En 30 segundos
In this February 2026 short, Matthew Berman humorously contrasts Claude Opus 4.6 with GPT-5.3-Codex. He portrays the former as making narrow edits and the latter as investigating and changing far more of a codebase, then says both approaches work well. The short is a personal comic characterization, not a measured comparison. It shows no actual task, code changes or timing data.
- 1A narrow-edit caricatureBerman depicts Opus 4.6 as selecting the most relevant files and returning after a few minutes.
- 2A wide-ranging caricatureHe jokes that GPT-5.3-Codex examines everything, changes every file and returns with an unfamiliar codebase.
- 3Both can workHis conclusion is that both styles work well; the short provides no benchmark or reproducible coding test.
Conclusiones
- Berman humorously portrays Opus 4.6 as focusing on files directly relevant to a task.
- He exaggerates GPT-5.3-Codex’s investigation and changes for comic effect.
- He says both approaches can work well, without presenting a controlled test.
Lee el texto editado por capítulos
Two coding assistants
Berman introduces the two models as having different approaches. The setup is a humorous comparison.
OpenAI and Anthropic have released new coding models, and I want to joke about how differently they can approach a task. This is a humorous impression, not a demonstration of an actual coding task or a benchmark.
Opus and a focused edit
He acts out a narrowly scoped response from Opus 4.6. The characterization is not a measured performance claim.
I portray Claude Opus 4.6 as saying it will make the changes that are most relevant to the request and return in a few minutes. It is a caricature of a focused approach rather than a promise about how the model always behaves.
Codex and a broad investigation
He exaggerates the scope and time of a Codex task. He concludes that both approaches work well.
For GPT-5.3-Codex, I exaggerate in the other direction: it researches everything, touches every file and returns half an hour later with an unfamiliar codebase. The joke is about a much broader approach; I do not show a real run or measurements to support those specifics.
The funny part, in my experience, is that both approaches can work well. The short does not establish a winner or a general rule about how either model changes code.
Más de este canal
34:21EN
Matthew Berman· 27 oct 2026Matthew Berman Examines Anthropic’s IPO Filing, AI Growth and GovernanceMatthew Berman, Sam Altman · Forward FutureThe presenter reviews the S-1’s financial metrics, compute strategy and voting structure while questioning the tension between rapid AI development and safety.
0:36ENMatthew Berman· 25 oct 2026Kimi K3.1 and a Rubik’s Cube Solver DemonstrationMatthew Berman · Forward FutureMatthew Berman shows a scrambled virtual cube being solved and gives the Kimi K3.1 test a favorable assessment.
0:21ENMatthew Berman· 24 oct 2026An AI-Generated Kimi Benchmark Chart Used Invented ScoresMatthew Berman · Forward FutureMatthew Berman asks for a graph of Kimi K3.1 Pro benchmarks, then discovers that the displayed figures are illustrative placeholders.
25:50ENMatthew Berman· 24 oct 2026Kimi K3.1 Pro Tested: Strong 3D Demos, Uneven Results and Competitive PricingMatthew Berman · Forward FutureMatthew Berman tests the open-weight model on coding, 3D rendering and visual explanations, then compares benchmark results and prices.
32:20ENMatthew Berman· 23 oct 2026Who Is Local AI For? The Costs, Trade-offs and a Hybrid FutureMatthew Berman · Forward FutureMatthew Berman weighs data control and independence against hardware cost, setup complexity and inference speed.
1:03:37ENMatthew Berman· 23 oct 2026Matthew Berman’s Team Tests Kimi K3.1 on Animation, Games, and ChartsMatthew Berman, Alex, Brian · Forward FutureA live team AMA turns into an informal test of Kimi K3.1, from personal-agent workflows to motion graphics, a browser game, and a benchmark chart.