GTM as code: stop rebuilding outbound lists by hand

The old way to build an outbound list was: open Apollo, build a filter, export CSV, dedupe in a spreadsheet, paste into a sequencer. The whole loop was clicks. In 2026, the growth teams we work with are doing it differently. They define the audience once, run it as a play, and let Claude Code adjust it the next time the market moves. The list isn’t a file. It’s a repeatable process.

What a play actually is

A play on Verbiflow is a short, repeatable pipeline that builds an audience and pushes it into a sequence. It has four parts:

• A config. One JSON file declaring the schema, the columns, and the views.
• Sources you bring. Local scripts that pull from wherever your data lives: Clay, Apollo, Apify actors, SERP, your own scraper, a Chrome session in your own browser. We don’t own the data layer. We make it easy to connect.
• A pipeline. A shell script that runs the steps in order. It can stop, restart, and run again, with all state in SQLite.
• An output. A clean audience, pushed into a Verbiflow sequence where the actual sending happens: connected mailboxes, deliverability, email and LinkedIn replies, CRM sync, and reporting.

Why this matters for one GTM lead

A growth-stage company doesn’t have a data team. The person running outbound is often the only one who knows where the lists come from and how to refresh them. When that work lives in spreadsheets and exported CSVs, two things happen. It rots fast (the list is stale in 30 days), and it’s impossible to delegate because the logic lives in someone’s head.

Plays fix that because the audience logic and run state live in code. Re-run the play next quarter and you get a fresh list with the same logic. A new GTM lead does not have to read the code by hand; they can ask Claude Code what the play does, change the filters, and rerun it from the saved workflow state.

Plays we’ve already shipped

• SaaS by category. Pulled 1,600 SaaS companies from a funding database and used Claude to analyze each homepage, then sorted by what each company actually sells. 4-step SQLite pipeline.
• Regional security firms. Used Google Maps and Census data to find every physical-security company in 1,935 US cities, classified them with Claude, fetched reviews, found negative reviewers. 7-step pipeline, restartable.
• YC trust atlas. Analyzed every YC company’s trust and security page to find which ones were missing certifications. The customer sequenced the gaps first.
• Company intelligence signals. 16 dimensions of company research (funding, customers, hiring, public signals, compliance posture) on every account in your CSV. Sourced once, queried any way.

None of these are static lists. Each one is a pipeline you re-run when the market moves.

What changes when Claude Code can edit the play

The Verbiflow SDK gives Claude Code the scaffolding to work on a play instead of guessing from scratch. Ask it to add a new source. Ask it to filter by a new signal. Ask it to swap the LLM analysis prompt. It can edit the workflow code, run the play, and show the changed rows or artifacts before anything gets pushed into a sequence.

This isn’t a demo. It’s how customers already extend plays. A change that would have been a half-day of Python becomes a short conversation with Claude Code.

The fastest GTM teams in 2026 aren’t the ones with the most enrichment tools. They’re the ones whose lists are code, and whose plays can be adjusted through Claude Code.

How this fits with sequencing

The play does the audience-building work, with sources you bring. Verbiflow’s core platform handles the work after launch: sequences, connected mailboxes, deliverability, email and LinkedIn replies, CRM sync, and reporting. The two connect with a push into a sequence, or a CSV when you want to review the list first.

We’re not a data provider. We’re the place all your data lands and turns into sent outbound. That distinction matters because it means we never compete with the tools you already use to build lists. We make them more useful by giving them somewhere clean to land.

If you’re the GTM lead at a growing company

The question to ask yourself: can the next person re-run your outbound work without asking you for the exact filter? If the answer is “they have to ask me,” you’re in spreadsheet territory. If the answer is “they ask Claude Code to explain and re-run the play,” you’re in code territory.

Verbiflow is built around the second answer.