At the plumbing level, yes; at the capability level, no. The cost of switching sits in the harness, not the interface. ENTRY 123 records a measured result: the same Opus 4.6 differs by 4.5 points when only the harness changes (ledger note: Bustamante, 5/3, single source).
The mainstream view says protocol compatibility makes vendors swappable at will. The diary parts ways here: compatibility only pushes plumbing cost to near zero, and plumbing was never a moat. Models are post-trained against a specific tool surface, so a new shell leaks capability. A product whose moat is "multi-model support" has a moat of zero.
DeepSeek quietly promoted V4-Pro to general release (ENTRY 123, 8/13; scope note: the announcement was small enough to need community digging, and 0813 weights are not open), and Claude Code can switch by changing base_url. In a measured test the same Opus 4.6 differs by 4.5 points when only the harness changes (Bustamante, 5/3, single source). Lindy's migration took 6-9 months, happened before protocol compatibility, and counts only as an upper bound.
For founders: before switching, build a regression benchmark on your own workflow and bind post-training to your own tool surface; if you only do routing and compatibility, find a second leg now. For investors: ask one diligence question, namely what the team uses to measure how much was lost after the swap; if there is no answer, count "can switch vendors anytime" as zero in the valuation.