Every claim on this site has a URL under it. Below is the full list of what this offer has put online — pages, packages, articles, pull requests — so you can check the work before talking to me. Nothing here is a promise: each line is a live link.
Three packages (Sprint $900, Engine $1900/mo, Full build $2900), FAQ, schema.org Service/Offer.
What is inside the retainer and what is explicitly not — one page you can forward to a client.
Agency vs freelancer vs marketing engineer, line by line, with the cost model behind each column.
Twelve artifacts from this campaign with their live links — the fastest way to judge the work.
A roster is not one funnel: every client carries its own floor and read-out date (13,914 per arm; 27,828 in total), a client at 2,000 contacts a month needs 14 months, and the smallest client readable inside 6 months is 4,638 a month.
Two funnels that must be readable at the same time, liquidity as the readable unit, why each side needs its own floor (13,914 per arm; 27,828 in total), and the four decisions taken before the first contact.
Six questions and what each answer really measures, the volume a verdict needs (13,914 contacts per arm; 27,828 in total; 56 days at 500 contacts a day), and the four things the interview cannot decide.
The four definitions — lead, source, reply, volume — that have to match before a B2B channel can be judged, the five things that exist before the first send, the contact floor an honest test needs, and the sending ramp. Also on dev.to.
The five numbers most SaaS plans are missing, where the stack stalls, and what ships in the first engagement.
The buying decision written out: who owns the number, how the work is checked, and six questions whose answers you can verify before paying a retainer.
Five steps and the artifact types that replace a portfolio of screenshots — written the same way the offer pages are: every claim openable.
The seven deliverables in one list, and the receipt behind each one — pages, packages, repositories, free tools.
The machinery, not the deck: agent skills, quotable pages, automation and kill rules — and what can be opened before paying.
One path built, instrumented and measured — the role, the packages and the test that separates an engineer from a marketer.
Sprint $900, Engine $1,900/month, Full build $2,900 — and the comparison with a freelancer, an agency and a hire.
Wilson interval plus the rule of three — the floor under “it is working”.
What a done-for-you pipeline actually contains, and where the handoffs are.
The four signals that separate an engineer from a channel operator.
Publishing is half the job; the other half is making the published work get read.
Getting quoted by AI answers: structure, schema, and the crawl paths that matter.
Where an agency earns its margin and where an engineer removes it.
The part-time seat: what the fixed monthly block contains and how it differs from a fractional CMO.
Nine checks in the order that removes risk — authentication, tracking domains, root-domain folding, MX, address-to-domain match, role rejection and send caps — with the funnel of a fifth source family measured in the same pass.
The operating mindset as an installable agent skill — npx marketing-mindset.
The same skill exposed as MCP tools: section index, test-size math, kill rule.
Enter a reply rate and the lift you care about — it returns the volume a test needs before a verdict.
SKILL.md, AGENTS.md, Q&A pages — MIT, on GitHub, release v1.0.4 tagged.
A playbook is the second step, never the first: proven moves only, ~6 working hypotheses before you can hire marketing, failed hypotheses archived in one line. New CLI: npx marketing-mindset playbook --hypotheses 3.
The Q&A page behind the release: what goes in, why it is step two, and why nobody should inherit your failures.
Test-size and channel questions answered as standalone pages.
What a marketing engineer does that an agency invoice does not cover.
Two submissions stopped being pull requests and became entries in repositories I do not own.
Not just opened: accepted. The maintainer reformatted PR #454 himself and committed the row (ec4927a0, 18 Sept 2026). Read back from the raw README, line 442 — HTTP 200.
PR #2477 merged (f846ecc0, 18 Sept 2026) into someone else's index; the profile page answers 200. Listed by their crawler, not by me.
Pull requests and registry entries, each one checked with a HTTP 200 on 19 September 2026 before it was written into this page. A link that stops returning 200 comes off.
TAPAC as a remote MCP server for verified B2B contacts — filed 18 Sept 2026.
The contact finder plus SMTP verifier, submitted to the list most MCP clients read first.
The remote-MCP variant, where no local install is required to call it.
Agent-native contact search — the entry an agent stack reads, not a human.
The same server for the Chinese-language list — submission date and title are visible in the repo.
marketing-mindset as an open agent skill for B2B marketing tests.
The skill submitted where agent builders actually look for skills.
marketing-mindset into the Hermes agent community list.
The analytics SDK for the GymCam product, version read from the PyPI JSON API.
The duplication-detection SDK — same check, same registry.
The contact engine as a library: find, verify, return.
Read on 19 September 2026 from the registry or the platform's own API. When a figure moves, the source moves with it, and this line changes.
The sample-floor skill, one line to install: npx skills add axelfreeman/marketing-mindset. Version read from the npm registry, not from memory.
The same floors exposed as an MCP server, so an agent can query the required volume itself instead of guessing it.
The B2B contact engine as one MCP tool: find, verify, return. Pay-per-use at $0.10–0.50 a contact, key at tapacapi.com/get-key.
Three published SDKs. Versions read from the PyPI JSON API on the same date.
The engine's published number, and the mechanism behind it: addresses are SMTP-checked before they enter a send queue, so a dead address is a row you delete, not a bounce you explain.
A posting engine runs a three-product rotation across 17 agent networks. Counters read from the engine's own state file on 19 September 2026.
MX records were resolved domain by domain: 353 accept mail, 28 do not and were flagged do not send instead of being mailed into the void. The queue ships with the flag in the row.
Before writing in a server, its invite API is queried: n8n 87,537 members · Design Buddies 104,078 · SillyTavern 80,035 · openrouter 53,799 · Cursor 39,279. A room of 1,071 is a different decision from a room of 104,078.
Open the link. If a page 404s, or a package does not resolve, that is a fact about my work and you should treat it that way. The same applies to numbers: where a figure appears on this site, it comes from a public source — a registry, a crawl index, or a report generated from my own analytics.
Six source families counted on a 9,266-domain working file, with the share that publish an address on their own domain: application marketplaces 0% at intake and 56% after walking 300 contact pages, agency directories ~60%, package registries ~15%, repository homepages ~11%.
Marketing engineer for startups — the three routes a seed-stage team compares, what the first 30 days produce, and the floors the numbers have to clear.
Marketing engineer for ecommerce — why last-click revenue is not a verdict, the spend gate, and the traffic floors a store test has to clear.
Marketing engineer vs growth marketer — what each role owns by default, the first readable result, and four signs that decide the hire.
Marketing engineer for B2B — the lead, source and reply definitions that decide whether a channel can be judged, the five things that exist before the first send, and the floor an honest test needs.
Marketing engineer job description — the four decisions and six lines to write instead of a tool list, the artifact due each week, the volume a strict test needs, and a free generator.
Marketing engineer retainer — the invoice line by line, the four decisions a retainer cannot make for you, and how many honest tests a month can actually hold at your volume.
Marketing engineer for fintech — why the readable unit sits after identity and funding, the volume a verdict needs, and the stop rule that has to exist before the spend.
Marketing engineer for devtools — why the readable unit is the first successful call rather than the signup, the volume a verdict needs, and the stop rule that has to exist before the spend.
Marketing engineer interview questions — the six questions that separate an engineer from a marketer with tools, what each answer really measures, the volume a verdict needs (13,914 contacts per arm; 27,828 in total), and the four things the interview cannot decide.
Marketing engineer for marketplaces — two funnels that must be readable at the same time, liquidity as the readable unit, why each side needs its own floor (13,914 per arm; 27,828 in total; 70 months for a 400-seller-a-month supply side) and the four decisions taken before the first contact.
Marketing engineer for agencies — an agency cannot pool its clients' traffic, so every client needs its own floor and its own read-out date: 13,914 per arm (27,828 total), 14 months for a client at 2,000 contacts a month, and 4,638 contacts a month as the smallest client readable inside 6 months.
Local B2B