Please reconsider killing Custom GPTs.
A Skill is not a Custom GPT, and capability is not continuity.
I understand moving technology forward. Architectures change and platforms migrate. But normally, we establish functional parity before decommissioning the system people already depend on.
That is what concerns me here.
Custom GPTs gave us persistent, purpose-built assistants. Many of us spent years developing their instructions, knowledge, behaviors, personalities, tools, and interaction patterns. Users learned how to communicate with those assistants, developed shorthand with them, and built real continuity over time.
Skills are powerful—I build and use them myself. But a Skill provides a capability or behavior. It does not currently reproduce the persistent assistant environment that a Custom GPT provides.
That continuity is functionality.
OpenAI’s own migration guidance acknowledges that the replacement isn’t identical: conversations don’t migrate, sharing doesn’t automatically migrate, selected models don’t transfer, Custom Actions require additional work, and migrated assistants may behave differently.
That isn’t functional parity.
This isn’t an argument against Skills or Plugins. I want them to succeed. But we shouldn’t remove working functionality before the replacement can reproduce it.
Build the replacement. Demonstrate functional parity. Let creators validate that what they spent years building actually survives. Then retire the legacy architecture.
In technology, we normally build the new bridge before tearing down the old one.
Doing it in the opposite order isn’t moving forward.
It’s a step backward while we wait for the new architecture to catch up.