“Still, Claude Code ain’t perfect, or even close.”
Our reading, the sentence that was checked:
Claude Code is not perfect or even close to it.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, The biggest advance in AI since the LLM, paragraph 12.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-the-biggest-advance-in-ai-since-the-4645c2-a2
We recorded 11 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2026-04-11. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
From the speaker
- Claude Code ain’t perfect
- or even close
Our interpretation
- The ordinary reading is that Claude Code retains substantial practical limitations and recurring failures, rather than merely one rare defect.
- No numerical accuracy threshold is implied or invented.
- Because this is an ongoing observation with no named period, use March 2026, the latest complete pre-post calendar month, as the observation window.
- Reports published after the post may establish March behavior if available by the evaluation date.
Counts as supported if: Documented recurring failures or substantial practical limitations in Claude Code during the target month show that users could not rely on consistently correct, unattended operation.
Counts as contradicted if: Broad, representative evidence establishes consistently near-perfect practical operation across the relevant uses and shows that alleged material failures were erroneous, isolated or outside the target period.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
Anthropic’s April 23 postmortem identifies a Claude Code defect introduced on March 26, 2026. After an idle-session threshold, it repeatedly discarded reasoning history, causing forgetfulness, repetition and odd tool choices [e1]. A firsthand report published April 2 records user-defined stop-hook interventions on multiple days from March 8–25, including 43 on March 18 [e2]. Together, these document recurring practical limitations during March and support the claim under the contract’s qualitative reading of “not perfect or even close” [e1, e2].
Limitations
- The evidence does not establish a representative failure rate across Claude Code users or tasks.
- The stop-hook counts measure user-defined workflow interventions, not incorrect code; the report’s proposed cause is not independently established [e2].
- Anthropic reports fixing the defect on April 10, 2026; this evidence does not establish that it persisted afterward [e1].
Evidence (2 items)
Supports the claim: Anthropic,
forgetfulness, repetition, and odd tool choices
Why it matters here: Anthropic traces recurring problems to a change shipped March 26 that repeatedly discarded reasoning history after an idle-session threshold. This documents a material product limitation during the target month.
- Publisher
- Anthropic
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: stellaraccident / GitHub,
43 violations
Why it matters here: The user's log analysis reports repeated workflow interventions rather than one isolated mistake. These are user-defined stop-hook violations, not a general coding-error rate, and the report's proposed cause is not independently established.
- Publisher
- stellaraccident / GitHub
- Kind of source
- Independent measurement
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, The biggest advance in AI since the LLM
Still, Claude Code ain’t perfect, or even close.
- Where
- paragraph 12
- Published
- 2026-04-11
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-the-biggest-advance-in-ai-since-the-4645c2-a2: Current. Outcome: Supported. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-the-biggest-advance-in-ai-since-the-4645c2-a2.