🦾Eric Nowoslawski

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

2 years of effort went into the Prospeo.io Company/Contact API and this is why it was WELL worth the wait.

2 years ago Jon asked me, "If there were a database that could do anything you want, what would you wish it would do?"

I had a laundry list of ideas from

- Data accuracy and transperancy
- Organization tools for the team
- API capabilities

And then, I didn't hear from them for a little while. We kept using their email finding tool in all our Clay tables as a happy customer of course but I thought they were just focusing on that.

Then Hugo reaches out to me and says, "Hey, we finished the database the way you wanted. Want to have a look?"

I sign in to Prospeo V2 and my mouth is on the floor of everything they built. Good things definitely take time and they crushed it on this one.

I go over my favorite parts in the video down below but here are my favorite features you should check out if you don't watch the video and go straight to the site.

- Data quantity - I did comparisons of searches against Sales Navigator and was extemely happy with the counts that come back for the same filters on people and company searches in sales nav vs. prospeo

- Data quality - Every profile I checked worked at the company Prospeo says that they work at. I tried to be sneaky and check a friend that I know changed jobs a week ago and they already had that in their database.

- Email quality - All emails are SMTP verified with bounceban on top of that.
The API is so great. I just spent the christmas break making a company database we are going to scrap and just use their's.

- When you enrich a person, you get the company data with the API response so you don't have to worry about double spending on companies!

- The platform is so fast to make searches in.

- Your team can visually make searches and then click a button to get all of the JSON of that search to quickly get that search with their API.

I think they are just starting to launch it and I would highly recommend checking it out to help with your company and contact finding workflows. 

We are starting to switch a lot of TAM links internally over to them so we can stop downloading and uploading so many CSVs!

Not paid to post this or anything. 

Just found it last week and thought they did a really great job and other people would like to see it!

Check it out here! 

https://lnkd.in/dJw9eAEy

Founder Growth Engine X | Clay Enterprise Partner

Tomorrow, join me and many other outbound minds to discuss the state of outbound in 2026. (of course focused on cold email because I'm not qualified to talk about anything else). 

We’ll cover the dreaded Gmail AI filter tweet that hasn’t even been rolled out yet that everyone is asking about. 

Deliverability trends we are seeing from sending 7M emails per month on behalf of our customers

The way data is getting more accurate and cheaper and combining this with AI can cheaply laser target your ICP. 

I’ll leave the link to register in the comments. Hope to see you there! 

Vaibhav Namburi
Harris Kenny
Nick Abraham
Wesley Hoang 🛠️♠️
Alex Fine🌲
Christian Oland
Matthew Lucero
Enzo Carasso 🧲

I built a way to monitor your customer's LinkedIn content with a RapidAPI endpoint and Cursor so you can reach out to your ICP and name drop a customer they have heard of. Here's the step by step on how to do it.

I think one of the best ways that we can reach out to an ICP is if they already know of one of our successful case studies. 

You could ask your customer for a referral, but that puts work on them. 

You can monitor who they already know and reach out as if it's a regular message, and the prospect will actually have heard of the social proof you're mentioning. 

To set this up, all I did was give a voice note while walking my dog and Cursor took care of the rest. 

I hope you can even copy and paste this post and give some more context and set it up yourself. 

Step 1

Get the Rapid API endpoint. If you comment "monitor", I'll send it to you. Sorry to play this game but LinkedIn hates external links. 

Step 2

Use the MCP in RapidAPI and connect that to Cursor

Step 3

Explain what you want it to do. 

"Hey Cursor, I want to use this endpoint to build a micro app that I can enter a comma separated list of linkedin companies. I want you to 

1) find all of the posts of the linkedin companies
2) Find all of the people that work at the company and find their posts
3) Pull all of the people that have liked and commented on the posts and filter for XYZ titles and company size. 

Step 4

Add a little bit more to the prompt above about how you exactly want this to work.

Step 5

Enter Plan mode in Cursor. Drop your prompts and watch the plan

Step 6 

Edit/Approve the plan. I was lucky that Gemini Flash got this done in one shot, so I'm hoping I can pass this on and help others use it too. 

Again, comment "monitor" and I'll send it to you but if you don't want to wait for me, searching rapidapi(.)com(user)apibuilderz should help those that want to search on their own!

This workflow I made with Cursor and hosted on Railway saves me 6 hours a week on email deliverability tasks and our customers get much more consistent results from it. 

Over the last 6 months, we really systematized our email deliverability infrastructure. 

No more guessing and then ordering domains that we thought felt right. I built an SOP for everyone that I can give away at the end of the post. 

We would miss customers. Over order for some, under order for others. It was crazy. 

So we got extremely scientific about it. 

We keep all domains marked as status, Active, Insurance, or Cancel in Supabase. 

We always want to keep AT LEAST 50% of our Active daily sending volume in Insurance so that we can always rotate our domains. 

Every Friday, I would use a python script that would get the reply rates of all the domains in our campaigns so I could make decisions of what should be cancelled and what should be rotated from Insurance to Active. 

Once I had the numbers, spreadsheet warfare would begin. 

Making sure that every customer had enough and we didn't cancel too many domains. Numbers wouldn't look right so I'd check them manually. It was a mess.

Now, every Friday, I run this little Preview Global Rebatch and it send me a slack message with every customer, their daily send goal, how many domains are active, how many domains do we suggest to cancel, how many domains I need to buy more of, etc. 

And I hit accept and not only does it change all the statuses in our Supabase, it also pulls all of our customer's campaigns and swaps out the old domains for new domains. 

Easily my favorite workflow for deliverability. 

If you want our email deliverability SOP that goes over how we do this and how we work with Hypertide.io and Zapmail.ai to get it done, comment "primary" and I'll comment it back to you!

Here's the prompt to get you started building your own google maps scraper with Claude code or Cursor. 

One of my favorite things about LLMs and coding is that my team can build things that impact our business and then everyone can use. Bharatt Arorah built this google maps scraper for us and it's been so easy to use. 

Then when someone else on the team wants to add to it, it'd be really easy for them. 

This scraper is built on a RapidAPI endpoint that gives us 3,000,000 requests per month for $100 a month. 

Here's the prompt I'd give for Cursor or Claude Code to figure this out for you. 

First connect to the Rapid API MCP and then run this because it'll ask you to restart. (I don't use the MCP to actually run the code. I just use it to connect to the documentation to help set everything up). 

If you need the list of zip codes don't ask me or your mother for it. Ask Chat GPT literally asking "Do you know of any GitHub repositories that have all of the zip codes in the United States, or any other public ways that you can download that information easily?" and it'll come up, I promise. 

Long Text not optimized for a linkedin post AT ALL. 

Hey cursor, can you explore this Google Maps endpoint and build a plan for me that would be able to create an input where I can say a Google category, and then you'd be able to scrape it?

Here is a file of all the zip codes around the United States so that you have something to parse by. Search every one of these zip codes with my keyword to then find the companies that are in those areas. Build something into it that includes logic that, if a company is in a different business category that is excluded from my results, I want to be able to filter for the category. Let me know if we could exclude the cities and also run multiple category searches in one run. Set up the database schema to pull the zip codes from SupaBase. Literally go one by one doing a search across each of the zip codes. I also might wanna specify that I want to include certain cities, exclude certain cities, or include and exclude states in the future as well. Let's also add the filters for that to be possible as well. 

Hope that helps and provides a good lead gen project for people trying to get into coding with Cursor or Claude code.

Send those companies over to a Clay webhook to have a contact finding waterfall and you're all set!

I have to admit, six months ago I didn't get the point of MCPs.         
                                         
This post is for people who were like me. Also for my employees who are reluctant to start using these tools.                     
                                         
What is an MCP (Model Context Protocol)?                    
                                         
Think of it as a gateway for an AI agent to understand a tool.         
                                         
Instead of asking an agent to read API documentation (it will miss things),  
you connect it to an MCP. The agent figures out:                
 - What endpoints exist                             
 - What payloads to send                            
 - How to actually accomplish your task                     
                                         
Then you ask it to write the TypeScript or Python to make those API calls   
directly.                                   
                                         
Way more efficient. Way fewer tokens.                     
                                         
 My biggest tip:                                
                                         
 We're not actually doing a lot of work with MCPs as in moving data or doing work with MCP.             
                                         
We use them so agents understand the API documentation better.         
                                         
The MCP helps the agent learn the tool. Then we have Claude write actual code that calls the APIs.                                
                                         
That's the workflow:   
                           
 1. Connect MCP                                 
 2. Agent learns the tool                            
 3. Agent writes the automation code                      
 4. You run the code directly                          
                                        
 If you've been confused about MCPs, hopefully this helps.

A couple of customers have asked me what I think about this announcement from Google and here's my humble prediction. 

I'm not going to be naive and say that this won't have any affect on cold email. But also, even the video doesn't explain very well what this will do. 

I think that there is a best case scenario and a worst case scenario. 

Best case - this is an overall summarization of the email meaning that the preview text we optimize our first lines for today won't be as important as optimizing for how the summary will come across. 

A recipient can see the offer (maybe even stated cleaner than what an average copywriter can do) and choose to open or disregard. 

Worst case - it becomes a new spam filter and labels all external emails you haven't received before as spam with no preview. 

I truly think the actual outcome will be somewhere in the middle and lead magnets will become more and more important so you aren't just pitching calls and getting hung up in the spam filters. 

Truly, no one knows but it's not in my inbox yet but excited to see it when it is so we can start playing with it and see how it affects the channel.

Make your GTM system impossible to copy.

We build bespoke AI intelligence orchestration systems that accelerate decisions and unlock unique approaches.

Joe Rhew

A lot of people sleep on CLAUDE.md.

It's the single most important file in any Claude Code project, and most people either skip it entirely or overcomplicate it.

In this video, I show the difference between having one and not having one. Same folder, same content, same question - but in one case Claude immediately finds what I need, and in the other it fumbles around guessing.

The reason: Claude Code automatically reads CLAUDE.md at the start of every session. No prompting. It just loads that context before you even ask your first question.

So whatever you put in there - project structure, how you like to work, critical instructions - Claude has it from the start.

A few things worth knowing:

You can nest them. Root-level CLAUDE.md for general context, subfolder CLAUDE.md for specific instructions. Claude reads both, prioritizes the nested one.

You can reference other files. If your context is getting long, break it into separate files and link to them.

Start simple. I've seen people use AI to generate massive CLAUDE.md files on day one and then wonder why things feel off. A few lines is fine. Add more as you learn what actually helps.

If you're using Claude Code without a CLAUDE.md, you're probably repeating yourself way more than you need to.

How do you use CLAUDE.md in your projects?

GTM Intelligence Orchestration | The Workflow Company

We rebuilt our entire outbound system from scratch. Before sharing what it looks like, a short story:

A few months ago I started working with a Series B AI company.

Hot product. 
Huge TAM. 
The kind of company VCs love.

But they had a problem: too many possibilities.

Their product worked for dozens of use cases, across multiple personas, in countless industries. They knew what was working - but they had no idea what they were missing.

What other industries might resonate? What use cases hadn't they explored?

So we started testing. And I immediately hit a wall.

The old way - building lists in Clay, writing copy, launching campaigns - couldn't match the pace we needed.

We weren't trying to run 1-2 campaigns a month.

We wanted to test 10+ hypotheses. Fast.

So we rebuilt everything from scratch:

1. Context Repo
A single source of truth: company narrative, product positioning, personas, campaign angles, offers. All documented. All version-controlled.

2. Micro-List Building
Instead of 5,000-person blasts, we build targeted lists of a few hundred people per hypothesis programmatically. Each list tests specific traits we have theories about.

3. Programmatic Copywriting
We feed the LLM the exact context it needs—campaign brief, persona, offer—pulled directly from the repo. No manual copywriting bottleneck.

4. Cost Optimization
Rapid experimentation = high token volume. So we built prompt caching and batching that saves 60-80% on costs.

5. Audit Trails
Every decision is documented. If we targeted the wrong pain point, then we can simply tweak the campaign brief, rerun. The system learns.

The result:

We now scientifically test outbound hypotheses with full traceability.

We're not guessing. We're not hoping our instincts are right. We're not rerunning a playbook that once worked for a similar company.

We're systematically exploring the space to find what actually works - not just what worked before for someone else.

The difference between a local maximum and a global maximum.

And it's why I believe the future of GTM isn't more SDRs running their own experiments in silos.

It's a centralized system that controls the testing - so you're not leaving hypothesis exploration to chance.

---

If you're in a similar situation as our client - drowning in hypotheses and frustrated with how slow it is to actually test them or if any of this resonates with challenges you're facing - happy to chat, DM me. 

I'm always down to go deeper on how we're thinking about this and how we've set things up.

Every time you use ChatGPT or Claude, you're starting from scratch.

"Here's my company. Here's what we sell. Here are our personas. Here's our competitive positioning."

Copy. Paste. Repeat. For every single conversation.

Claude Code solves this by writing files. Which means you can build a context repository - markdown files that capture everything an AI needs to know about your business.

Company overview. Products. Personas. Messaging. Competitive positioning. Voice and tone.

And it's not a black box memory - you know exactly what files exist and what they say.

Once it's saved, you never repeat yourself again. You just say "read the context files" and Claude is already up to speed.

In this walkthrough, I walk through building one from scratch for Anthropic as an example:

> Spinning up parallel research agents to gather intel
> Synthesizing raw research into structured context files
> Using that context to brainstorm outbound messaging

The whole process took maybe 20 minutes. And now that context is permanent - ready for the next task, and the one after that. 

(Granted this isn't production grade, but it's enough to get started with for most teams.)

This is the compounding advantage of Claude Code. Every piece of context you save makes every future conversation smarter.

If you're tired of being a broken record every time you open a chat window, this one's for you.

For those of you already deep in Claude Code - what do you wish you'd set up on day one?

--

This is the second video in a series I'm doing on Claude Code for Sales & Marketing teams. Link to the first video in the comments.

Git is the single best investment you can make when starting with Claude Code.

I know it sounds like a developer thing.

But - Claude Code modifies files on your machine. Without version control, you're one bad edit away from losing work you can't get back.

With Git + GitHub, you get checkpoints you can always roll back to, cloud backup, and the ability to pick up your work from any machine.

Here's the mental model that made it click for me: think of Git like packing and shipping boxes.

▶ git add = putting items in the box
▶ git commit = sealing it and writing a label
▶ git push = shipping it to the warehouse

The reason you "stage" files before committing is so you can group related changes together. If you update your persona doc and that triggers changes to your campaign angles and pain points, you probably want all of those in the same box. That way you can roll back the whole thing if your hypothesis was wrong.

If you're still confused, that's ok. Watching this video will help clarify things for you.

In this video, I walk through:
→ What Git actually is (and why it's more powerful than Google Docs versioning)
→ The add → commit → push workflow
→ How to pull changes when working across machines

What's holding you back from using Git?

---

This is the fourth video in my Claude Code series for GTM teams. Links to the previous episodes below, along with the full 31-min long YouTube version of this video which includes: 
- setup instructions
- protecting your API keys with .gitignore (!!!)
- using Claude Code to run Git commands for you

---

Previous episodes:
- 001 Setup: https://lnkd.in/gyWz4Fm6
- 002 Company Context: https://lnkd.in/g6M9WzbN
- 003 CLAUDE-dot-md: https://lnkd.in/gRCGjbuq

Full YouTube version: https://lnkd.in/gk_qNW_7

Claude Code isn't just for developers.
                                                              
At The Workflow Company, we've been using it daily for GTM work - research, competitive analysis, content, outreach, internal ops - and it's completely changed how we operate.
                                                              
Most people still think of it as a coding tool. It's not. It's an AI assistant that can actually take action on your computer. Write files. Search the web. Run scripts. Chain tasks together.
                                                              
It's a different dimension compared to chat interfaces.
                                                              
If you're in GTM and haven't tried it yet, I made this video to get you started. It's the first in a series where I'll share how we actually use it - not for coding (although we do a fair bit of it), but for the GTM work you're already doing.
                                                              
Claude Code is the single tool I'd recommend to anyone who wants to push the boundaries of what's possible in GTM right now. Hopefully this helps you get over the line.

What's your take on Claude Code? Would love to hear about your experience with it.

I try not to have regrets but here's a big one from 2025.

We went through a complete website refresh - new branding, new pages - last year. The design was great (Thanks team Odama), and we were really happy with it.

Since we already had a site up on Webflow, inertia took over and I made the decision to develop the new website on Webflow. Looking back, this was clearly the wrong decision.

Webflow and other low-code tools were amazing in the pre-AI era. They abstracted away the technical nuisances for the less technical users. Instead of writing code, you could click and drag your way around and build something semi-presentable.

But with the cost of writing code approaching an asymptote of 0, the cost of these no-/low-code abstractions is inversely approaching ∞. 

Why should a user need to learn the idiosyncrasies of a platform that is detached from code, when they can simply ask an LLM to create a coded solution?

The cost of this decision often appears in maintenance and iteration speed.

For us, practically speaking, developing our site on Webflow means...
- it takes longer for us to make tweaks to it
- it takes longer to experiment / iterate - spin up new pages, run A/B tests, etc
- it costs more to change anything (hiring a Webflow developer vs. using Claude Code)

The point isn't to specifically call out Webflow, but just use it as an example to illustrate a point that's becoming more and more obvious:

Given LLMs' current capability, the cost of abstraction continues to increase.

And the flip side: the value of code-based skills keeps compounding. 
Code is portable.  
Code is composable.
Code is extensible.
And perhaps most importantly, AI is really good at working with code.
Platform-specific knowledge is none of these things.

So here's the takeaway. If you're evaluating tools in 2026, ask yourself: 

Am I learning a siloed platform's quirks, or building with code that compounds and gives me leverage?

Initial thoughts on Claude Coworker, the Claude Code wrapper for non-technical folks:

> You can give Coworker access to a file directory with files you want to work with, and in the background it loads up a Linux filesystem sandbox with clear isolation from the rest of your system. Good safety/privacy feature without the user having to do anything on their end

> Coworker, as far as I can tell, does not yet have the ability to revert any changes it makes. I can already see a lot of tears shed over Claude accidentally deleting and overwriting files

> Because Coworker has access to your local file system, I see this first iteration being largely single-player. But what makes Claude Code awesome is that you can sync to git, and make it multi-player and collaborative

> Coworker's interface feels more session-based compared to Claude Code's directory-based. Meaning, Coworker would be great for working on a particular issue at the moment, but the work feels more disposable versus compoundable. Claude Code, with its directory-based approach, feels like it brings more permanence to the work you do, allowing you to iterate on the same set of files and build on top of your previous work

> Feels snappier than Claude Code; most likely has to do with the streaming text - Time-To-First-Token

Coworker will still feel immensely powerful for folks who have not started using Claude Code yet. I still believe that the UI layer is collapsing, and that's the case for Coworker as well. The polished interface compared to Claude Code's terminal is less overwhelming, but to someone used to Claude code, the UI still feels like friction.

Overall, I think Coworker could be a great transient entry point for non-technical folks to eventually graduate to Claude Code. The Anthropic team did mention they released this super early to iterate, so I'm keen to see how it evolves and whether it can stand on its own and have as seismic an impact as Claude Code has had.

We've built a GTM context repo (link to a post with more details in the comments), with outbound being the first use case. 

Here's how we're thinking about extending the application of the contexts.

> Horizontal
We've built the repo to store the contexts and do nothing else on purpose. Fully decoupled from the current application of the context = outbound. Because it's self contained and modularized, we can plug in the same repo to other use cases: 
1. content generation
2. qualification
3. enrichments

> Vertical
The repo is driving outreach sequence generation. For each prospect, we have a clear view into the contexts that were pulled, the resulting sequence and the rationale behind it. But it's not been tied yet to the outcome, i.e. response from the prospect or lack thereof. We want to close the loop, so that AI can help with pattern-matching:
1. Contexts are programmatically retrieved and composed
2. An outbound sequence is generated by an LLM based on 1
3. A positive reply to 2 is received
4. LLM analyzes 1, 2, and 3 and generates a hypothesis to test
5. Back to Step 1, the hypothesis from 4 is fed as a context

Personally really excited about step 4 - it essentially unlocks self-learning.

Claude Code literally 10x'd our iteration speed. 

We don't outsource thinking to it - we're still very much in control of the dialogue and the final decision. But it's great at pumping out different brainstorming ideas that we can apply our judgment against.

It really is as easy as feeding it existing context files and asking it to be a brainstorming partner. We talk through the ideas together, agree on the final output, which Claude generates. 

Another piece of context created that we can feed in a future Claude Code session for further refinement and iteration.

What did we do before CC?

Here's my roadmap for automating context repo updates.

(I'll never reach Stage 4 but that's ok.)

---

First - what's a context repo? A central place where you store everything an AI needs to know about your business. Positioning. Personas. Products. Voice.

Load it once, and you stop repeating yourself in every conversation with an LLM.

The most common question I get after people set one up: "How do I keep it updated?"

Because your business evolves. You refine your messaging after a big client win. You sunset a product line. Your ICP shifts.

Meanwhile, your context repo is stuck in the past. And every AI output built on stale context drifts further from reality.

Here's how I think about the progression:

---

1️⃣ Stage 1: I do, AI helps

I drive the updates. AI assists with the drafting.

"Hey Claude, I've changed my thinking on how we position ourselves. Pull up the last few client calls for context, then update the positioning doc. Show me what you changed before saving."

2️⃣ Stage 2: AI surfaces, I decide

AI monitors my raw context sources - call transcripts, LinkedIn posts, Slack threads - and surfaces drift and decay.

"I noticed your last 3 posts emphasize 'context as code' but your messaging doc doesn't mention it. Want me to suggest changes?"

I'm still in full control. But I'm not the one catching drift anymore.

3️⃣ Stage 3: AI does, I approve

AI makes changes autonomously - but staged, not live. It creates a pull request. I review and merge.

Full audit trail. Full veto power.

4️⃣ Stage 4: AI does

AI updates my context repo directly. No review. No approval. Full trust.

I don't think I'll ever get here. And I don't think you should either.

---

Here's why:

Context is the foundation of every AI output you produce. Content. Campaigns. Etc.

Get the context wrong, and everything downstream is wrong.

The goal isn't to remove yourself from the loop, but rather to remove the tedious parts - the monitoring, the drafting, the formatting - while keeping yourself where it matters: the judgment call.

Stage 3 is the sweet spot. You get 90% of the efficiency with 100% of the control.

Stop optimizing for full automation, and instead optimize for the right level of human oversight.

---

Where are you on this spectrum? Most are between 0 and 1.

What are the building blocks of a great cold outreach message?

I have been building a system to programmatically pull together contexts to generate an outreach sequence. Here are all the types of contexts that go into it:

-------------
Our contexts
-------------
1. Company
> Context about the company, its origin, mission, founders, funding history, etc.

2. Narrative
> The narrative - why now, why us - around the company and the product. More macro than messaging

3. Messaging
> Key messaging pillars, tone of voice, specific phrases to use/avoid

4. Product
> Product overview, both technical details and a JTBD framing

5. Channel Best Practices
> Email / LinkedIn / etc. - unwritten rules around best practices, accepted norms

6. Personas
> Various ICP personas that the company has a hypothesis on. Includes pain points, aspirations, JTBD, buying triggers, owned KPIs, etc. Only the relevant personas get pulled into the final composed context

7. Campaigns
> Campaign angles that specifically go deeper on all of the contexts above - for a specific persona in a specific industry, what product features and pain points do we highlight, what should our messaging tone be, etc. Only one campaign brief gets pulled in

8. Offers
> Different offers we can make in a cold outreach context. Should we ask for a meeting, can we send them resources, could we ask for permission to send them a personalized lead magnet? Offers we want to test get pulled in

--------------------
Prospect's contexts
--------------------
1. Account research
> Deep web-based research; disambiguation is tricky, and verifying all the facts are about the target company is most of the work

2. Contact professional experience
> Mostly based on LinkedIn data. What's their current role, career trajectory, what words do they use to describe themselves?

3. Contact social media presence
> Mostly LinkedIn - works great for active posters/engagers; not so much for lurkers

---

This feels fairly comprehensive to me, but I'm sure there are other considerations that I'm overlooking. So what else should be in this list of contexts?

One thing I think about a lot these days is the durability of our ideas, products, and services.

When ChatGPT 3.5 came out (Nov 2022, only just over 3 years ago) - all of us had this collective expansion of ideas and possibilities. With the help of LLMs, any idea seemed feasible.

Today, working inside of multiple Claude Code terminals for most of the day, every idea feels ephemeral. Most of us are probably feeling some version of this tension - as companies opt to build versus buy, and mature and seemingly resilient businesses go out of business overnight.

What is durable in this age of disposable software?

I've been sitting with this question, and here's where my thinking is landing:

Coding agents compress build time. What took weeks now takes hours. But they don't compress the human elements - trust, relationships, community, domain expertise. These still compound at a human pace.

The tempting advice is "just stay in the game long enough." But I don't think that's the answer - at least not for pure software businesses. If the moat is just code, someone can rebuild it tomorrow.

What seems more durable: businesses where human elements are architecturally intertwined with the software. 

Some examples:

► Marketplaces, where network effects are built on human trust - supply and demand that took years to cultivate. Can't clone that overnight.
► Forward-deployed engineering, where humans are embedded in the value prop itself.
► Service-as-a-software = human expertise + judgment + taste + software

Maybe that's why we're seeing the rise of services-enabled software, forward-deployed engineering as a discipline, and a rethinking of the "pure SaaS" playbook.

I'm still working through this, but I think the question isn't just "what should I build?" It's: how deeply are human elements woven into the business itself?

Curious to get folks' take on this. What feels durable to you right now?

Same workflow. Different clients. Different fit assessments.

We built a person research workflow that passes a client_id into every enrichment request.

client_id=wfco → pulls The Workflow Company context.
client_id=client_a → pulls Client A context.

The research agent sees the same person's LinkedIn activity. But evaluates fit against different criteria.

A technical founder posting about developer tools might be:
→ High fit for our dev tool client
→ Low fit for our enterprise sales client

One enrichment pipeline. Infinite context variations.

Why did we separate this out instead of hard coding our context?

Imagine I have multiple clients. I can pass in their IDs and look at the same person through different lenses. Someone who's not a good fit for The Workflow Company might be a great fit for a client of ours.

Building this composability into your workflows means you don't have to repeat yourself. You don't have to build separate enrichment workflows per client with the same logic duplicated everywhere.

Update one prompt, forget to update the others? Not a problem when the enrichment logic lives once and the context is the variable.

The context about your business should live in a programmatically accessible source. Your enrichment modules just point to it and pull whatever the latest version is.

p.s. the context database is doing a lot of the heavy lifting here - how it's structured, how data is stored, updated, and retrieved. Get that right and this architecture is easy to set up and maintain

My good friend Matteo Tittarelli ⚡️and Jared Waxman are back with the third cohort of GTM Engineer School. 

"GTM engineering" as a discipline is nascent and evolving quickly. If you've been on the fence and unsure of where to start, this course would provide a great starting point to ground yourself and understand the landscape of frameworks and tools - after which you can go on your own and start exploring areas that interest you. 

You'll be learning from some of the most forward thinking operators in the space, how they think about problems and how they use their favourite tools to solve them.

DM me if you're interested in taking the course - I've asked Matteo and Jared to tack on my affiliate fee as an additional discount.

taplio

The Battle

🦾Eric Nowoslawski

Joe Rhew

Followers

Posts

Engagements

Likes

Performance Score🦾Eric Nowoslawski's LinkedIn Performance

Performance ScoreJoe Rhew's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

🦾Eric Nowoslawski

Joe Rhew

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 ScoreJoe Rhew's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance Score🦾Eric Nowoslawski's LinkedIn Performance