🦾Eric Nowoslawski

Founder Growth Engine X | Clay Enterprise Partner

I am a direct response outbound lead generation expert. And by that jumble of words, I mean that I leverage cold email to build predictable revenue pipelines for my clients. 

I leverage over 20 databases to merge Firmographic, Technographic, and Persona data to build highly targeted playbooks with my clients. 

I believe we need to arm sales teams with the knowledge of their total Target Addressable Market and teach them how to score that Market for which companies have the most pain and need their solution. 

The best way to do this is to look at public indicators such as open jobs data, employee growth/attrition, social media growth, web traffic, new product launches, employee sentiment, individual employee tech investment, company tech stacks, etc.

With this data, not only are you only messaging with people that you know have a problem that you can solve, you also are armed  with powerful messaging that cuts through the noise. 

I teach sales teams to embrace technology and multiply what is possible per employee.

Founder

There is actually a limit to cold email reply rates.

Knowing this can hopefully free you from testing things that won't matter.

Let me explain.

I have worked with customers in the past that are obsessed with getting the reply rate as high as possible. They think if they send a good enough email, every single person who could possibly see it is going to respond.

We've run tests that show this just isn't the case.

We had the opportunity to work on behalf of Forbes at Web Summit. We had the list of speakers. We had all their emails. We cold emailed them from infrastructure that had warmed up for four months. The pitch was simple: "This is our booth. We want to interview you for Forbes. Book a meeting." And we sent them a link to Forbes.com.

That only got a 30% reply rate. Basically all positive.

When I say "only" 30% — if I were a speaker at an event and Forbes wanted to interview me, I would think the reply rate would be even higher than that.

We also helped a customer email their own customers. We found data that showed they needed a new product, and we could fix the problem for free.

Still only a 10% reply rate.

I say all this because these two examples are pretty undeniably good offers. And they still did not get 40, 50, 60% reply rates.

When you're sending emails, there are so many reasons why somebody doesn't respond. They didn't even see it. It got mixed up with everything else in their inbox. It's just not the right time for them.

When you've run a couple of experiments and the reply rate is staying pretty steady, I would think more about experiments around lists and experiments around offers versus trying to change the copy to see if an AI personalization or a fancier way of saying something is going to move the needle.

List segmentation and changing your offer matter way more.

What experiments have you run that changed your thinking on reply rates?

Brandon Charleson

AI-powered junkie / Clay & Automation Expert / Cold email strategist / Marketing & Growth Advisor / Headhunter

(TL;DR) As an AI and automation expert, I help companies develop their lead-generation and go-to-market strategy to implement AI/automation workflows which allows teams to scale quickly, save time and capital. Some people call us an "automation wizard" because we just make some magic happen on the backend!

Just like you, I wear many hats in life: Digital Marketer, Executive Recruiter, Director/Filmmaker, Full Stack Web Developer, Air Force Veteran, and major techie embracing the AI revolution with generative AI. But above all, I'm a family man to my wife and kids.

I thrive on helping people and companies succeed. Whether it's working with a startup to get off the ground, helping a parent who lost their job find their next big career, or telling a compelling story through video, I love connecting with people.

As an AI and automation expert, I push the limits of automation to help businesses grow at scale and reduce time spent on mundane, repetitive tasks. If you're a business owner struggling to find that "go-to-market" strategy or if you and your team find yourselves doing the same manual steps day after day, let's chat.

I am also a filmmaker/digital creator/part 107 drone operator and went to LA Film School where I got to work with and be taught by big budget films. Perhaps you have some upcoming video ideas and need help bringing them to life? Let's chat about that too!

My job is to help you and your business succeed. I do the heavy lifting in areas like marketing, video production, and automation. Having worked with Forbes as a Top 100 industry expert in tech, I bring a wealth of experience and knowledge to the table.

If you're building an awesome SaaS product/business and need help with marketing or have a video idea but aren't sure how to get it on paper, let's connect.

I am your AI Automation Consultant, Recruiter, Filmmaker, Video Editor, Script Writer, Website Developer, Creative Director, UI/UX Designer, AI Prompt Engineer, Digital Marketer, and most of all, your teammate to see you and your business succeed!

PS - why list so many titles? Three letters: SEO. 😉

When I'm not working, I enjoy scuba diving 🤿, shooting video/photos 📸, hiking 🏔, high-intensity training, lifting heavy weights, shredding snow on the mountains 🏂, and teaching people how to live their best life through nutrition coaching and CrossFit training. 🏋🏼‍♂️

Recruiting Partner & Owner

TechRecruiterPro is a Forbes.Jobs endorsed and veteran-owned executive search and recruiting partner and NPAworldwide recruiting partner. We consult and help companies find the right candidates and other tech specialists for much needed roles.  

Experience with small VC and seed-funded startups, established SMBs, and Fortune 100 corporations. 

Proven history in the technology space including Frontend, Backend, and Full stack software developers with various modern and legacy tech stacks. 

Specialties include:
- Fintech Web3 (blockchain, crypto, NFT) development roles
- Artificial Intelligence (AI), Prompt Engineering, generative AI and NLP
- Application and Web software development & engineering, IT/Technical roles (Frontend, Backend, Full Stack Developers)
- Cybersecurity, database, quality, and implementation roles
- C-Level Executive and Director roles
- Creative roles
- Consulting roles (including IT Strategy and Business Process Outsourcing (BPO))
- Marketing and advertising roles (programmatic)
- Defense & Space, DoD and cleared professional roles
- Manufacturing
- Technicians
- Accounting

If your company is hiring and struggling to find the right person for the job, reach out to us and we are happy to help. 

Contact info:
brandon@techrecruiterpro.com
(415) 862-1960

Last week I open sourced instantly-cli — 156 commands giving AI agents full control of Instantly.ai.

This week, I'm going bigger.

I just built & open-sourced a new CLI + MCP server for Clay

13 commands

But this changes 𝗲𝘃𝗲𝗿𝘆𝘁𝗵𝗶𝗻𝗴 about how you use Clay.

Here's the shift no one is talking about:

If you use Clay today, you know the workflow. You live 𝗶𝗻𝘀𝗶𝗱𝗲 the table. You're the operator. You build the enrichment steps, you watch the rows populate, you export the CSV, you paste it somewhere else.

You are the middleware.

With clay-gtm-cli, Clay becomes something completely different:

𝗘𝘃𝗲𝗿𝘆 𝘁𝗮𝗯𝗹𝗲 𝗶𝘀 𝗮 𝗰𝗮𝗹𝗹𝗮𝗯𝗹𝗲 𝗳𝘂𝗻𝗰𝘁𝗶𝗼𝗻. 𝗘𝘃𝗲𝗿𝘆 𝗿𝗼𝘄 𝗶𝘀 𝗮𝗻 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻 𝗹𝗼𝗴.

You define the logic once — enrich lead, find email, validate, company lookup — whatever you want. Then your agent calls it like any other tool.

One command:

clay fire enrich-lead --data '{"linkedin_url":"..."}' --wait

That's it. 

The agent fires the webhook. Clay runs every enrichment step. A cloudflared tunnel catches the callback. Structured JSON comes back. No dashboard. No CSV export. No human in the loop.

But here's the part that makes this 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 important:

𝗬𝗼𝘂𝗿 𝗔𝗣𝗜 𝗸𝗲𝘆𝘀 𝗻𝗲𝘃𝗲𝗿 𝗹𝗲𝗮𝘃𝗲 𝗖𝗹𝗮𝘆.

Hubspot, Ocean.io, ContactOut. Clearbit. Instantly.ai. LeadMagic. Prospeo, (all the things..)

All API credentials stay inside Clay's environment.

Your agent only ever sees a webhook URL. That's it. Zero-trust by design.

And Clay's real moat in 2026? 

150+ pre-built integrations behind a single credit system. No signing up for 12 different vendors. No juggling billing. 

One login, one currency, all providers.

GTM is about 𝘀𝗽𝗲𝗲𝗱. Fast sprints. Fast validation loops. Test an ICP Monday, validate the offer Tuesday, pivot Wednesday. Clay + clay-gtm-cli makes that possible because you skip the entire vendor evaluation cycle. 

What your OpenClaw agent can now do autonomously:

➡️ Enrich any lead end-to-end (LinkedIn → email → validation → CRM)
➡️ Batch company lookups across 10+ data providers
➡️ Find and validate emails — only return deliverable
➡️ Log events to Clay for enrichment and routing
➡️ Monitor row limits and auto-reset when you hit 50K
➡️ Run a morning enrichment pipeline on new signups

The stack is now complete:

→ n8n for automations (the plumbing)
→ 𝗠𝗖𝗣 for connectors (Claude/Cursor)
→ 𝗶𝗻𝘀𝘁𝗮𝗻𝘁𝗹𝘆-𝗰𝗹𝗶 for outbound (campaigns, leads, replies)
→ 𝗰𝗹𝗮𝘆-𝗴𝘁𝗺-𝗰𝗹𝗶 for enrichment (Clay as a function platform)

Each layer does what it's best at. Your agent orchestrates all of it.

I wrote the full deep dive — how it works, why this architecture, how to set up OpenClaw with it, & the complete command reference:

Want the full setup — MCP config, OpenClaw agent, Clay table setup, and workflow playbook?

➡️ Connect with me
➡️ Drop a "🧱" or "CLAY ME!" in the comments and I'll DM you the full stack

Let's turn Clay into your agent's enrichment engine. 🤖

I just open-sourced something called OpenShart. 💩

Yes, really. Let me explain.

AI agents are handling your most sensitive data — patient records, financial projections, API keys, M&A strategy. And where does that agent memory live? In a database with the same encryption as your grocery list.

One breach. Everything exposed.

With all these agents that we're standing up daily internally and for our clients, the concern is very real. 

So I created it. I put all the security knowledge I had from working with the Air Force, SOC 2 compliant clients, medical companies I've worked with on HIPAA, and really dove into what could be the next version of agentic security. 

OpenShart fixes this by making agent memory fundamentally unbreachable:

→ Shamir's Secret Sharing splits every memory into K-of-N shards
→ AES-256-GCM encrypts each fragment with its own derived key
→ No single storage location ever holds a complete memory
→ ChainLock temporal sequence locks detect automated extraction
→ Bell-LaPadula mandatory access control — no read up, no write down

It works with OpenClaw 🦞, CrewAI, LangChain, LlamaIndex, AutoGen — any framework where agents touch confidential data.

Four security presets, one line of config:
• Standard — SaaS apps
• Enterprise — regulated industries (SOC2, HIPAA, GDPR)
• Government — classified environments (FIPS 140-2, Bell-LaPadula)
• Classified — TS/SCI, air-gap ready

Zero runtime dependencies. MIT licensed. TypeScript. Node.js crypto only.

The name is intentional. The security is not a joke.

If your business is deploying AI agents with access to proprietary or confidential information — and you're not fragmenting and encrypting that context — you're one breach away from a very bad day.

🔗 https://openshart.dev
💻 https://lnkd.in/gtPfU8k2

I'm open sourcing a CLI that gives AI agents like OpenClaw full control of Instantly.ai.

156 commands. 31 API groups. Free. MIT licensed.

Here's why this matters:

I built the n8n integration for Instantly — the evergreen automation layer. 

Webhooks fire, workflows trigger, leads move. That's 𝗽𝗹𝘂𝗺𝗯𝗶𝗻𝗴. Valuable plumbing, but plumbing.

Then I built Instantly MCP server to give access via Cursor/Claude code and run more agentic workflows. I'm still very bullish on MCP. Thousands of users and agents use it every day. It works. The problem with MCP is that it bloats context windows which makes it limiting.

But in 2026, something shifted.

Autonomous agents don't live in dashboards. They don't live in chat windows. 

They don't click buttons or navigate UIs. 

𝗔𝗴𝗲𝗻𝘁𝘀 𝗹𝗶𝘃𝗲 𝗶𝗻 𝘁𝗵𝗲 𝘁𝗲𝗿𝗺𝗶𝗻𝗮𝗹.

We went from the terminal → to GUIs → to chat interfaces → and now back to the terminal.

The circle is complete.

The terminal is the native environment of compute. Before visual interfaces made software accessible for humans, this was it. The command line. Now AI is taking us back.

So I built 𝗶𝗻𝘀𝘁𝗮𝗻𝘁𝗹𝘆-𝗰𝗹𝗶 to meet agents where they actually live:

→ A CLI command for every single Instantly endpoint → Skill-based tool definitions so your agent makes 𝗮𝗰𝗰𝘂𝗿𝗮𝘁𝗲 𝘁𝗼𝗼𝗹 𝗰𝗮𝗹𝗹𝘀 without bloating the context window

The problem with dumping 156 tools into an agent's context? It chokes. Too many options, too much noise, worse decisions. Context windows are precious and finite.

Instead, tools are organized into 𝘀𝗸𝗶𝗹𝗹𝘀 — grouped by what the agent is actually trying to do. Campaign management. Lead ops. Reply handling. The agent loads what it needs, when it needs it. Precise tool calls. Clean context.

What your agent can now do 𝗮𝘂𝘁𝗼𝗻𝗼𝗺𝗼𝘂𝘀𝗹𝘆:

➡️ Launch campaigns from scratch 
➡️ Import leads, write sequences, activate sending 
➡️ Handle replies (read, classify, respond, update CRM) 
➡️ Morning infra check (SMTP/IMAP, DNS, warmup) 
➡️ Duplicate and scale winning campaigns 
➡️ Weekly analytics across all active campaigns 
➡️ Lead list cleanup, dedupe, blocklist management

No dashboard. No GUI. Just the terminal.

n8n for the automations. MCP for the connectors (Claude/ChatGPT). CLI for the agents (OpenClaw)

Each layer does what it's best at.

Want the full setup guide + MCP config + agent workflow playbook?

➡️ Connect with me 
➡️ Comment "SEND IT!" and I'll DM you everything

Let's give your agents hands. 🤖

Earlier I open sourced instantly-cli — 156 commands giving AI agents full control of Instantly.ai.

Then clay-gtm-cli — turning every Clay table into a callable function.
Then hubspot-cli — putting your HubSpot inside your agent's world.

I said the stack was complete, but there was just another missing piece to the GTM stack which we all know (hint... you're on it now).

I just built & open-sourced heyreach-cli — a full CLI + MCP server for HeyReach.io.

47 commands. 10 API groups. Free. MIT licensed.

And this one is different.

Here's why:

Every CLI I've built so far touches transactional channels.
Email hits an inbox. Enrichment pulls a record. CRM logs a note.

𝗟𝗶𝗻𝗸𝗲𝗱𝗜𝗻 𝗶𝘀 𝗿𝗲𝗹𝗮𝘁𝗶𝗼𝗻𝗮𝗹.

Connection requests. Multi-step sequences. Conversation threads across 10 different LinkedIn accounts.

The entire motion is designed to feel human.

HeyReach made it scalable.

heyreach-cli makes it autonomous.

Here's what no one mentions about HeyReach:

They 𝗱𝗼 have an API.

But there's 𝗻𝗼 𝗼𝗳𝗳𝗶𝗰𝗶𝗮𝗹 𝗖𝗟𝗜.

No way to pause a campaign from a terminal. No way for an agent to read your unified inbox across all LinkedIn senders. No way to fire webhook subscriptions, tag leads, or check connection status without opening a browser.

LinkedIn was the last human in your outbound loop.

What your OpenClaw agent can now do 𝗮𝘂𝘁𝗼𝗻𝗼𝗺𝗼𝘂𝘀𝗹𝘆:

➡️ Launch LinkedIn campaigns — create list, add leads by profile URL, resume
➡️ Manage your unified inbox — read threads, reply, mark seen across all accounts
➡️ Tag and track leads through your entire outbound sequence
➡️ Subscribe to 12 LinkedIn event types via webhooks (reply, accept, InMail, completion)
➡️ Check connection status before reaching out — network intelligence built in
➡️ Pause underperforming campaigns, pull reply rates, pivot in real time
➡️ Manage multiple LinkedIn accounts from one API key

One command:

heyreach campaigns add-leads --campaign-id $ID --leads-json '[{"firstName":"Brandon","profileUrl":"https://lnkd.in/dntTJs9x"}]'

JSON back. 

No dashboard. No browser. No human.

And yes — every command is simultaneously an MCP tool for Claude, Cursor, and any MCP-compatible agent.

→ n8n for automations (the plumbing)
→ 𝗰𝗹𝗮𝘆-𝗴𝘁𝗺-𝗰𝗹𝗶 for enrichment (Clay as a function platform)
→ 𝗵𝗲𝘆𝗿𝗲𝗮𝗰𝗵-𝗰𝗹𝗶 for LinkedIn outbound (relationships, at scale)
→ 𝗶𝗻𝘀𝘁𝗮𝗻𝘁𝗹𝘆-𝗰𝗹𝗶 for email outbound (campaigns, leads, replies)
→ 𝗵𝘂𝗯𝘀𝗽𝗼𝘁-𝗰𝗹𝗶 for CRM (the close)

Prospect → Enrich → LinkedIn sequence → Email sequence → CRM.

Full cycle. Fully autonomous. Your agent runs the motion. You run the agent.
I wrote the full deep dive — all 47 commands, MCP config, agent workflow playbooks, and how to wire it into OpenClaw:

Want the full setup — API key config, OpenClaw agent, and workflow playbook?

➡️ Connect with me
➡️ Drop a "🔗" or "REACH ME!" in the comments and I'll DM you everything

Let's give your agents LinkedIn superpowers. 🤖

I gave my AI agents eyes into my live Chrome session.

Not a headless browser that has to log in from scratch.

My actual Chrome. The tabs I already have open. The accounts I'm already logged into.

Here's what that unlocks:

There are certain times when you need to run a session "headless" or have your AI go research. Then there are times where you already have the website open, you're logged in, and you just want to pull what you currently have in front of you into your AI Agent's context.

Consider it done. This is truly where the AI Agent works "with" you where you're at.

Imagine what this unlocks for things like LinkedIn RPS or other difficult ways of extracting data (hint hint 🫢) 

All in one shot. No API keys. No re-authentication. No scraping blocks.

That's what chrome-cdp enables.

Most browser automation tools launch a fresh, empty browser.

Chrome CDP connects to the Chrome you're already running, so your agent inherits your entire authenticated state automatically.

I forked the original chrome-cdp-skill and extended it with enhancements for GTM, web scraping, and CLI agent workflows — packaged as a pi skill (one-command install for any CLI agent).

For the technical folks: it connects via CDP WebSocket, holds a persistent daemon per tab so Chrome's modal only fires once, and handles 100+ open tabs reliably — something Puppeteer-based tools fail at constantly.

The way I think about the browser layer for agents now:

→ Public web scraping → Firecrawl
→ Authenticated live sessions → Chrome CDP
→ Structured APIs → CLI tools

Three layers. Every data surface your agent needs to reach. Just conversations right within a platform like Slack. 

Full resource page + install instructions 👇🏻

https://lnkd.in/gK7ZExPq

Earlier I open sourced clay-gtm-cli — turning Clay into a callable function for AI agents.

Last week I open-source instantly-cli turning Instantly.ai's entire API into agent-native actions. 

This week, I'm closing the loop.

I just built & open-sourced hubspot-cli — a full CLI + MCP server for HubSpot CRM.

55 commands. 11 API groups. Free. MIT licensed.

Here's what no one mentions about HubSpot:

They DO have an official CLI.

It only covers CMS — themes, templates, design files.

There is 𝗻𝗼 𝗼𝗳𝗳𝗶𝗰𝗶𝗮𝗹 𝗖𝗥𝗠 𝗖𝗟𝗜.

No way to touch contacts, companies, deals, or tickets from a terminal. No way for an AI agent to log a note, move a deal, or search your pipeline without opening a browser.

200,000+ companies run their entire revenue operation in HubSpot. 
(this one is for you Eoin McGuinness & Yamini Rangan 😉)

And their CRM is 𝗰𝗼𝗺𝗽𝗹𝗲𝘁𝗲𝗹𝘆 𝗶𝗻𝘃𝗶𝘀𝗶𝗯𝗹𝗲 to autonomous agents.

Here's the gap that's been killing fully autonomous GTM:

Your agent could already:

→ Send cold email sequences (instantly-cli)
→ Enrich any lead end-to-end (clay-gtm-cli)

But when a prospect replied? When a deal needed to move? When a note needed logging?

𝗛𝘂𝗺𝗮𝗻 𝗵𝗮𝗱 𝘁𝗼 𝗹𝗼𝗴 𝗶𝗻𝘁𝗼 𝗛𝘂𝗯𝗦𝗽𝗼𝘁.

Every. Single. Time.

hubspot-cli removes that last manual step. Your CRM is now part of your agent's world.

What your OpenClaw agent can now do 𝗮𝘂𝘁𝗼𝗻𝗼𝗺𝗼𝘂𝘀𝗹𝘆:

➡️ Create contacts and companies, link them together instantly
➡️ Move deals through pipeline stages on verbal commits
➡️ Log notes, calls, meetings, and tasks on any CRM record
➡️ Pull a full account brief — contacts + deals + activity
➡️ Build contact segments with cross-object property filters
➡️ Research any account and surface stale opportunities
➡️ Book and log discovery calls without touching HubSpot

One command:

hubspot deals update $DEAL_ID --dealstage closedwon --closedate 2026-03-15

That's it. JSON back. No login. No GUI. No browser. 

Your Agent handles it. You Manage it.

And yes — every command is simultaneously an MCP tool for Claude, Cursor, and any MCP-compatible agent.

The stack is now complete:

→ n8n for automations (the plumbing)
→ 𝗶𝗻𝘀𝘁𝗮𝗻𝘁𝗹𝘆-𝗰𝗹𝗶 for outbound (campaigns, leads, replies)
→ 𝗰𝗹𝗮𝘆-𝗴𝘁𝗺-𝗰𝗹𝗶 for enrichment (Clay as a function platform)
→ 𝗵𝘂𝗯𝘀𝗽𝗼𝘁-𝗰𝗹𝗶 for CRM (the missing link)

Full cycle. Fully autonomous.

Prospect → Enrich → Sequence → CRM.

Your agent handles the entire motion. You manage the agent.

I wrote the full deep dive — architecture, all 55 commands, MCP config, and agent workflow playbooks:

Want the full setup — Private App config, OpenClaw agent, and workflow playbook?

➡️ Connect with me
➡️ Drop a "🧡" or "HUB ME!" in the comments and I'll DM you everything

Let's give your agents a CRM. 🤖

Cloudflare just shipped a web crawler.

The same company that spent years — and billions — defending the internet against crawlers.

The irony is real. But more importantly, the signal is loud.

Here's what it means for your business:

𝗬𝗼𝘂𝗿 𝘄𝗲𝗯𝘀𝗶𝘁𝗲 𝗻𝗼𝘄 𝗵𝗮𝘀 𝘁𝘄𝗼 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿𝘀.

Not metaphorically. Two distinct visitors with completely different needs.

People browse. They read your copy, click your CTAs, judge you visually.

Agents crawl. They parse your structure, read your schema, check your llms.txt — and decide whether to recommend you to the human who asked.

𝗔𝗴𝗲𝗻𝘁𝘀 𝗮𝗿𝗲 𝗻𝗼𝘄 𝘃𝗶𝘀𝗶𝘁𝗶𝗻𝗴 𝘆𝗼𝘂𝗿 𝘀𝗶𝘁𝗲 𝗺𝗼𝗿𝗲 𝘁𝗵𝗮𝗻 𝗵𝘂𝗺𝗮𝗻𝘀.

And most sites aren't ready for them.

Recently, one of my clients got a cold inquiry from someone at FIFA.

Out of curiosity, they asked: "How did you find us?"

"𝗖𝗵𝗮𝘁𝗚𝗣𝗧 𝗿𝗲𝗰𝗼𝗺𝗺𝗲𝗻𝗱𝗲𝗱 𝘆𝗼𝘂."

No ad. No outbound sequence. No referral.

And last week, we got an inbound at TOFU. Right on the call they said:

"𝗖𝗹𝗮𝘂𝗱𝗲 𝘁𝗼𝗹𝗱 𝗺𝗲 𝘆𝗼𝘂'𝗱 𝗯𝗲 𝘁𝗵𝗲 𝗯𝗲𝘀𝘁 𝗽𝗲𝗿𝘀𝗼𝗻 𝘁𝗼 𝘁𝗮𝗹𝗸 𝘁𝗼 𝗮𝗯𝗼𝘂𝘁 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝘀𝘆𝘀𝘁𝗲𝗺𝘀."

AI search is real pipeline. Most people don't know they're missing it because there's no UTM parameter. You're invisible by default.

Now — back to Cloudflare.

They didn't kill the scraping market. They 𝗰𝗼𝗺𝗺𝗼𝗱𝗶𝘁𝗶𝘇𝗲𝗱 𝗮𝗰𝗰𝗲𝘀𝘀.

Here's how the landscape actually shakes out now:

→ 𝗖𝗹𝗼𝘂𝗱𝗳𝗹𝗮𝗿𝗲 𝗖𝗿𝗮𝘄𝗹𝗲𝗿 — infrastructure level, free, single endpoint. The commodity play is dead.

→ 𝗙𝗶𝗿𝗲𝗰𝗿𝗮𝘄𝗹 — still the best for LLM-native output, JS rendering, site-wide crawls built for language models.

→ 𝗭𝗲𝗻𝗥𝗼𝘄𝘀 — still the go-to when you need anti-bot bypass and protected sites.

→ 𝗖𝘂𝘀𝘁𝗼𝗺 𝘀𝗰𝗿𝗮𝗽𝗲𝗿𝘀 — still essential for authenticated sessions, proprietary formats, complex SPAs. Engineering work doesn't commoditize.

The bigger shift: 𝗲𝘃𝗲𝗿𝘆 𝘀𝗶𝘁𝗲 𝘄𝗶𝗹𝗹 𝗯𝗲 𝗰𝗿𝗮𝘄𝗹𝗲𝗱. Not could be. Will be.

Design accordingly, or leave the opportunity on the table.

I wrote the full breakdown — the two-customer model, Cloudflare vs the scraper landscape, and a concrete agent-ready checklist with exactly what to implement, including a free AI-search 'ready' report. 

➡️ https://lnkd.in/g355g7f6

Deploying and using OpenClaw Agents within Top of Funnel along with each and every client we serve having their very own has been a *massive* game-changer. 

We've gotten to the point where clients are wanting more and having deploy into other departments for various use cases (not to mention them going fully geeking out and having fun). (yay for client retention! 😆)

One of THE biggest changes we've seen is being able to fully train an OpenClaw agent with the client's context on everything they know and want to feed into the Agent, just like if you are training an employee. The OpenClaw agent stores it to memory, and it will always know this. 

If the context is confidential or needs to be encrypted, we implement OpenShart, a frameworks I built last week and is working great. 

Then we (as TOFU) can also give the OpenClaw agents tools and resources that we know best to do the job our clients need it to do (like sales and outbound!). 

Speaking of tools and resources, one area that is a concern is how to handle API keys. After all, the OpenClaw agent even has access to it's own environment variables and can do whatever it wants with it. (hint don't do that!)

And for anyone out there, it's too easy to believe the Agent only for it to say "just tell me your API key and I'll get it going!"

Not so fast....

Here are two scenarios we can easily navigate quickly:

1️⃣ If we are doing a "GTM sprint", we will use Clay simply because of two things. Many integrations and credits under one login and a collaborative environment to analyze data that we are all very familiar with. No-brainer.

We use webhook tables to send the payload to that table from the OpenClaw agent, where it can execute a normal workflow like anything else we do traditionally. This serves as a massive benefit because the OpenClaw Agent just did the heavy lifting, where we (as humans) are verifying for quality and accuracy. 

Another benefit is that Clay is effectively our "execution logger" where we can apply strict controls like conditional formulas to ensure it always meets standards and catch edge cases. 

After all, we are spending credits and want to have complete control versus the OpenClaw Agent.

2️⃣ Another example is doing the same for any other type of use case that we want, which to use n8n webhook workflow as a "proxy" endpoint instead of giving it direct access to your CRM (like Airtable in this case). This allows us (again) to control the tools, see the 'execution logs' and know that we have control over these and know that the OpenClaw agent has zero access to raw API keys. 

TL;DR and KEY takeaways:

OpenClaw is doing massive amounts of the heavylifting. Complete unlock is an understatement.

Humans verify two things: Quality and Accuracy. 

This is just scratching the surface, but this is my type of 2026 GTM Engineering. 🍻

The Instantly.ai MCP server alone is used by thousands of users and agents across the globe. I've been deep in this space — building integrations, connecting AI models to external systems, watching how agents actually work in production.

I'm still very bullish on MCP.

But MCP as we knew it has evolved. It's become "connectors" — which is really what it always was. Enabling AI agents and models to have connections to external resources. Your CRM, your database, your email platform. That's just plumbing. Valuable plumbing, but plumbing.

We've also seen the rise of "mini-apps" inside chat interfaces. ChatGPT's "back to the chat" pattern. Claude's artifacts. The idea that you should be able to do everything inside the chatbot without leaving the conversation. 

This is powerful — for humans.

But then OpenClaw came out.

And all the swarms of fully agentic, autonomous workflows started to boom. Research agents feeding into writing agents feeding into deployment agents. Multi-step workflows running end-to-end with zero human intervention.
And where do these agents live? Where do they do their best work?

The terminal. The CLI.

After all — this is where it all started. The terminal is fundamentally the core of compute. Before the visual UI came along to make software accessible for humans, this was it. The command line.

Now AI is taking us back.

The AI model is the brain — the intelligence, the reasoning engine. But just like a human who needs external tools (the internet, books, databases) to do useful work, AI agents need external systems to pull in context. APIs, file systems, databases. And the most efficient way to access those? The CLI and REST APIs.

Even Peter Steinberger, the creator of OpenClaw, has purposely not implemented MCP — because the CLI and REST APIs have been the digital freeway of software communication for decades. Proven. Battle-tested. Universal.

Programming languages like Python, JavaScript, C — these are just frameworks we all agreed to communicate on. They define the programmatic logic of how software interacts with data. But what happens when AI models start developing their own ways of executing at the OS level?

That's a question for tomorrow. For today, the answer is clear:

The CLI and REST APIs are the easiest, most efficient way to enable AI agents to do their job — without overwhelming the context window, without unnecessary overhead, without reinventing the wheel.

We went from the terminal → to GUIs → to chat interfaces → and now back to the terminal.

The circle is complete.

Full article: https://lnkd.in/gNr_x6Qw

Almost St Patty’s day. 

That means…

➡️ wear something green

➡️ feel lucky

➡️ build a leprechaun trap with your kids

➡️ go run a 5K race downtown  

➡️ last, but certainly not least, have a 🍻 and be around people who fill your bucket


🍻☘️

Elyse Charleson

No alternative text description for this image

AI agents handles sensitive business conversations every day — investor updates, revenue targets, hiring decisions, customer deals.

The problem? Most AI agents store everything they remember in plain text files. That's like writing your passwords on sticky notes and leaving them on your desk.

So I built OpenShart — an encrypted memory system for AI agents running on OpenClaw.

Here's the simple version:

My agent BB-8 now has two types of memory:

1. Open memory — for everyday stuff like preferences, task lists, and personality. Think of it like a notebook on your desk.

2. Encrypted memory — for anything confidential. Business numbers, strategy, people stuff, credentials. This gets military-grade encryption before it's stored. Even if someone accessed the files, they'd see nothing but scrambled data.

The agent decides in real time which memory to use. I tell it "remember that our Q1 revenue target is $2.5M" and it automatically recognizes that's confidential business data and routes it to encrypted storage. No manual tagging, no extra steps.

This wasn't a planned project. It came out of necessity internally and for our clients (and very quickly). When your AI agent is sitting in on real business conversations, you realize very quickly that unencrypted memory isn't an option.

Built it, tested it, and it's working in production now.

If you're giving AI agents access to your real business — and a lot of us are — the security of what they remember matters just as much as the security of your email or your database.

I know this is a needed frameworks much bigger than me so I have open sourced and made it completely free so we can all build this thing together.

openshart.dev

There's a buyer in 2026 that I don't think most people are thinking about yet.

For the entire history of software, the relationship has been the same: humans discover apps, download them, bookmark them, and use them. 

As a business, your job was to get a human to your front door (or Zoom call) and convince them to live in your platform.

But something fundamental has shifted.

When someone sits down with an AI agent and says "build me an app with payments and a database" — the agent doesn't browse ProductHunt. It doesn't sit through a demo. It doesn't download your app to try it.

It scrapes your website. It reasons about your product. And if it trusts what it finds, it installs you with a single terminal command.

Commands like:

npm install stripe
npx supabase init
npx vercel deploy

This is not theory. This is already happening. There's a reason every "vibe-coded" app in 2026 uses the exact same stack.

These platforms win because they are:

→ Discoverable — clean docs, semantic HTML, strong authority signals that LLMs can reason about

→ Installable — one CLI command, zero friction, no demo required
The implications for GTM are massive:

Your website is no longer just a landing page for humans — it's a data source for AI agents. 

Your documentation isn't just for developers — it's the sales pitch that agents read before recommending you. 

Your CLI isn't just a developer tool — it's the frictionless checkout for the agent economy.

If an agent has to manually sign up, verify an email, and book a demo just to use your product? It finds an alternative.

I broke this down in a full resource covering:

• Why the saturation problem exists (downloading apps, bookmarking sites)
• The two pillars: Be Discoverable + Be Installable
• Why Stripe, Supabase & Vercel are winning the agent economy
• A practical GTM readiness checklist for 2026

Link here: https://lnkd.in/gHaj3p_b

If I had to put one word on Clay's new pricing model it would be:

Necessary.

Why?

The product has evolved. GTM has evolved. And the wider industry has certainly evolved.

Even with all the massive changes since the early Clay days years ago, this is the first major move they've made that directly hits everyone on pricing and usage. They're splitting the old single-credit system into two meters: Data Credits (for enrichment) and Actions (for platform orchestration). Data costs are dropping 50-90% on the most common enrichments. Actions are priced at fractions of a penny.

Having been deeply connected with the Clay team for years, one thing has always been true — they care about what's fair, what's competitive, and what problems they can actually solve.

The thing they've always solved for me and my team at TOFU is simplicity inside complex GTM stacks. One login. One subscription. 150+ integrations. When the name of the game is speed, tight feedback loops, and validation — that matters.

Just a few days ago, I released an open-source CLI (clay-gtm-cli) that makes Clay fully agent-native — AI agents can fire enrichment workflows, get structured results back, and pipe them into any GTM system. Under this new pricing, that kind of programmatic usage becomes dramatically more accessible.

This pricing change asks all of us to reevaluate how we use Clay and run the numbers on what actually helps you or your clients scale. That can only be decided by you.

But if anyone recalls what happened to Blockbuster — history has a tendency to repeat itself. 

The ultimate equation is evolving to what is necessary to stay relevant and valuable, which I think we can all agree on. 🍻

Before starting Top of Funnel a few years ago and focusing heavily on AI/automation, web scraping, and outbound for sales/marketing, the roots of Headhunting still are largely there. 

(After all, it's all the same where you're generating leads!)

There's a new way to Headhunt though, and that is via our OpenClaw agents. 

In this picture, I have my trusted agent, "M-O" who did everything a full-desk recruiter would do. 

➡️ Business Development: Has TAM Surveillance on all jobs in a niche of interest and watches them like a hawk. Reaches out to the Hiring Manager.  

➡️ Candidate Sourcing: Is able to source candidates, score and segment them based on any criteria, and enrich, personalize on a deep 1:1 message, and manage the entire outbound flow. 

In this image, we received a job order from a good friend of mine who is looking for a specific role. So I tasked M-O, who understands full context on the company, the role, who a good fit is, and ran through a scoring criteria based on their location and specialty, and automatically segments into campaigns (C1 through C7) for prioritized outreach. 

We used ContactOut to enrich personal emails, and LeadMagic for business emails. Then M-O used Instantly.ai to send emails out at scale with our large infrastructure, optimized for deliverability.

And because for any true Headhunter out there, we know that "everyone" is open to the right opportunity for the right salary/timing. 🙃 

So we reach out to both emails at the same time (with a unique approach that is in good taste) and therefore our success rate is much better. 

The REAL value here, is not only time to source and reach out to people like what M-O did, but you can see in this image that it asked me if we wanted to update his record and also proactively reach out to his personal email to engage there. This keeps things very organized and ensures we can reach out to every person who is qualified for this opportunity is going to hear from me quickly. 

By the way, this was done in about an hour's worth of "human" time. 

This is definitely the new way of recruiting/headhunting. 🍻

With a massively growing fleet of agents for both internally at TOFU and for our clients, the fastest way you can communicate your thoughts is via voice. 

What better way than use Slack audio clips to talk to your agents no matter what device or where you’re at?

WALL-E even puts the green check mark on its own and responds as expected with his own personality. 

Agents are the new “operators”…

We (as Humans) are the managers…

Verifying quality and accuracy of the outcomes/deliverables.

What a time to be alive. 🍻

graphical user interface, text, application, chat or text message

My wife and I usually have our midday "walk/touch grass" date every day at 11 AM to check in on each other and spend time.

But over the last couple days, we have gotten dumped on with snow ❄️.

Kids are on their SECOND snow day in a row 🙄, so they are out throwing snowballs, building snowmen...you know...fun stuff on a "normal Wednesday."

But on our walk, I noticed two teenagers with snow shovels walking around our neighborhood and knocking door-to-door, particularly on houses with driveways with snow on it. 

It was clear they were helping fellow neighbors around shovel their driveway and charging a reasonable price or being tipped for their work. 

But I think this actually a good reminder and teaches the true foundations of any sales/business development opportunity. 

That is revolving around Signals, Relevance, and Triggers/Actions to follow:

➡️ TRIGGER -- clearly we got a lot of snow. The weather triggered the signal that drives new pain points and needs all over.

➡️ SIGNAL -- There will be driveways that have lots of snow, and there will be driveways that are completely cleared. 

➡️  RELEVANCE -- Some people have the time, the right shovel, and health to not break their backs and clear their own driveway (not relevant).

And there are others who likely don't have time, don't have a snow shovel, or a healthy back to do it. (Hint...this is opportunity...)

➡️ ACTION -- These two teenagers took it upon themselves to see this opportunity and went around to houses that defined their ICP (which is a driveway with snow!).

➡️ NEW DEALS -- It's mid-day Wednesday, so they will likely get some houses that don't answer (aka not reply to your email), some "no's" (not interested), and they will likely also get some "yes please!" 💰

The point is:

1️⃣ They saw the trigger (weather forecast)
2️⃣ They saw the signal (snowy driveways)
3️⃣ Saw the relevance (why ask people who have clear driveways!?)
4️⃣ They took action and did cold outreach.
5️⃣ After enough houses, they are surely going to close some deals. 💰

These are the fundamentals in any channel (cold calling, cold emailing, door-to-door) around relevance and bringing true value. 🍻

🚨Value Webinar Alert!

Next week, Thursday Feb 26th, I'll be joining n8n showcasing a live demo on how you can level-up your cold outbound & sales workflows using Instantly.ai.

Absolutely thrilled for this one and couldn't be more honored to be linking up with such amazing people in helping businesses grow and operate more efficiently.

Link is in post! 👇🏻

Next Thursday, February 26th: join us at our monthly Community Livestream!

We'll unveil our new monthly Community Challenge, welcome the Instantly.ai team for a live demo, go through product updates, and so much more...

Whether you're an automation tinkerer or a workflow wizard, you won't want to miss it.

Sign up now 👉 https://bit.ly/3OgS3bL

The very talented Jacob Dietle just posted a video which I think we should all recognize that "you're not alone" is very real. 

This is an "old" but "new" feeling that many of us are feeling, but is often difficult to recognize and address it. 

Without summarizing his wonderful video that he posted, I think it's never more important to know that in today's world, it is far too easy to feel anxious like "you're never doing enough...", and "you're always behind.."

When at the end of the day, we (as humans) are still on this Earth for a very finite amount of time. 

And it all comes down to three things:

➡️ You only get one body...

➡️ You only get one family...

➡️ You only get one life ...

Are you making efforts to prioritize of all three?

🟠 Context Operating Systems for GTM | Amplify Your Taste & Scale Expertise

Agentic Anxiety: The Lie Of Optionality and "Don't Get Left Behind"

You ever feel like there is too much to do, never enough time and there's a massive opportunity cost to not running your claude code 24/7?

You're not alone. I've felt this way many a time. It is part of a larger feeling of anxiety and uncertainty that you're not "staying up to date with AI enough" that I think we can all intuit already - but today I wanted to talk about two more specific trends, one new and one old, that are now at the forefront. 

Information Asymmetry: Since the printing press there's effectively been more overall knowledge than time in someone's life to work through it. In the 1940's Vannevar Bush, think like just raw invention output of Edison + he lead America's entire national scientific coordination through WWII and beyond, identified this exact problem

"...publication has been extended far beyond our present ability to make real use of the record. The summation of human experience is being expanded at a prodigious rate, and the means we use for threading through the consequent maze to the momentarily important item is the same as was used in the days of square-rigged ships."

He proposed a physical machine that ran on microfilm and an analog computer called the Memex. Interesting the vision for this machine could pretty well describe what is now possible with Obsidian + Claude Code today. You are part of a much longer history than you realize. The smartest people in the world decades ago were grappling with it the same core problem (too much info, not enough energy or time) that you are today 

Agentic FOMO: The second, more acute and recent trend is unique to our times. It is the lie of digital optionality (endless feed scrolling, one more post, one more video, one more x) compounded by the ability of LLMs to generate information extremely relevant to your life - all tied up in the hype/fear cycle of "AGI."

In this video I discuss these problems in depth, how I've noticed them pop up in my in own life (I now find myself checking on my claude code agents in between sets at the gym), the ways I'm trying to solve them and what I think "solvable" vs something closer to acceptance and active engagement with limitations. 

I will not even pretend I have anything close to all the answers, but this is my POV and experience, and I hope its useful to you at the least :) 

Figuring out that this is a problem people have been experiencing for far longer than I even considered was comforting and helped me step back. 

Anyone else feeling the agentic fomo? Why and how are you dealing with it? 

---

https://lnkd.in/eKqkX-nG

Open video: Agentic Anxiety: The Lie Of Optionality and "Don't Get Left Behind" on youtu.be

Agentic Anxiety: The Lie Of Optionality and "Don't Get Left Behind"

n8n Livestream: new Community Challenge & Instantly.ai

Join us for the February n8n Livestream, your monthly dose of product updates, live demos, and community stories.

This time:

Our new monthly Community Challenge: we'll unveil our first monthly Challenge: think 'mini-hackathon' meets 'hands-on learning'. Our first challenge will focus on AI evaluations, with a new theme revealed each month in the Livestream.

Integration spotlight: The team from Instantly.ai joins us for a product overview and a live build using their n8n node.

Come hang out, learn what’s new, and connect with the n8n community!

n8n Livestream: new Community Challenge & Instantly.ai

taplio

The Battle

🦾Eric Nowoslawski

Brandon Charleson

Followers

Posts

Engagements

Likes

Performance Score🦾Eric Nowoslawski's LinkedIn Performance

Performance ScoreBrandon Charleson's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

🦾Eric Nowoslawski

Brandon Charleson

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}🦾Eric Nowoslawski's LinkedIn Performance

Performance ScoreBrandon Charleson's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance Score🦾Eric Nowoslawski's LinkedIn Performance