The Semantic Dispute

• The statement "coding is solved" by Boris—a software developer who reportedly stopped writing code by hand in November 2025—is only defensible if one defines coding strictly as typing code or turning a spec into a pull request.5:58
• If coding is defined broadly as end-to-end software engineering, the claim collapses; critical tasks like strategic planning, architectural integrity, and long-term maintenance remain human-centric.6:35

Verification and Tooling

• The visible failure in the Claude Code Desktop app, where update text was truncated due to poor UI rendering, highlights that code review cannot replace the actual verification of a running product.2:51
• Desktop Electron apps are structurally difficult to verify at scale because they often require per-thread graphical virtual machines, whereas web applications permit simpler parallel testing.16:51
• Improving agent reliability requires investing in verification infrastructure, such as remote dev servers via Tailscale and data snapshots, rather than relying solely on raw code-generation capability.24:30

Strategic Implementation

• Agents operate best as "tactical tornadoes" that execute well-specified changes, but they currently lack the autonomous judgment required to detect bugs outside of their immediate prompt parameters.30:27
• Significant engineering overhead, such as the thousands of lines of code added to T3 Code for test support, is often necessary to provide agents with the reliable feedback loops required to prove their work.