DEV Community

brian austin
brian austin

Posted on

Ghostty left GitHub. Your AI might be next.

Ghostty left GitHub. Your AI might be next.

This week, Ghostty — the fast terminal emulator — left GitHub. The HN thread hit 3,000+ points and 900+ comments. That's not just a GitHub story. It's a platform dependency story.

The thread title says it all: "GitHub is no longer a place for serious work."

Ghostty's maintainer moved because they didn't trust that GitHub would keep the rules the same. And they were right not to trust blindly. Platforms change. Terms change. Pricing changes.


The Pattern Is Always the Same

  1. Platform offers free or cheap access to build your workflow around
  2. You build the workflow. Deeply.
  3. Platform achieves lock-in
  4. Platform changes terms

We've seen this with:

  • GitHub (now owned by Microsoft, adding AI features to your repos)
  • Twitter → X (API pricing went from free to $100/month overnight)
  • Heroku (killed the free tier with 30 days notice)
  • npm (rate limits, registry disruptions)

And now, the same pattern is playing out with AI.


Your AI Access Has the Same Problem

If your team's workflow depends on ChatGPT or Claude.ai:

  • OpenAI just added ads. Sponsored recommendations in ChatGPT. Your AI is now serving advertiser interests alongside yours.
  • OpenAI just joined Amazon Bedrock. Great for enterprises with AWS budgets. Worse for individual developers who now face another layer of IAM, VPCs, and AWS pricing complexity.
  • Claude.ai went down last week. For hours. Anyone whose workflow depended on claude.ai was blocked.

The platform owns your workflow. When the platform hiccups, you're stopped.


What Ghostty Did About It

Ghostty self-hosts its code and community. They don't need GitHub to exist. GitHub can change its terms tomorrow — Ghostty keeps working.

This is the right instinct. Own the layer you depend on.


What Developers Are Doing About AI Platform Risk

The equivalent move for AI is to build against the raw API — not the consumer product.

Here's the difference:

# Platform-dependent (consumer product)
# You log into claude.ai and type
# If it's down, you're down
# If they change the interface, your workflow breaks
# If they add ads, you see ads

# API-dependent (you own the integration)
import anthropic

client = anthropic.Anthropic(api_key="your-key")

message = client.messages.create(
    model="claude-opus-4-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Review this code for security issues"}]
)

print(message.content[0].text)
Enter fullscreen mode Exit fullscreen mode

When you call the API directly:

  • You choose when to upgrade models
  • You choose the interface (terminal, IDE plugin, custom app)
  • You handle your own uptime with retries
  • No ads, no sponsored recommendations
  • Anthropic's business model is API revenue — not advertising

The Cost Argument

The reason most developers stay on the consumer product is cost. Claude.ai Pro is $20/month. That's the platform tax.

But API access doesn't have to cost $20/month.

SimplyLouie gives you Claude API access for $2/month — flat rate, no per-token billing anxiety, no usage caps, no ads. It's what I use when I want the API without the enterprise pricing.

The math:

  • ChatGPT Plus: $20/month (with ads now)
  • Claude.ai Pro: $20/month
  • Direct Anthropic API: pay-per-token, unpredictable bills
  • SimplyLouie API: $2/month flat

The Lesson from Ghostty

Ghostty's maintainer didn't wait for GitHub to change the rules. They moved proactively, while it was still their choice.

The developers I see getting hurt by AI platform changes are the ones who built deeply into a consumer product — a specific interface, a specific conversation history, a specific feature — and then the platform pulled it.

Build against the API. Own the integration. Keep the platform at arm's length.

That's the Ghostty lesson applied to AI.


SimplyLouie is a flat-rate Claude API at $2/month. 7-day free trial. simplylouie.com

50% of revenue goes to animal rescue.

Top comments (0)