DoubleChecks

What was claimed

fable is better than you in every branch of software [...] it has been trained on all the code of the internet, it knows almost every trick in the book [...] unless you are 99th percentile expert in some niche domain of software, it simply is not possible [to beat it or be more creative]

Our verdict

Inaccurate

Anthropic states Fable was trained on a proprietary mix of publicly available internet data, public and private datasets, and synthetic data, not literally "all the code of the internet.". No cited evidence supports this universal claim about human performance or creativity. Benchmark results only evaluate specific tasks, and some models and humans can outperform Fable in particular settings.

Combined assessment19 sourcesChecked Oct 17, 2026

Why this verdict

  • Unless you are a 99th‑percentile expert in some niche domain of software, it is not possible to beat it or be more creative.

    Incorrect
    Two assessments reached this conclusion; one differed·ChatGPT: Can’t verify

  • Fable has been trained on all the code of the internet.

    Incorrect
    Two assessments reached this conclusion; one differed·Perplexity: Can’t verify

  • Fable is better than you in every branch of software and knows almost every trick in the book.

    Misleading
    Two assessments reached this conclusion; one differed·Perplexity: Incorrect

  • Fable knows almost every trick in the book.

    Can’t verify
    One assessment reached this conclusion; one differed·Claude: Misleading

Evidence reviewed

19 sources

Perplexity, Claude and ChatGPT sources are combined here.

usefollowed.comanthropic.comsnorkel.aisnorkel.aiawesomeagents.aidatacamp.comarxiv.orgdatacamp.comlindy.aisnorkel.ai

Individual assessments

Assessment summary

  • The response makes sweeping claims that are not supported by the available evidence.
  • Benchmark results show Fable can be strong on some coding tasks, but other models outperform it on several evaluations, and no evidence confirms that it was trained on all internet code or is more creative than nearly all software experts.usefollowed.com anthropic.com snorkel.ai

Claims reviewed

Unless someone is a 99th-percentile expert in a niche software domain, they cannot beat Fable or be more creative than it.

Incorrect
No cited evidence supports this universal claim about human performance or creativity. Benchmark results only evaluate specific tasks, and some models and humans can outperform Fable in particular settings.anthropic.com snorkel.ai lindy.ai

This claim may mislead readers without added context.

Fable is better than you in every branch of software.

Incorrect
Available comparisons show mixed results rather than universal superiority. Opus outperforms Fable on several shared benchmarks, while other tests show Fable winning or tying.anthropic.com snorkel.ai

This claim may mislead readers without added context.

Fable knows almost every trick in the book.

Can’t verify
This is a vague, non-testable claim. Coding benchmarks can measure selected capabilities but cannot confirm near-complete knowledge of software techniques.anthropic.com snorkel.ai

Assessment summary

  • The statement makes several absolute claims that are contradicted by evidence.
  • While Fable is indeed a highly capable model, the claim that it is "better than you in every branch of software" is an unfounded absolute.
  • Research shows AI systems have well-documented limitations in complex architecture, novel problem-solving, and business context understanding.
  • The claim about training on "all the code of the internet" is also false—Fable has a knowledge cutoff date and cannot access real-time information.

Claims reviewed

Unless you are 99th percentile expert in some niche domain, it is not possible to beat it or be more creative

Incorrect
Research shows AI systems cannot think critically or innovate in the way humans do. source unavailable Creativity in coding involves problem-solving, design, and originality—areas where AI is fundamentally limited. source unavailable source unavailable Humans at all skill levels can outperform AI in novel problem-solving, architectural design, and creative innovation, not just 99th percentile experts. source unavailable source unavailable

This claim may mislead readers without added context.

Fable has been trained on all the code of the internet

Incorrect
Fable 5 has a knowledge cutoff date (January 2026 for Fable 5, June 2026 for Fable 5.1) and cannot access real-time code repositories or information beyond these dates. support.claude.com Additionally, no AI model has access to literally all internet code due to access restrictions, private repositories, and proprietary codebases.

This claim may mislead readers without added context.

It knows almost every trick in the book

Misleading
AI systems rely on pattern recognition from training data rather than genuine comprehension. arxiv.org They struggle with unseen scenarios, novel problems, and lack intuition and common sense. arxiv.org source unavailable The phrase 'almost every trick' overstates capabilities and misrepresents how AI systems actually work—through memorization and pattern matching, not true understanding.

This claim may mislead readers without added context.

Assessment summary

  • The response mixes some true points about Fable's strong software performance with exaggerated and verifiably false claims.
  • Anthropic documents say Fable was trained on a proprietary mix of public, private, and synthetic data, not "all the code of the internet" anthropic.com anthropic.com.
  • Performance claims about being better "in every branch" or impossible to beat except by a 99th‑percentile expert are unsupported and misleading given available benchmarks and red‑team results anthropic.com arxiv.org.

Claims reviewed

Fable has been trained on all the code of the internet.

Incorrect
Anthropic states Fable was trained on a proprietary mix of publicly available internet data, public and private datasets, and synthetic data, not literally "all the code of the internet." anthropic.com anthropic.com

This claim may mislead readers without added context.

Fable is better than you in every branch of software and knows almost every trick in the book.

Misleading
Fable shows strong software engineering performance in benchmarks and safety testing, but claims of superiority "in every branch" and omniscience are broad, subjective, and unsupported by comprehensive cross‑domain evidence. anthropic.com arxiv.org

This claim may mislead readers without added context.

Unless you are a 99th‑percentile expert in some niche domain of software, it is not possible to beat it or be more creative.

Can’t verify
No public, authoritative measurement defines a "99th‑percentile" threshold or proves human experts cannot outperform Fable across all tasks; comparative performance varies by task and context. anthropic.com arxiv.org

Check your own claim

Paste any statement, headline, or AI answer. Independent assessments check it and compare sources.

Share this result