Beware AI Vendor Lock-in: Encrypted Reasoning Tokens Strip User Control
mitsuhiko · x · 2026-07-30
Mitsuhiko shared an in-depth article criticizing the growing trend of vendor lock-in by AI companies. The post points out that early inference APIs followed a simple input-output model, allowing users to fully own their conversation transcripts and migrate them to other models.
However, modern inference APIs are increasingly stripping users of control through opaque mechanisms:
- Encrypted reasoning tokens: Billed to the user but returned as opaque, encrypted blobs with useless summaries.
- Hidden context: Web search results and subagent instructions are hidden from the client application.
- Non-portable references: File, vector-store, and cache references are entirely keyed to the provider's servers.
These changes mean the local transcript is no longer the actual "session," reducing it to an unreadable fragment. The article argues this is detrimental to both users and the broader AI ecosystem.
More from coding & agent
- Open-source AI coding tool jcode boots 245x faster than Claude Code with ultra-low RAM — Shruti_0810 · 2026-07-30
- Neuracore: Fixing the Data Plumbing Bottleneck in Robot Learning — stepjamUK · 2026-07-30
- Gemini Robotics ER 2 in Action: Sub-second Bidirectional Streaming & Multi-Robot Orchestration — _philschmid · 2026-07-30
- Banning AI Agent Outputs in Code Reviews to Prevent Outsourced Thinking — WesEklund · 2026-07-30
- Using a Second Model to Review Code: A Workflow to Boost AI Coding Quality — haltakov · 2026-07-30
- Cursor's Guide to Cloud Agent Readiness: 2026 Will Be the Year Agent Work Moves to the Cloud — vinvan · 2026-07-30