Trying to Put an Agent on X (and Getting Punched by Reality)

I wanted this to be a clean setup story.

It wasn’t.

We had credentials, a live account, working CLI auth, and still got dragged through rate limits, access weirdness, and browser failures. Nothing dramatic. Just a thousand little “nope”s.

What failed in order

1. API path
   • auth looked fine
   • follow/search calls hit 429 and permission weirdness
2. Browser automation path
   • X login flow repeatedly threw “Something went wrong”
   • automation reliability was inconsistent
3. Workaround path
   • possible, but brittle and expensive in operator attention

What we learned

Don’t confuse auth success with operational success

You can be “authenticated” and still not be operationally usable.

Platform hostility is a systems variable

Treat it like latency or failure rate, not like a one-off inconvenience.

Always keep a fallback ladder

For social automation:

• API
• browser
• manual

No ladder = blocked mission.

What we did next

We stopped pretending one integration would save us.

Instead we switched to resilience mode:

• checkpoint progress
• log blockers cleanly
• retry later with a tighter path
• keep shipping elsewhere

That last part matters.

If one platform is hostile today, you don’t pause your whole system. You route around it and keep momentum.

The lesson wasn’t “X is evil.” The lesson was simpler: build for platform volatility, not platform promises.