News AI
Latest news about AI-assisted development, OrkestrAI updates, and industry insights.
A Repository Is Not an Instruction
A repository is full of instructions, but not all of them are for the agentThe fastest way to make an AI codin…
Human Gates for Irreversible Agent Actions
Speed is not the same as permissionAI coding agents are most useful when they remove toil from bounded work, a…
Dependency-fix agents still need a human with the last word
The best place for an agent is a narrow, reviewable scopeDependency updates are one of the best places to use …
Coding agents need test harnesses, not faith
Benchmarks are not the same as proofAI coding agents can write code, but a software team still needs proof tha…
You Can't Govern What You Can't See
Visibility is the real control planeThe biggest risk in AI-assisted development is not that agents are too fas…
Agent PRs Still Need Strict Human Review
The diff is not the decisionAgent-generated pull requests are a throughput story, but review is still a judgme…
Coding agents need human policy, not just a prompt
The prompt is not the policyAI-assisted development is no longer a toy autocomplete box. Coding agents can ins…
Code Review Rules Keep Humans in Charge
Agents can write more code, but they should not decide aloneThe turning point in AI-assisted development is no…
Human Approval Is the Control Plane
The best agentic workflow is boring on purposeThe most useful AI-assisted development workflows are not the fl…
OpenCode, the terminal agent that speeds up senior engineering work
Why terminal agents matter nowSenior engineers do not usually lose time because they cannot type fast enough. …
Coding Evaluations Need Human Judgment
When a benchmark becomes a product decisionEvaluation is not a side quest in AI-assisted development. It is th…
When AI Debugs, Humans Must Stay in Control
AI debugging is not AI diagnosisThe appeal of coding agents is obvious when a bug is blocking a team: they rea…
MCP Needs a Human at the Controls
Every new MCP server is a new permission boundaryThe appeal of MCP is easy to understand: a model that can rea…
Reviewing Agent PRs Without Giving Up Human Judgment
Review is now the scarce skillAgents made code cheaper to produce, and review became the new bottleneck. GitHu…
Keep Coding Agents Under Control
Bounded autonomy is the real upgradeThe fastest way to make an AI coding agent useful is not to let it do ever…
The After-Code Gap: Why Humans Still Need to Own AI Output
The after-code gap is where AI still needs humans AI coding tools are getting good at the part everyone can se…