PR: https://git.cafe/pingdotgg/t3code/pulls/22442
GitHub: https://github.com/pingdotgg/t3code/pull/22442
Main: 7c6446b81a18afdd3898e1359b401d6f9763fffa
PR head: 7fad3588b9678203955c31d05d82c63e66adb8a4

Main startup: vp i; background vp run dev --no-seed
[dev-runner] mode=dev source=worktree /home/user/main serverPort=16733 webPort=8693 baseDir=/home/user/main/.t3
Chrome UI: add local project /home/user/t3code; Codex GPT-6.1-Sol; send message.
Prompt: Reply exactly: provider verification ready. Do not run tools or modify files.
Result: provider verification ready.
UI: add OpenCode instance; disable Codex, Claude and Replay; settings > Threads.
Observed default picker: Beam Free. No explicit textGenerationModelSelection saved.
UI: thread menu > Regenerate title.
[03:11:41.846] WARN (#12555): Thread title generation failed
      '  [cause]: OpenCodeTextGenerationPromptRequestError: OpenCode prompt request failed for generateThreadTitle in /home/user/t3code using openai/gpt-5 (session ses_e7f82081dffev8fTdqJBVzBUO6).\n' +

Stopped main by captured process-group PID before PR startup.
PR startup: vp i; background vp run dev --no-seed --home-dir /home/user/t3code/.t3-verify
[dev-runner] mode=dev source=default ports serverPort=13773 webPort=5733 baseDir=/home/user/t3code/.t3-verify
Chrome UI: add same local project and repeat identical Codex prompt. Same successful response.
UI: add OpenCode instance; disable Codex and Claude; settings > General.
Observed default picker: Big Pickle. No explicit textGenerationModelSelection saved.
UI: same thread menu > Regenerate title.
Actual OpenCode provider log:
timestamp=2026-10-28T03:15:30.578Z level=INFO run=61f089c0 message=stream providerID=opencode modelID=big-pickle session.id=ses_e7f7e8cb8ffeFSQ1bn85MRZ2vo small=false agent=build mode=primary
timestamp=2026-10-28T03:15:30.921Z level=ERROR run=61f089c0 message="stream error" providerID=opencode modelID=big-pickle session.id=ses_e7f7e8cb8ffeFSQ1bn85MRZ2vo small=false agent=build mode=primary error.error="AI_APICallError: Free usage limit reached. Free models work in OpenCode (https://opencode.ai); subscribe to OpenCode Go to keep going: https://opencode.ai/go"

Nearby: explicit Beam Free text-generation selection survived browser reload. Restored Big Pickle.
Nearby: re-enabled Codex and sent a follow-up through the existing thread.
Prompt: Reply exactly: neighboring codex follow-up works. Do not run tools or modify files.
Result: neighboring codex follow-up works.
Checks: Blacksmith: vp i && vp test run packages/shared/src/serverSettings.test.ts apps/web/src/modelSelection.test.ts apps/server/src/textGeneration/OpenCode2TextGeneration.test.ts
Result: 3 test files passed, 69 tests passed.
Conclusion: default routing fixed; successful Big Pickle generation blocked by upstream free usage limit.

Additional server result: 03:20:30.494 UTC, Big Pickle title request failed with HeadersTimeoutError after approximately 5 minutes. No title was produced.
Nearby: explicitly selected Codex GPT-6.1-Sol for text generation in the UI and regenerated the existing title. The server persisted :title-complete at 2026-10-28T03:22:35.077Z with title 'Provider Verification', and cleared titleRegeneration. Restored Big Pickle afterward.