Big Idea

The argument in one line.

Cold-outbound lead generation has shifted from single-source tools and rigid no-code workflows to AI-orchestrated waterfalls across data sources, self-correcting scheduled agents, and free open-source replacements for expensive scraping and enrichment tools.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You run or manage cold email / cold outbound campaigns and currently rely on one list-building tool (LinkedIn scraper, Google Maps scraper, or a single AI directory).
You've hit a wall finding contacts for companies with no LinkedIn presence or only a generic website email.
You already use Claude Code or Codex for other tasks and want to see a concrete GTM/sales-ops application.
You're paying for enrichment tools like Clay, ZenRows, or BuiltWith and want to know what a self-hosted alternative looks like.

SKIP IF…

You don't do outbound sales or lead generation in any form.
You want a no-code, point-and-click tool walkthrough - this assumes comfort directing an AI coding agent.

TL;DR

The full version, fast.

Claude Code and Codex have replaced several single-purpose lead-gen tools with orchestration: instead of picking one company database (LinkedIn-based, Google Maps-based, or AI-search-based), an agent waterfalls across all three, starting with the highest-fidelity source and using the expensive AI directories only to fill gaps. The same shift applies to contact finding, where custom scrapers plus tools like Parallel.ai and Open Web Ninja now surface people with no LinkedIn footprint. Scheduled tasks become self-correcting agentic cron jobs instead of brittle no-code workflows, and goal mode lets you define a plan and pass/fail criteria up front rather than hand-holding execution. Campaign optimization itself becomes a daily auto-research loop that studies replies and ships improvements. Underneath all of it, free open-source tools (Browser Use, Gemma 4, HTML-to-text libraries, simple tech-detection scripts) replace paid equivalents like Clay's scraper, ZenRows, and BuiltWith, cutting the cost of running the whole operation.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 01:36

01 · Cold open + agenda

States the thesis that lead gen has changed since February 2026, claims 200-300 daily positive responses, previews the six shifts and promises to open-source the skills.

01:36 – 03:17

02 · Change 1: Company waterfall sourcing

Explains the three tiers of company-list tools (LinkedIn-based, Google Maps-based, AI-search-based) and how Claude Code orchestrates a waterfall across them, starting cheap and filling gaps with AI directories.

03:17 – 04:33

03 · Change 2: Contact finding

Old binary choice (LinkedIn lookup vs. scraped generic email) replaced by AI-routed contact discovery; live demo finding a gym owner's name via Open Web Ninja.

04:33 – 06:29

04 · Change 3: Agentic cron jobs

Scheduled tasks become self-correcting agents (OpenClaw, Hermes) instead of brittle no-code workflows; used for daily prospecting and for reply-rate/bounce-rate campaign health checks.

06:29 – 07:36

05 · Change 4: Goal mode

Define a plan with clarifying questions, then set explicit pass/fail completion criteria before execution - applied to both list building and campaign setup.

07:36 – 08:51

06 · Change 5: AI auto-research on campaigns

A recurring loop analyzes reply data (job title, industry, company size, messaging) and proposes campaign changes automatically, replacing manual analysis.

08:51 – 10:00

07 · Change 6: The open-source stack

Free/self-hosted replacements for paid tools: Browser Use for browser automation (demoed on a live sales-tax compliance check), Gemma 4 for local custom-variable generation, open-source HTML-to-text libraries, and simple self-hosted tech detection.

10:00 – 10:13

08 · Sign-off

Recaps that a lot of this wasn't possible in February, points to the skills linked in the description, and closes.

Atomic Insights

Lines worth screenshotting.

Company list-building tools split into three tiers - LinkedIn-based, Google Maps-based, and AI-search-based - and waterfalling across all three beats picking just one.
AI-search list-building tools like Exa.ai and ocean.io are the most expensive option, so the efficient play is starting with cheaper high-fidelity data and using AI tools only to fill gaps.
A marketing agency with 10-200 employees is a LinkedIn-list problem; an HVAC company with 200 Google reviews is a Google-Maps-list problem - the right source depends on how the target market self-identifies.
Contact finding used to force a binary choice: a LinkedIn profile or a scraped generic website email, with no middle ground for people who had neither.
Running the same search through OpenWebNinja instead of a plain Google search surfaced a business owner's name that a standard AI overview search missed.
A cron job used to require pre-built workflows that broke on any edge case; an agentic cron job can see a new edge case and solve it instead of erroring out.
Agentic cron jobs are now used for daily reply-rate and bounce-rate checks on outbound campaigns, replacing manual daily monitoring with a Slack-delivered report.
Goal mode means defining pass/fail completion criteria before execution, rather than hoping the agent interprets the task correctly along the way.
A practical goal-mode routine: dump context, have the agent ask 10-15 clarifying questions, lock a plan, then define what done looks like before starting.
AI auto-research on a campaign can analyze reply data by job title, industry, and company size and ship suggested changes without the operator reading through the raw data first.
Browser Use, an open-source browser automation harness, can walk a checkout flow all the way to the cart to check whether a store is charging sales tax correctly in a given state.
Google's Gemma 4 12B open-weight model, released around June 3, 2026, is being used locally to generate custom outreach variables after local tuning.
Paid scrape integrations (like Clay's or ZenRows) can often be replaced by an open-source HTML-to-text library feeding the same content into an AI model at near-zero cost.
Detecting what technology a company's website runs is just pinging the site and checking the HTML for known code signatures - a task that doesn't require paying for a tool like BuiltWith.

Takeaway

AI turned single-source lead lists into orchestrated waterfalls.

WHAT TO LEARN

The gains described here come from directing an AI agent to combine multiple existing data sources and self-correct on schedule, not from any single new tool.

Splitting company databases into LinkedIn-based, Google Maps-based, and AI-search-based tiers, and querying cheap/high-fidelity sources first, avoids overpaying for AI-search directories on every lookup.
A target market's own conventions (does it self-identify on LinkedIn, or show up on Google Maps instead) should decide which data source to start from, not habit.
Contact discovery no longer has to stop at no-LinkedIn-profile-no-obvious-email - routing the same query through multiple contact-finding tools surfaces people that a single source misses.
A scheduled task that used to break on edge cases can instead be handed to an agent that recognizes the edge case and adapts, cutting the maintenance burden of automation.
Defining pass/fail completion criteria before an agent starts a multi-step task (not just instructions) reduces the chance of a technically-finished-but-wrong result.
Recurring analysis of reply data by job title, industry, and company size can be delegated to a scheduled AI loop that ships a report, instead of requiring a person to manually mine the data.
Several previously paid tools (site scraping, technology detection, campaign-variable writing) now have viable free or self-hosted equivalents, which changes the cost floor of running outbound at scale.

Glossary

Terms worth knowing.

Company waterfall sourcing: Running a company list-building search across multiple data sources in sequence - starting with the cheapest, highest-fidelity source and falling back to more expensive AI-search directories only to fill gaps.
Agentic cron job: A scheduled task handed to an AI agent instead of a fixed no-code workflow, so the agent can adapt to edge cases and errors instead of failing silently.
Goal mode: A way of directing an AI agent by defining explicit pass/fail completion criteria up front, rather than only giving step-by-step instructions.
AI list-building tools: Search platforms (e.g. Exa.ai, Parallel.ai, ocean.io, Disco) that have crawled and indexed large portions of the open web to find companies or contacts not present in LinkedIn or Google Maps data.
Technology finder: A tool or script that detects what software or platforms a website runs by scanning its HTML for known code signatures.
Open-weight model: An AI model whose trained parameters are published for anyone to download and run locally, rather than being accessible only through a paid API.

Resources

Things they pointed at.

01:40toolBlitz API

01:40toolClay ↗

01:40toolProspeo

01:33toolGoogle Maps

01:33toolOpenMart

01:33toolD7 Lead Finder

01:45toolExa.ai

01:45toolParallel.ai

01:45toolDisco

01:45toolocean.io

03:51toolOpen Web Ninja

04:33toolOpenClaw agent

04:33toolHermes agent

08:07toolBrowser Use

08:49toolGoogle Gemma 4 (12B)

09:13toolZenRows

00:15book$100M Leads (Alex Hormozi)

00:00linkCold outbound skills repo ↗

Quotables

Lines you could clip.

00:02

“Lead generation in 2026 looks absolutely nothing like what lead generation looked like in February 2026.”

tight, stat-free opening claim that frames urgency→ TikTok hook↗ Tweet quote

00:25

“We're using these tools to generate anywhere from 200 to 300 positive responses per day on behalf of our customers with just cold email.”

concrete, specific claim with a number→ IG reel cold open↗ Tweet quote

04:34

“Because they're agentic, they'll be able to see the problem and then solve the problem for you.”

clean one-line definition of the agentic-vs-workflow distinction→ newsletter pull-quote↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphorstory

00:00Lead generation in 2026 looks absolutely nothing like what lead generation looked like in February 2026. The space is moving so fast with quad code and codecs that it's really difficult to keep up.

00:11And so if you're a person who's still manually clicking around and building workflows, I promise you there's a better way to do it. We're using these tools to generate anywhere from 200 to 300 positive responses per day on behalf of our customers with just cold email. In this video, we're gonna go over six shifts that we've seen with Cloud Code and Codex so that you can implement that in your business as well.

00:31I'll be giving away all the skills and open sourcing everything. Let's jump in. So I've made this little presentation for you guys, and I think the six shifts come down to the ability to waterfall company sourcing, which we'll talk about why that's so important, to waterfall contact finding outside of just basic LinkedIn scraping.

00:47Number three is the ability to add agentic cron jobs. Number four is goal mode. Number five is the ability to use auto research.

00:57And then number six is the ability that ClaudeCode and Codex give you to access open source technology. So the first thing that I mean by company waterfall sourcing is that, in my opinion, there are three different kinds of tools that help you build a list of companies. You have tools that live off of LinkedIn scraping, tools that are based on Google Maps scraping, and then AI search tools.

01:16These can be bucketed into tools like Blitz API, Clay, and Prospio. Their company database and their their contact database is based on LinkedIn, which for a lot of use cases is great. It's really high fidelity data.

01:29It's the the basis for a lot of these tools, but sometimes it could be missing things. Specifically, when you're looking for local businesses, which then you need a Google Maps based list.

01:37You could scrape Google Maps directly. You could use platforms like OpenMart or d seven Lead Finder. And then finally, you have AI list building tools like exit.ai, Parallel dot a I, DiscoLike, and ocean.io.

01:48And so what those tools are, they're probably the most expensive tools that you can use, but they've scraped the entire Internet and they have a directory so that if you're looking for really, really niche things and they're not really on Google Maps and they're not really on LinkedIn, but they have a website, you can get them there.

02:02And so what we've done in the last three months is now with Claude Code, we're orchestrating being able to start with the highest fidelity data that we possibly can, and then cleaning all of that up with AI, and then moving to the AI company builders. The AI list building tools are absolutely amazing, but they're just really expensive if you wanna get the entire list of every company that you wanna reach out to in your market.

02:23So we start with our LinkedIn based database tools, or we start with the Google Maps based database tools, depending on the list that we're trying to build. So if you're looking for marketing agencies with 10 to 200 employees, totally a LinkedIn list. If you're looking for HVAC companies that have 200 Google reviews, totally a Google Maps list.

02:43And then in either situation, because of the way that people self identify their industries on LinkedIn and the way that not every company has a Google Maps listing, then as long as you can target somebody with a website, ocean.io, DiscoLike, Parallel, Exa do a really great job of filling in the gaps.

02:59Being able to get in one shot a really awesome company list because you start with the high fidelity data and then cram in the AI to fill in the gaps, just an absolutely amazing game changer by just using your voice and using the skill file that I'll give away in the description as well too. Change number two is contact finding.

03:16Now in the past, what you were basically limited to was if you could find a contact on LinkedIn, you would be able to reach out to them, which usually is great, but everyone is basically using LinkedIn. And then if you could find an email on the website and then use AI to pull out the name from the email, you might be able to do that.

03:31Or you just scrape any email off the website, that was your only other option. You were stuck between both of those things. You didn't have a good middle ground.

03:38So you didn't have a way to find contacts even though they didn't have a LinkedIn profile, and you didn't wanna get stuck with the generic email finder on the website. Now, you can orchestrate with Cloud Code using custom scrapers, and then, again, parallel.ai, exit.ai are not only great for finding companies, they're also phenomenal for finding contacts.

03:56But then you can use Open Web Ninja as well to find some of these things. If you were to just Google for the owner of E and A Paramus, which is a gymnastics gym that I used to work at, it's Craig Zappa with the AI overview in here. And so then if you use OpenWebNinja to run the same exact search in here, see how we also get Craig Zappa is the owner of USAG Elite, owner of ENA Paramis.

04:15So now we are not limited to just they have a LinkedIn profile or they have a generic email on the website. Now we can use AI to plug in the gaps on this contact finding as well too, which is our second biggest game changer. The third game changer that I would say AI brings is the ability to have agentic cron jobs.

04:33And so if you don't know what a cron job is, it is just basically a scheduled task. And if you wanted to set these things up in the past, it used to be you had to set up a workflow, and then connecting it to AI, it had to run on prompts that you made beforehand. But if there was an edge case, it didn't really work and you had to build the workflow manually and just really annoying to set up.

04:49Now I can literally go to my OpenClaw agent or my Hermes agent and just say, hey, every day, can you find me a list of at least five companies that are hiring for their first go to market engineer, connect to exa.ai or parallel.ai in order to get this information, or do your own searching, whatever it might be.

05:08When you built these workflows in the past, you might have edge cases that cause problems that then you would just get errors and you'd have to fix it for all the edge cases and all this. But because they're agentic, they'll be able to see the problem and then solve the problem for you. You can use these agentic cron jobs to obviously do more things.

05:23But if you wanted just a really great list of, hey, I'm gonna reach out to 500 people per day automatically, but I wanna reach out to five people manually that are super good fits, awesome job for an agentic cron job. We're also using agentic cron jobs to be able to run reply rate analysis, bounce rate checks, the daily checks that we would usually be doing on all of our campaigns.

05:44Now it's just a Slack message away to be able to set up a daily, hourly, a fifteen minute check during the day. Absolute game changer over here as well too. For go to market, I do not see enough people talking about goal mode as well.

05:57With goal mode, we are using that to do really great list building. So like I talked about here, when we're doing the company waterfall sourcing up here with Claude Code, we have a skill, but we will outline and we will say, hey, you know, want you to use all of these tools and you're not done until you've checked all of these databases, compiled it into a CSV, normalized all the columns, etcetera, etcetera.

06:18Keep working until you get all of that done. Goal mode is really great for setting that up. Goal mode is also really great for setting up campaigns.

06:24When you say, hey, these are the inboxes that you need to attach to the campaign. This is the messaging. This is the list.

06:30And here's the custom variables that you need to have happen. You need to be able to audit them and verify them and all those other things. Goal mode is just absolutely amazing to be able to set these things up.

06:38What I do is I set up a plan and I just dump a bunch of context into Claude code, and then I say ask me 10 to 15 clarifying questions to just make sure we get the context really right. Then that plan gets set in, and then I say, okay, create a pass fail for the goal being completed and what that criteria is based on my plan.

06:56I review that. It's all set up. Ready to rock and roll.

06:59Change five, I would say is definitely the fact that you can now use AI to kick off an auto research loop on your campaign. Now instead of you analyzing all of the data and coming up with all of the insights, you can see what are the prospects that are most likely to respond. What's their job title?

07:14What industry are they in? How many employees do they have? What's the messaging that they responded with?

07:18What's the messaging that they send back when we send them whatever messaging? All these things can now just be done on a loop on a daily or a weekly basis so you could get these insights and then have them presented to you instead of you having to collect the data yourself, read through the data, and analyze it all completely.

07:33You can now just get a report sent to you and then you could spot check the data afterwards. And then the final thing that I would say for the changes is the ability for you to get access to open source technology. So not only am I gonna give away all the skills that we have for this video, I've got the company waterfall sourcing skill, I've got the contact finding skill.

07:50The AI auto research on your campaigns is a skill that we have. These skills are being open sourced by creators and companies and all these things. And that these tech companies that you can get access to amazing technology for absolutely free is just mind blowing.

08:03So for example, Browser Use Harness is absolutely one of my favorite open source repositories right now that lets you use browser automations to move through websites. You can do research on websites.

08:14I even used it over here. We set up a campaign, you can see right here, that we even took a product all the way to the shopping cart and then tested to see where they're collecting sales tax in and if they're properly collecting sales tax. So here, they should be collecting sales tax in California, but you can tell that they're not.

08:29And so now we can email them and say, hey, you're not collecting sales tax properly. You need to do it better, blah blah blah. And you can change that up.

08:35And so browser use makes that possible. The open source models are getting so freaking good as well too. So for one, I am not a proponent of using a Chinese model.

08:44That's just never gonna happen. I'm never touching those models. So when I saw that Google came out with their Gemma four line of open source models, I was extremely excited.

08:53Specifically, the 12,000,000,000 parameter one that they just came out with, I think it was June 3. I started using it locally to write custom variables for our email campaigns, and that's working really well. I had to tune it a little bit locally, but AI makes this stuff so easy.

09:07It's so freaking it's just so crazy. HTML to text. I used to pay hundreds of dollars a month for either the scrape website integration in clay or if we were using Zenrose.

09:17And so Zenrose is an amazing scraping tool, but we the use case that we were using it for was quite expensive to literally when I just wanted literally just the text from the home page, really expensive. Now you can just run an open source HTML to text library, get all the content from the website, put it into your AI model.

09:32And then even things like technology finders. A lot of people don't know when you find technology on somebody's website, all is is just pinging the website to look in the HTML for some code that denotes that they have this technology on their website. Super easy thing.

09:45Why do we need to pay for built with or any of these other tools? This is a very, very, very easy thing to do now. I think that these are basically the six main shifts that Claude Code brings to GTM and Cold Outbound in particular.

09:58A lot of this wasn't even possible in February. The cost of data is going down on a lot of things.

10:03I hope you could agree with me that a lot of things are changing. You can check out all the skills that I'm gonna be giving away in the description below. I hope you got value from this, and I'll see you on the next video.

The Hook

The bait, then the rug-pull.

A cold-outbound agency owner claims his stack now produces 200-300 positive email responses a day, then spends ten minutes showing exactly which six workflow shifts got him there.

Frameworks

Named ideas worth stealing.

01:36model

Company Waterfall Sourcing

LinkedIn-based databases (Blitz API, Prospeo, Clay/Apollo)
Google Maps-based databases (Google Maps, OpenMart, D7 Lead Finder)
AI list-building tools (Exa.ai, Parallel.ai, Disco, ocean.io)

Query the cheapest, highest-fidelity company database first, clean it with AI, then fill remaining gaps with the more expensive AI-search directories rather than starting there.

Steal forany B2B list-building process that currently relies on a single data source

01:36list

The Six Shifts

Company waterfall sourcing
Contact finding
Agentic cron jobs
Goal mode
AI auto-research on campaigns
The open-source stack

The video's full agenda - six discrete ways Claude Code/Codex changed his agency's cold-outbound operations.

Steal forstructuring any what-changed explainer video

CTA Breakdown

How they asked for the click.

VERBAL ASK

10:00link

“You can check out all the skills that I'm gonna be giving away in the description below.”

Soft, single closing CTA pointing to a free GitHub skills repo plus a free-campaign application form and a Clay affiliate link in the description - no hard pitch on-screen.

MENTIONED ON CAMERA

01:40toolClay ↗

00:00linkCold outbound skills repo ↗

FROM THE DESCRIPTION

AFFILIATECommission earned if you click.