Founders Insight

Most "GTM databases" are just contact directories flexing "50M companies"
→ but how many are actually active?

Why account tiering matters:
Horizontal SaaS serves everyone. You need to prioritize based on timing, activity, and traits - not surface filters.

The problem:
Techno-graphics work for qualification, not timing. They don't show what's changing, why it matters, or who's involved.

Static databases → wasted credits on inactive accounts.

Companies move in time.
Last month's non-priority could be Tier 1 today (like adopting usage-based billing).

Agents don't think in filters - they think in meaning.
They combine world knowledge with user context to craft semantic queries based on intent, not keywords.

The problem: The industry is stuck translating natural language into pre-set SQL filters. Rigid schemas. Static filters.

Our agent-native data layer: Vector embeddings + semantic matching for company activities:

• What functions are they building?

• What migrations are happening?

• Who's joining and leaving? Static filters for deterministic data:

• Headcount, location, funding stage The bet: Agents are smart enough to query both static and semantic fields. We're building for Agents that reason around meaning.

We're studying GTM timing, movements, and real-time signals that reveal pain-points.

One thing became obvious: People and Company APIs are still CRUD operations.

They weren't built for GTM search or insight discovery.

Getting to meaningful insights is slow, inefficient, and fragmented.

The shift: Many are refactoring code-bases to make them agent-ready. The same shift is coming to GTM.

A data layer for agentic search looks nothing like CRUD APIs.

Converting natural language to API params repeatedly is wildly ineffective. We're just masking text-to-SQL as search.

Inference providers like Groq unlock two shifts:

1. LLM-native architecture We replaced deterministic code with LLMs from day 0.

Example: Natural language blocklists. Instead of hardcoded lists, we use Groq + web search in real-time. "Exclude marketing agencies" just works.

2. Rapid prototyping with model swaps

• OSS models for intermediate reasoning

• SOTA models for complex reasoning

v1 fast > perfection. Optimize later based on usage.

The insight: Orchestrate a hierarchy of models - cheap/fast for most flows, expensive/smart only when needed.

With competitors appearing weekly, timely and precise outreach wins.

If you're using static filters - even technographics - you're too late or off on timing.

Funding signals are commoditized. Everyone has them, reacts, and it's signal slop.

You need creative timing and insight - leading indicators of pain with nuance.

Great GTM teams spot live, nuanced movements:

• Job posts revealing plans

• Website/pricing changes

• Hyper-specific layoffs

• Sub-departments static too long

Find leading indicators of pain.

That's where timing wins and GTM teams succeed.

Assume models won't get better Building your moat around "GPT-5 can't do X yet" is a death sentence.

Your defensibility must be orthogonal to model capabilities.

1. Treat evals as performative Without rigorous evaluation frameworks, you're flying blind. Evals are your early warning system.
   
   

2. Vibe code production features Ship fast, not recklessly. In AI products, trust compounds slowly and evaporates instantly.
   
   

3. Skip in-person at SF The density of AI talent, customers, and capital in SF is unmatched. Serendipity still matters.
   
   

4. Never build a data moat Proprietary data and feedback loops are real differentiators.
   
   

5. Treat non-AI infrastructure as secondary Brilliant LLM + broken integrations = customers leave.

Which mistake do you see most often?