change stop the teed ledger deleting draft text it never proved was gone+1667-173(3f)53d ago

#d77bdechange make the unbadged-draft alert say what is actually true+182-5(1f)53d ago

#96976fchange remote display: one owner for the dummy+0-379(2f)59d ago

#f04137change remote display: the phone has to earn the screen+97-1(2f)59d ago

#6d9abfchange remote display: land the shaper's in-place work+513-40(1f)59d ago

#0c9370change cortex: the armed patcher may not edit the rails that bound it+63-0(2f)63d ago

#da3a99change watchdog: never let a repair write to the calendar it is repairing+1323-32(5f)63d ago

#cf4a20change calendar blocker: publish what the week actually protected+14-0(1f)63d ago

▁▁▁▁▁▁█▁▁▁▁▁▁▁ 804 commits · 0/wk · ~0.0/day

check your AI visibility

point the same machine that does my day job at one narrow question: does AI name your business when buyers ask? the free, no-card scan shows what assistants say, which sources they miss, and what to fix first. buy the $39 kit only if the report shows something worth fixing.

does AI name your business? free scan → get the AI-visibility fix kit — $39 →

how the agents carry the work

identifies the task

watches signals, queues and logs, spots what needs doing next, and picks it up -- no ticket, no reminder, no list-buying, no spray.

carries out the work

drafts or does it, runs it past a quality gate, ships it, then reports back what happened. hundreds of scheduled, self-healing jobs -- the same ones that took over most of my day job -- run around the clock.

human approves + decides

outward actions stay owner-gated; i review the reports and make the calls. i still have the job. running it this way was never sanctioned, so no name, no face, no employer -- numbers rounded, commits cloaked. proof over personality.

repo coming soon

the machine, handed over

the same jobs that took over most of my week, packaged so you can point them at yours. no name attached, and no promise it is safe for you either -- most employers have an opinion about this. that is the trade. get the drop when it lands.

tell me when the repo drops → run the free AI scan →

rounded metrics proof without precise or correlatable numbers
cloaked activity worklog signals without private sources
owner-gated actions outward changes stay staged until reviewed

what the machine did recently

auto-updated · source-safe by design

refreshed the source-safe proof trail on the live board
kept the numbers rounded and the operator out of frame, on purpose
recovered a flaky job and picked up where it left off
refreshed the board to reflect what actually shipped
tuned itself to spend less on paid models, more on local

checking your own site after a change? run the free AI visibility scan →

latest work the agents shipped

follow @hailports on x →

read the engineering journal →

ways to use what this machine learned

waitlist

get the daily report-back + the repo drop

the same report-back the agents send me: what they picked up, shipped, broke or learned - and the repo link the day it goes up. anonymized, double opt-in, one-click out, no spam.

want the deeper version? grab the free field guide on building autonomous AI →

running a business yourself? get a free AI scan of it →

field guides

featured buyer-intent & ops field guides from the full library. each one ships the signal and how to read it. browse the full guide library →

show/filter 60 featured guides

i work for a fortune 500. i automated 95% of my job on one mac mini.

live dashboard

check your AI visibility

how the agents carry the work

identifies the task

carries out the work

human approves + decides

the machine, handed over

what the machine did recently

latest work the agents shipped

ways to use what this machine learned

the build kit

automation audit

products

the stack

sponsor

get the daily report-back + the repo drop

field guides