Watch on YouTubeBoris says more general models tend to win as models improve exponentially in capability.
Claude can now be used on GitHub, according to Boris.
Boris uses Claude Code to review his Git commits and report what he shipped, and says it makes TDD easier to use in practice.
Boris, a technical staff member at Anthropic and the creator of Claude Code, gives a short demonstration of how he uses coding models in daily work. He says more general models tend to win as their capabilities grow exponentially. He mentions that Claude can now be used on GitHub. For his weekly stand-up, he asks Claude Code what he shipped that week, and the tool checks his Git commits so he does not have to remember each change. Boris also discusses test-driven development. He has found TDD difficult to use consistently, but says coding tools make it work much better in practice. He suggests this may be because the model handles the process instead of him. The clip is brief and focuses on two concrete workflows rather than showing detailed implementation steps.
Boris introduces himself as a member of technical staff at Anthropic and the creator of Claude Code. He connects the tool to a broader claim about model progress: more general models tend to win, while model capability increases exponentially. The introduction is brief, and he moves quickly from that claim to examples of how he uses Claude Code in his own work.
Boris says Anthropic announced that Claude can now be used on GitHub. He presents this as one of several areas connected to the increasing capability of general models. The clip does not explain the integration or show its setup, but it establishes GitHub as a place where he expects Claude to take part in coding work.
Boris describes a weekly stand-up workflow. He asks Claude Code, "what did I ship that week?" The tool looks through his Git commits and tells him what he shipped. This removes the need for him to keep track of every change himself. The example is grounded in a routine task rather than a special demo, since Boris says he does it pretty much every day on Monday.
Boris says he tries to use test-driven development, but finds it hard to apply in practice. He now thinks coding tools make TDD work much better. His explanation is that the model may be doing the process rather than him. He does not describe a specific test or repository, so the claim stays at the level of his own workflow experience.
Boris contrasts his own difficulty with TDD against the results he gets from coding tools. He suggests that the improvement comes from assigning the work to the model. The point is practical: a workflow he finds hard to maintain himself becomes easier when the coding model carries it out. He ends after making this observation, without giving a longer account of how the model performs the TDD steps.
"The more general model always wins and the model increases in capability exponentially."00:00