Agent WorkflowsHarness engineering is the missing layer of coding agentsCoding agents are not just models with a terminal. Their reliability depends on the instructions, tools, context boundaries, hooks, and checks around the model.September 5 2026//2 min read#harness engineering#coding agents#context engineering#verificationOpen post→
Developer ToolsMTPLX made local LLM speed feel like a product featureMTPLX is a native Apple Silicon runtime for local models that makes speed, exactness, and API compatibility part of the same workflow.September 5 2026//2 min read#mtplx#local ai#apple silicon#mlxOpen post→
Developer ToolsSWE-bench stopped being enoughWhen a benchmark stops separating frontier coding systems, the repo needs local evals more, not less.April 8 2026//3 min read#benchmarks#coding agents#evalsOpen post→
Agent WorkflowsClaude 4 made long-running agent work a planning problemClaude 4 pushed the conversation toward sustained coding and agent workflows, which raises the bar for supervision.May 27 2025//3 min read#claude#coding agents#reviewOpen post→