Download PDF

Weave. Wise. Webs.

We create thinking solutions.

Agency · Consultancy · Lab

ARACHN

We create thinking solutions.

We believe the better this technology gets, the less a brand should need an agency. That is not a threat to our business, it is the reason we started one. A company that runs its own marketing, on its own systems, with its own people, is stronger than a company renting the ability. So we build toward the day you can take it in house, and we say so before you ask.

Getting there is a partnership, and both sides have real work in it. Ours is the integration: plugging intelligence into the tools you already pay for, wiring your systems to talk to each other through their APIs, and putting harnesses around them so the machine can be trusted with the repeated parts. Yours is the judgment. People review and approve, and people decide what the business should do next.

What comes back is time. When the systems talk to each other, and to the people who run them, the week stops going into keeping the business running and starts going into building it. The ideas about what to build next come from both sides.

What we are for
ARACHN
02 / 17

Arachn is an intelligence shop with three main service areas: an agency, a consultancy and a lab. The agency runs our instance of Arachn for you. The consultancy helps design your version of Arachn. The lab helps build you your Arachn, wired into your tools and hands it over with the runbook and the training.

Webster Reed is the shop's executive producer, live since July 2026 on its own infrastructure. It works the way a good producer works: takes the brief, staffs the specialists the job needs, walks the work through every check, then brings it to a named human and waits. It has never published anything without a yes, because it structurally cannot. It is a colleague with a job description, not a tool somebody switches on.

Who we are
ARACHN
03 / 17

Most agencies are built to be hard to leave. The accounts sit in their name. The passwords live on their machines. The knowledge walks out the door when the account manager does. Nothing gets written down, because written down is how you get replaced.

Arachn is not only our company, it is also the one product we sell at the same time: an operating model where software does the repeated work, a named person approves anything that leaves the building, watchers report when a piece of it stops, and every action leaves a record. We build it, run it, break it, fix it and improve it here first, every week, on our own business. It is better this month than last month, and it will be better next month, whether or not anyone buys it.

So the product we sell is built to be handed over. The goal is a company that could eventually run this without us, which is why there is one product with three ways in rather than three separate businesses. They differ mostly in how much of it you want in your own hands.

Agency

Hire the agency and we run our instance of it for you.

Consultancy

Hire the consultancy and we design your version of it: what moves to software, what never does, and where the approvals sit.

Lab

Hire the lab and we build you your own, wired into your tools, and hand it over with the runbook and the training.

Many services, one product, three ways in
ARACHN
04 / 17

Agency

Solutions that run on a cadence and answer for results over time.

The running things: campaigns, social programs, media buys, email programs, search work, reporting rhythms. The work that has to happen every week, run by us, on your accounts, in your name. You approve; it goes out; it gets logged. Nothing publishes, sends, spends or goes live without a named person saying yes, and if that person says nothing, nothing happens.

Social media management

We plan, write, design, gate, publish and log one piece a week on your accounts.

DEMONSTRATED
Social content production

One idea becomes a card set and a vertical video, rendered in-house on a locked template, every week.

DEMONSTRATED
Marketing operations

Every job passes an independent check and a named gate, and leaves a record you can read.

DEMONSTRATED
AI search optimization (GEO)

We make your site something an answer engine can read and quote, then measure every month whether it is quoting you.

DEMONSTRATED
DEMONSTRATED2
AI answer advertising DEMONSTRATED
Paid media management DEMONSTRATED
BUILT1
Marketing analytics and reporting BUILT
NO RECEIPT YET3
Email marketing management NO RECEIPT YET
Search engine optimization (SEO) NO RECEIPT YET
Community and reputation management NO RECEIPT YET
DESIGNED1
Outsourced marketing department DESIGNED
Many services, one product, three ways in
ARACHN
05 / 17

Consultancy

Thinking that guides decisions and defines what gets built.

The written things: audits, assessments, POVs, strategies, roadmaps, architectures. The decisions the other two parts build from. Where intelligence belongs in your operation, what it must never touch, what you actually stand for, and the order to do things in. Every recommendation carries what it is based on and how sure we are. It is yours to keep and act on, with us or without us.

AI visibility audit

A written before-measurement of what machines can read about you, what they currently say about you, and what is missing.

DEMONSTRATED
Brand strategy, positioning and naming

A position you can defend, in words your own people repeat, plus the naming and the recognizable assets that carry it.

DEMONSTRATED
Messaging architecture

One architecture every channel writes from, with proof attached to each claim.

DEMONSTRATED
AI operating model design

We design the model first: which work moves to the machine, which decisions stay human, where the stopping points go and who holds each one.

DEMONSTRATED
DEMONSTRATED1
AI readiness assessment DEMONSTRATED
BUILT1
Content and claims audit BUILT
NO RECEIPT YET6
Digital marketing strategy NO RECEIPT YET
Martech and data architecture NO RECEIPT YET
Measurement and attribution planning NO RECEIPT YET
Audience research, personas and journey mapping NO RECEIPT YET
Pricing strategy NO RECEIPT YET
Hiring and interview design NO RECEIPT YET
Many services, one product, three ways in
ARACHN
06 / 17

Lab

Creations that are proven in front of you before being handed over.

The made things: sites, stores, portals, agents, pipelines, content systems, creative. We build them, we prove them in front of you, and we hand them over with the runbook and the training so your team can hold them.

AI agent development

A named agent on its own infrastructure doing a real piece of your weekly work, where nothing reaches the outside world without your person saying yes.

DEMONSTRATED
Website design and development

Fast, editable sites built by the shop that will have to keep them running afterward, which changes what gets built.

DEMONSTRATED
Rapid prototyping

A working thing, built against your real problem, shown live.

DEMONSTRATED
Multi-agent systems

A bench of specialists behind one producer, with a single approval point at the edge, so capacity grows and accountability does not spread out.

DEMONSTRATED
DEMONSTRATED5
Content systems DEMONSTRATED
Workflow automation DEMONSTRATED
Data engineering and integrations DEMONSTRATED
Ecommerce, portals and applications DEMONSTRATED
Brand identity and design systems DEMONSTRATED
BUILT1
Your own Arachn, built in-house BUILT
NO RECEIPT YET2
Campaign creative and video production NO RECEIPT YET
Email design and development NO RECEIPT YET
Many services, one product, three ways in
ARACHN
07 / 17
Many services, one product, three ways in
ARACHN
08 / 17

Work moves in loops, not phases.

Think, Create, Solve, repeating. It crosses a gate on every pass.

What a gate is

A person says yes before anything reaches the outside world, and silence never counts as a yes. The check in the middle is a quality pass by someone who did not build the thing, with screenshots as proof.

Think Consultancy · Wise Create Lab · Weave Solve Agency · Webs The first look a person, before it is built on The check independent QA The last door a named person, always
the work crossing a gate the work reaching Think, Create or Solve
How we work
ARACHN
09 / 17

The shop is three.

A 108-role bench in three teams plus the people who check them, seven industry contexts, and seven runbooks, deployed and staffable.

An intelligence
Written, designed
and built

Software does the making, the measuring and the running.

An executive producer
ProducedWebster Reed

It takes the brief, staffs the specialists, moves the work through its checks, and then waits for a person to say yes.

A founder
Judged, gated
and approved

A founder’s whole attention against a locked cap of three clients.

The bench108 specialists, staffed per job
Weave25
Create
Wise21
Think
Webs45
Solve
Checks17
Every job
Our team
ARACHN
10 / 17

Receipts, not case studies.

Owners disclose gaps upfront. Arachn is our own first client: everything we sell gets built, run, and broken here first, and work done here is real work. It is the proof. Every receipt in this document is an artifact you can open, with a date on it, from a system that was actually running.

A case study is a story about work, written afterward, by the people who want the credit. A receipt is the work itself. Each one is paired with the service it belongs to and the marker that service carries, so you can see exactly what the evidence covers and exactly where it stops.

There are two ways to sell a capability. One is to describe what you would assemble if somebody paid you. The other is to build it, run it, live with it, and then hand over the logs. We do the second one, and this document is arranged so you can tell the difference line by line.

The cost of working that way is that the failures are ours. A vendor cut off our publishing access for good and we switched rails the same day. A guard stopped a scheduled job rather than let it ship something wrong. We counted our own bench and found that all 108 files carried a fake model name and would have died at launch. Five AI platforms were asked who we are and they answered with Greek mythology, and we published that too. None of it happened to a client, because we run everything here first. A shop that has never published a failure has either never shipped anything or is not telling you.

The receipts
ARACHN
11 / 17
Publishing · Arachn's own three channels

The week the rails broke, and the same day they were replaced

Running on X, Instagram and TikTok. The whole operation logged, including the failure.

2026-07-31 12:36:33Z
Live publish test FAILED. TikTok 403 reached_active_user_cap at /v2/post/publish/video/init/, identical to the 2026-07-25 block. Auth, token refresh and creator_info all succeeded. The cap is on the vendor's app.
2026-07-30, in advance
The decision rule, agreed with Mike before the test and recorded as journal prediction P-007: if the test fails, Arachn switches its publishing layer. No debate on the day.
2026-07-31 11:05am CT
Rails switched, and the same day the reserved video PUBLISHED on Arachn's own account, verified by read-back: tiktok.com/@spydyrs/video/7668724012360437023
2026-07-31 5:25pm CT
On Mike's word (“Yes swap, new post with new audio”) the post was replaced with the final audio bed. Old post deleted. Verified closed 2026-08-03.
2026-08-03 9:09pm CT
Next week scheduled on Mike's named approval, all three channels, staggered: X 10:05, Instagram 10:40, TikTok 11:10, Wednesday 2026-08-05. All three verified PENDING by read-back.
2026-08-03
The same piece published as a permanent page: arachn.com/thread/the-tool-bought-to-fix-the-last-two-tools
Source: arachn/social/SOCIAL-LOG.md, every action logged with its date. Vendor names, account ids and post ids held back; the live URLs are public.
Social media management
DEMONSTRATED

We know we should be posting. It has been three weeks. Nobody has time to make anything, and when we finally do, it does not sound like us.

This runs on Arachn's own three channels every week since 2026-07-25, with live posts you can open on X, Instagram and TikTok, and the whole operation logged, including the week our publishing vendor's TikTok access failed permanently, a pre-written decision rule fired, and we switched rails and published the same day.

The receipts
ARACHN
12 / 17
Webster Reed · live since 2026-07-22

The useful half of the receipt is the failure

On its own infrastructure, on a schedule, with a person in front of every outward step. The week published on time anyway.

The scheduled job, read back today
arachn-social-weekly-draft0 9 * * 1
enabledTrue, scheduled
last run 2026-08-03error
next run2026-08-10T09:00
Why it errored
“Skipped to prevent unintended spend: global inference config drifted since this job was created (model 'claude-opus-4-8' -> 'claude-sonnet-5'), and this job is unpinned. No inference call was made.”
The guard did what it was built to do: no inference call was made.
What he produced anyway, and who let it out
2026-08-03 · buildsocial-weekly-2026-08-12-dryrun
“Dry run of the weekly social pipeline (Aug 12 calendar row) exercised end to end: beats written, cards rendered and visually verified, TikTok video assembled with the audio bed and verified…”
2026-08-03 · reviewsocial-weekly-2026-08-12
“Mike approved all three channels (X, Instagram, TikTok) for the Aug 12 calendar row post, to publish Wednesday Aug 5. Assets and captions staged; scheduling handoff…”
gate outward by mike verdict approved
What the guard set off, in order, UTC
the review polish job stops on the same guard2026-08-03 12:40Z
the weekly social draft stops on the guard, nothing published2026-08-03 14:00Z
Webster runs the pipeline by hand instead, one real defect found and fixed2026-08-03 18:12Z
Mike's outward gate: approved, all three channels2026-08-03 18:32Z
Aug 5 post scheduled to all three channels, staggered…2026-08-04 02:10Z
Sources: /home/hermes/.hermes/cron/jobs.json (droplet) · arachn/brain/ops-mirror/records/runlog/2026-08.jsonl · arachn/social/SOCIAL-LOG.md. Job and account ids removed from the quoted error.
AI agent development
DEMONSTRATED

I want something that just does the thing every week without me remembering. But I am not letting a bot email my customers.

Webster Reed, live since 2026-07-22 on its own infrastructure, running twice a day on a schedule. It drafts Arachn's weekly social package, renders the cards, stages the assets and sends the approval to a phone. It has never published without a yes because it structurally cannot. The more useful half of the receipt is the failure: on 2026-08-03 its scheduled job errored on a model-drift guard, the guard did what it was built to do rather than shipping something wrong, and the week published on time anyway.

The receipts
ARACHN
13 / 17
Unattended · read 2026-08-04

Twelve things run here without being asked

Four automation workflows on our own server, four scheduled jobs on the assistant's machine, four on the founder's.

4 · n8n, the server
$ n8n list:workflow --active=true
W2 Pulse: checks every 30 minutesactive
W2 Pulse: Monday proof-of-lifeactive
W1 Work records: sweep and nightly summaryactive
W3 Weekly review: Sunday assemble, Monday sendactive
4 · the assistant's machine
.hermes/cron/jobs.json
social-weekly-draft0 9 * * 1
social-monthly-calendar0 9 25 * *
weekly-review-reminder0 8 * * 1
weekly-review-polish40 7 * * 1
4 · the founder's machine
$ launchctl list | grep arachn
campaign-watch00:15 07:15 12:15 18:15
morning-sessionSun-Thu 06:45
ops-mirror22:40
spend-snapshot08:07 22:30
The pulse, every 30 minutes, no model involved
2026-08-05T01:30:00Z OK gateway spend box expiries jobs 2026-08-05T02:00:00Z OK gateway spend box expiries jobs 2026-08-05T02:30:00Z OK gateway spend box expiries jobs 2026-08-05T03:00:00Z OK gateway spend box expiries jobs 2026-08-05T03:30:00Z OK gateway spend box expiries jobs
What it caught on its own
2026-08-03 01:30Z · snapshot-stale
“PULSE: the spend snapshot is 46 hours old (expected daily). Money checks are running on old numbers. Fix: check the Mac's nightly job.”
Last run status, read back today: social-weekly-draft error · social-monthly-calendar ok · weekly-review-reminder ok · weekly-review-polish error. A failure is a receipt too: a guard stopped the weekly draft rather than let it publish on a model it was not built for.
Sources: n8n list:workflow --active=true (droplet, hermes) · /home/hermes/.hermes/cron/jobs.json · ~/Library/LaunchAgents/com.arachn.*.plist · arachn/brain/ops-mirror/pulse/checks.log and incidents.md, read 2026-08-04
Workflow automation
DEMONSTRATED

Half my week is stuff a machine should be doing. Every time we automate something it quietly breaks and nobody finds out for a month.

Twelve things run unattended here right now: four automation workflows on our own server, four scheduled jobs on the assistant's machine, and four on the founder's. A model-free pulse checks the machinery every thirty minutes and sends a proof-of-life note every Monday, so a quiet week is proof of nothing. The system has caught its own failures: a guard stopped a scheduled job rather than let it publish something wrong, and a spend watcher stands ready to force-stop an ad campaign that runs past its approved window. All of it is logged.

The receipts
ARACHN
14 / 17
Observed · baseline 2026-08-04

Cited nowhere, on all five surfaces

The citation baseline for arachn.com across Perplexity, ChatGPT, Google AI Mode, Google AI Overviews and Gemini, run on our own account.

$ citation_monitor.py --replay <run> --domain arachn.com
replayed 14 saved checks against arachn.com platform checks capable cited Perplexity 3 3 0 ChatGPT 2 1 0 Google AI Mode 3 3 0 Google AI Overviews 3 3 0 Gemini 3 3 0 TOTAL 14 13 0
The questions asked
“Who is Arachn?”5 surfaces
“What agencies do GEO?”5 surfaces
“Who can help me show up in ChatGPT?”4 surfaces
What the machine says instead · Perplexity
“Arachn refers to Arachné (Arachne) in Greek mythology, the mortal weaver who challenged the goddess Minerva and was transformed into a spider. There…”
Asked again, disambiguated · Perplexity
“Arachn (Agency Consultancy Lab) appears to be ArachnidWorks, a marketing and web/brand services firm that positions itself as blending strategy and execution for…”
Fourteen checks, five surfaces, zero citations of arachn.com. The baseline is not a low score, it is an absence, and the machine fills it with a different agency.
Sources: ~/.arachn/citation-monitor/2026-08-04T13-32-15Z.json (14 checks) and latest.json (“Who is Arachn (Agency Consultancy Lab)?”), replayed with bin/arachn-ops/citation_monitor.py. Replay costs nothing and touches no network.
AI visibility audit
DEMONSTRATED

Before I spend a dollar on any of this, tell me where we actually stand.

This exact deliverable, performed on arachn.com on 2026-08-03 and published rather than buried. Findings on ourselves: every machine file returning 404, zero structured data, not one heading on the page, roughly 3,400 characters of machine-readable text of which the brand content was about a line and the rest was the privacy policy, and non-branded queries returning four other companies and never us.

The receipts
ARACHN
15 / 17
The operating layer · live since 2026-07-31

The base runs, and it keeps score against itself

Scheduled jobs, a work-records pipeline, a nightly summary, and a ledger that records the predictions it got wrong.

The operating layer's own record, assembled nightly
month 2026-08 events 27 across 13 jobs by kind check 5 build 13 review 2 research 2 publish 4 admin 1 by actor architect-seat 24 webster 2 mike 1 gates 7 passed spend DigitalOcean $1.66 month to date rebuilt 2026-08-05T02:50:00Z (nightly, unattended)
The jobs it swept, this month
agency-roster · arachn-social · capabilities-deck · media-direction · media-direction-d9 · media-direction-observed-leg · morning-session · offering-source · pulse · social-weekly-2026-08-05 · social-weekly-2026-08-12 · social-weekly-2026-08-12-dryrun · thread-section
The predictions ledger, written before the fact
4right
1wrong
1overtaken
7open
The wrong one, in full
P-007, 55%: a vendor said publishing worked again, so we predicted the test would pass. OUTCOME: WRONG. “The 55% was too generous to a vendor signal with no verifiable artifact behind it.”
What it found broken in itself
“FIXED 2026-08-04. Every one of the 108 agents had a fake model name (model: craft on 80, judgment on 26, mechanical on 2)” Every agent would have died on launch. STANDING RULE: never write a model name into an agent file.
Nothing has been sold. That is the honest headline. Everything above is capability, and the copy-and-fit motion has never been run for anybody. The base is inventory; the engagement is unproven.
Sources: arachn/brain/ops-mirror/records/SUMMARY.json · arachn/brain/research-journal/PREDICTIONS.md · arachn/brain/STATE-OF-THE-BUILD.md, read 2026-08-04
Your own Arachn, built in-house
BUILT

I want the whole thing in-house. My people, my systems, not a retainer forever.

The base exists and runs: 108 roles, an operating layer live since 2026-07-31, twelve scheduled jobs verified running, a work-records pipeline, a model-free pulse checking the machinery every thirty minutes with a standing proof-of-life note, and a weekly review that assembles itself on Sunday and reports Monday morning unattended. Total infrastructure cost for the month so far, metered nightly: about ten dollars. What has never happened is the copy-and-fit motion, for anybody. The base is inventory. The engagement is unproven.

The receipts
ARACHN
16 / 17

An agency, a consultancy and a lab, genuinely built and run in partnership with intelligence rather than with intelligence bolted onto the side. That operating model is the product.

It is available whenever you are ready, and to whatever degree you want it in your own house: run by us, designed for your people, or built and handed over. Inspect it before you buy any of it.

This document was written, designed and built by Arachn's intelligence, produced by our executive producer Webster Reed, and judged, gated and approved by its founder.

That sentence is our capabilities statement.

Weave. Wise. Webs.

Agency · Consultancy · Lab

Thank you.
ARACHN
17 / 17