# Cold Email Drip: Condense Error Logs

Target ICP: AI engineers, LLM app developers, MLOps specialists, cost-conscious startups using Claude API.

## Email 1: Pain Hook

**Subject:** Your error logs are burning $$ on API calls

Hi [name],

Just watched your app drop a 50KB error dump into Claude for analysis. 

You know what that costs? If you're feeding verbose logs to Claude 500 times a month, you're burning ~$400 on context that Claude doesn't actually need to read.

The errors matter. The 10KB of noise around them? Doesn't.

Worth 2 minutes to see how much you're leaving on the table?

[condense-error-logs.com]

Wes

---

## Email 2: The Gap

**Subject:** Re: Your error logs

Fast follow up.

Most error analysis tools are cloud-only. Log aggregation, analysis, piping to LLMs. You need the vendor. You need their infrastructure.

Condense runs locally. One binary. Filters verbose noise from your logs *before* they hit your API. No signup, no rate limits, no vendor lock.

Just: input verbose log, output signal only.

Smaller context = faster responses + lower bills.

[See how]

Wes

---

## Email 3: Real Number

**Subject:** Case: $380/month saved

Hi [name],

One of our users, a 4-person AI team building on Claude:

- Before: 60KB average log context per API call, ~800 calls/month = $480 in context costs
- After: 4.2KB filtered output, same 800 calls = $32 in context costs
- Savings: $448/month. Filters built in 10 minutes.

Their biggest gap? Kubernetes get -o yaml dumps. 90% noise, 10% signal. One filter rule.

Your logs probably have the same pattern.

[Run the audit]

Wes

---

## Email 4: Why Now

**Subject:** Claude's context window is expensive this way

Quick context:

You probably have one of two approaches right now:
1. Pipe raw error output to Claude and hope it finds the signal (expensive, slow)
2. Pre-process logs by hand (time tax, fragile)

Condense sits in the middle. Pluggable filters for each command. Your team writes them once, runs them forever.

Not a replacement for your logging infra. A lightweight layer between you and the API.

[Learn how it works]

Wes

---

## Email 5: Low-Risk Start

**Subject:** 15 min to prove it

Hi [name],

No pressure, but if you're building on Claude and your error logs are >10KB on average, you almost certainly have $200-400/month of waste sitting in your repo right now.

15 minutes to build one filter. See the difference.

If it saves you even $50/month, it paid for itself in 3 weeks of focus.

[Get started]

Wes