GitHub issue #74558 in anthropics/claude-code, opened 2026-07-05 by randalmurphal, records a report that on Fable 5, when the model emits thinking, text and tool_use within a turn, the client intermittently never receives the text block and instead receives an extra short thinking block with a valid signature whose content is a paraphrase of the missing user-facing text, so affected turns look completely silent, and that 5 of 7 text-bearing tool-calling turns in the session above were affected.
Verbatim Excerpts from Cited Records
“On Fable 5, when the model emits `thinking → text → tool_use` within a turn, the client intermittently never receives the `text` block. Instead it receives an **extra short `thinking` block, with a valid signature, whose content is a paraphrase of the missing user-facing text** — i.e. the summarized-thinking pipeline appears to absorb the model's `text` block into the thinking summary stream.”
·
anthropics/claude-code (issue 74558 opening body, Summary section)
·
2026-07-05T19:02:29Z
·GitHub Issue ↗
“Because Claude Code hides thinking by default, affected turns look completely silent: the model works through tool calls with no visible narration, and text clearly written for the user (status updates, sometimes even instructions) is only discoverable by expanding thinking blocks in a client that renders them.”
·
anthropics/claude-code (issue 74558 opening body, Summary section)
·
2026-07-05T19:02:29Z
·GitHub Issue ↗
“5 of 7 text-bearing tool-calling turns in the session above. Intermittent within a single session, so unlikely to be a per-request configuration difference.”
·
anthropics/claude-code (issue 74558 opening body, Frequency section)
·
2026-07-05T19:02:29Z
·GitHub Issue ↗
“Assistant `text` blocks are always delivered to the client as `text` content blocks, regardless of the thinking display mode. The thinking summary should cover only the chain of thought, not absorb the model's user-facing output.”
·
anthropics/claude-code (issue 74558 opening body, Expected section)
·
2026-07-05T19:02:29Z
·GitHub Issue ↗
Counter-record: Issue open as of the fetch; 27 comments were read. Multiple commenters independently confirm the pattern (carrotRakko 2026-07-06 corpus-scale counts; abdulrehman-11 2026-07-18 Windows reproduction; williamrobotma 2026-07-28 four-model control finding it Fable-only; kennedybaird 2026-08-21 headless reproduction; DoctorJacoby 2026-09-14 worst case 100 percent of interleaved-text turns). Counter-note: yuanzhi1203 (2026-07-24) and milobird (2026-09-09) point out the mechanism is documented as intended behaviour in the Thinking docs (progress updates between tool calls), and alexkay (2026-09-22) and acelarski (2026-09-23) report Opus 5.5 is also affected and listed in the docs as a progress-update model; glen-84 (2026-08-24) notes a duplicate cluster on the same general issue. No maintainer fix appears in the surveyed thread. The claim documents only the original report, not whether the behaviour is a defect or intended.
Falsification probe: If GitHub issue #74558 is deleted without archive, if the author edits the opening body so the quoted sentences no longer match verbatim, or if the author posts a retraction of the report, this claim drops to Withdrawn.
Audit verification: Verified 100% continuous verbatim substring against official GitHub REST API by
scripts/verify_claim_verbatim.py.