MCP sucking is a harness problem, not an MCP problem
MCP unlocks behavior that is fundamentally impossible to get via CLI or APIs
Bad auth, too much context usage, all get solved with an execution layer - your agent writes code to progressively discover and call tools
@tunguz The @perplexity_ai app is ok, but the real strength is in their API offerings. Models such as sonar-reasoning-pro are showing noticeably stronger performance in web-grounded data retrieval and synthesis.
@psomkar1 Yes, It was for an internal data review app (can’t share publicly). Next.js on the frontend, Python backend, completely built using Agent mode in Github Copilot and then switched to Codex CLI with no manual code changes.
We built layers of software for decades.
Dashboards.
Workflows.
Approvals.
Abstractions on top of abstractions.
And we called it progress.
#ClaudeCode#Codex@opencode