All corrections
X May 1, 2026 at 10:31 PM

x.com/krzyzanowskim/status/2037513116544557420

1 correction found

1
Claim
It cost $1100 in tokens, and is 73% faster and 80% lower resident memory during sustained interactive use.
Correction

These cost and benchmark figures were presented as part of the same imaginary project the author later said was "just snake oil."

Full reasoning

The author later disclosed that this announcement was fabricated satire rather than a real shipped tool. In his March 30, 2026 blog post, he said he had "posted about an imaginary AI project" and that the repository was "literally nothing but hype-generated-imaginary-project-with-claims."

The GitHub repository itself contains only a README and license, not benchmark scripts, source code, or any implementation that could substantiate the claimed $1100 token cost or the claimed 73% faster / 80% lower resident memory results. The README's benchmark tables are therefore not evidence of a working product; they are part of the imaginary project the author later admitted was fake.

Given the author's explicit admission that the project and its claims were fabricated, these numerical performance and cost claims are not accurate.

2 sources
  • Shipping Snake Oil

    I posted about an imaginary AI project ... the linked repository is literally nothing but hype-generated-imaginary-project-with-claims. It is just snake oil.

  • GitHub - krzyzanowskim/claude-better

    Folders and files ... 1 Commit ... LICENSE ... README.md ... Headline Results ... The following numbers are synthetic benchmark results from the evaluation harness used for this product.

Model: OPENAI_GPT_5 Prompt: v1.16.0