Author
M. Patel writes DevMoment field notes on AI dev workflow, tested on real work rather than demos.
Comparisons of Claude Opus, Sonnet and Haiku score capability, speed and price per token. None of them mention that the model string is part of your request contract. Swap it and thinking can turn itself on, your answer can move out of content[0], and a max_tokens you never touched can start truncating. Eight measured differences, September 2026.
A proxy changes where Claude Code's requests go. It also changes which of its own safety nets are still running. The credential variable is a header choice, the base URL alone does not move your billing, and a gateway context limit turns off auto compact recovery. Field log, September 2026.
Claude Code's /fast is not a speed dial, it is a mode. The first toggle in a conversation reprices your whole context, the toggle moves your model one way and does not move it back, and asking for fast is not the same as getting it.
The desktop app and the CLI run the same engine and read the same files. On three points they resolve them in a different order, and Anthropic's own docs call one of them a departure from the CLI scope hierarchy. Read 10 September 2026.
Eleven of fifteen documentation sites serve a real llms.txt. Zero of the five major AI crawler operators document reading yours. Measured on 8 September 2026 with a plain HTTP client, including the commands, plus the one part of the file that does pay off.
ACP and MCP sit on opposite sides of the agent: MCP connects an agent to tools, ACP connects an editor to the agent. The spec calls itself MCP-friendly and passes your MCP servers inside session/new. The catch is the transport floor: stdio is mandatory, HTTP and SSE are optional capabilities.
OpenCode Go's own page states usage limits as requests per five hours, per model, across fourteen models running 110 to 45,300. The two third-party pages that rank for the term describe the same plan as a three-tier dollar budget on a lineup that has largely turned over. A field log on which source to trust and why.
Agent teams are off by default in August 2026, and turning them on also changes how ordinary subagent delegation behaves. A documentation field log on the env var everyone writes wrong, the 7x token figure nobody quotes correctly, and the five minute cache bucket.
Every comparison ranking for this query calls goose Block's open-source agent. Block donated it to the Agentic AI Foundation at the Linux Foundation in April 2026, alongside Anthropic's MCP. A field log on what that changes and what it does not.
Every page-one comparison treats this as a choice between two products. Google's own model docs say Antigravity offers Claude Sonnet 4.6 and Opus 4.6, free tier included. That makes it a harness comparison, not a model one, and it changes the answer.
I built the same rate-limiting feature twice, both inside Warp. Warp's own docs argue you should switch off Claude Code, then document running it inside Warp. Here is what that actually means for a solo dev in 2026.
I ran the same task in Aider and Claude Code on the same model. The code converged; the workflow did not. The real split is where your undo lives: git or the agent.
