Co-Founder at Expnd | Building go-to-market teams & systems for tech startups | Certified Clay partner

We build and ship GTM systems for SaaS companies - no slide decks, no fluff, just real infrastructure that moves pipeline and unlocks revenue.

Our work isn’t consulting or campaigns. We set up the automation, AI signals, data flows, and live reporting your team actually needs to scale. If you’re seed to Series B and want to get out of spreadsheet hell and manual lead routing, that’s what we fix.

We’ve worked inside sales teams, taught ourselves to code, and now engineer operating systems for SaaS founders who care about measurable results, not just good ideas.

If you want a team that ships, not one that talks - let’s chat.

Claude Code > n8n > Clay

This was due for a while.

Haven't touched n8n in about 2 weeks now. And I'm faster, better, building more complex stuff.

But I don't think people realize how big this shift is.

Going from Clay to n8n was trading a spreadsheet for a workflow engine.

Going from n8n to Claude Code is trading a workflow engine for a software engineer.

You just describe what you want. And it builds it. (Databases, APIs, full systems...)

I think this might be the biggest jump in tooling I've seen since I started.

PS. I do think using n8n in the early days and being able to use that as context in how we think/ setup claude code was huge in this migration. You shouldn't outsource your thinking to it.

PPS. In NYC next week, still have some open evenings if anyone wants to grab a drink.

Laurens Nys

Founder @ GTM Sigma │ GTM Engineering for B2B SaaS

Just spend a week in NYC as a Belgian founder. 

Here's what it taught me about GTM: 

Nothing. 

Beautiful city though, def going back there in the near future.

Context engineering for GTM: progressive disclosure

Most people build AI workflows by feeding their agents docs after docs after docs. 

Assuming more context equals better answers. 

But LLMs suffer from context rot. 

There is a cognitive "smart zone" for agents (usually around 50k–60k tokens). 

Once you bloat the context window beyond that, the AI's attention degrades, hallucinations spike, and reasoning falls off a cliff. 

If you've ever had a long conversation with an agent and wondered why it suddenly started getting worse, that's context rot. 

Enter progressive disclosure:

We need to optimize for relevance.

aka showing the AI exactly what it needs to solve a specific problem, and absolutely nothing more.

Here's one way to solve this;
(core idea:  structuring the knowledge base so the agent discovers information in deliberate tiers, rather than loading everything at once)

1. Atomic files: Breaking knowledge down. Instead of 20-page SOPs, each file carries exactly one specific piece of information or one core claim. 

2. Maps of Content: Using central hub notes that act as routing layers, describing to the AI what it can find in different conceptual neighborhoods.

3. Descriptions as retrieval filters: Every file has a short description. The agent loads just the descriptions first. This acts as lossy compression, allowing the AI to ask: "Does this warrant loading the full file?" before it commits to spending context window space on it.

TL;DR

Stop force feeding your agents entire hard drives of unstructured GTM docs. 

Treat the context window as a highly curated attention span. Use progressive disclosure to let the agent filter the noise before it commits to reading the depths.

PS. Nice side effect, this creates a beautiful graph if you do it right.

taplio

Laurens Nys's Linkedin Analytics

Laurens Nys 🔵

Performance ScoreLaurens Nys 🔵's LinkedIn Performance

What is Laurens talking about?

Laurens Nys 🔵's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?