Why everyone HATES GPT-5 (and how to fix it)

Value of the Information & Strength of the Argument

The video provides substantial value by directly testing the most common complaints against GPT-5, offering concrete examples and comparisons. Wolfe’s approach is methodical: he recreates user-reported issues, compares outputs across multiple models, and evaluates the validity of each gripe. His argumentation is balanced, acknowledging both valid criticisms (e.g., coding performance, accuracy) and those he considers less valid (e.g., personality, based on his tests). He also contextualizes the model routing issue, explaining how a broken router could have made GPT-5 seem dumber. However, the analysis is limited by its anecdotal nature; the tests are not comprehensive or statistically rigorous, and personal experience may not generalize. Wolfe also speculates on OpenAI’s motivations (e.g., cost-saving) without direct evidence, which weakens the argumentative rigor.

Scientific Rigor, Source Quality, Title Accuracy

The video references several sources, including Reddit posts, tweets from Sam Altman, and comments from AI experts like Ethan Mollick, but these are presented without formal citations or links in the description. The description includes links to FutureTools.io, a newsletter, and a sponsor (Globant), but no direct references to the cited claims. The title accurately reflects the content, which is a critical review of GPT-5. The video’s scientific rigor is moderate: it uses real-world tests but lacks controlled conditions and peer-reviewed sources. The adéquation between title and content is strong, as the video indeed explores why people hate GPT-5 and suggests fixes. However, the lack of verifiable sources for the claims made (e.g., benchmark charts) reduces the overall reliability.

The title accurately reflects the content, which systematically addresses common criticisms of GPT-5 and explores potential fixes.

7/10

The video provides a balanced, hands-on evaluation of user complaints about GPT-5, testing claims with direct experiments and referencing community feedback. However, the analysis is anecdotal and lacks rigorous methodology or independent verification.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:57Overview of all gripes against GPT-5
• 04:36Gripe 1: Removal of old models and model routing issues
• 09:32Gripe 2: Personality comparison between GPT-5 and GPT-4o
• 16:09Gripe 3: Coding performance test with Balatro clone
• 24:19Gripe 4: Accuracy tests with logic puzzles
• 33:09Final analysis and summary of gripes

• FutureTools.io — Matt Wolfe's AI tools and news platform, mentioned as a resource for exploring AI tools.
• Weekly Newsletter — Sign-up for Matt Wolfe's weekly AI newsletter.
• Globant Enterprise AI — Sponsor segment promoting Globant's AI services, not directly related to the video's content.
• Matt Wolfe's LinkedIn — Social media profile for Matt Wolfe.
• Matt Wolfe's Threads — Social media profile for Matt Wolfe.

• Reddit post on GPT-5 accuracy — User-reported logic puzzle failure, consistent with the video's test results.
• Sam Altman's tweet on router issue — Acknowledgment that the auto-switcher broke, supporting the video's explanation for perceived dumbing down.

• OpenAI's official benchmark claims — OpenAI's published benchmarks suggest GPT-5 is superior, but the video's tests show mixed results, indicating a discrepancy between official claims and real-world performance.