The week the coding verdict flipped
We changed our agentic coding pick for the first time in four months, OpenAI set a hard shutdown date you can't ignore, and Google quietly matched a capability everyone said was a moat. Nine things mattered this week. Two change how you should work.
47 announcements scanned · 9 mattered · 2 verdicts changed
Verdicts changed
Best for agentic coding
Claude Opus 4.8
31% fewer dead-end loops on long-horizon tasks in our pipeline runs. Opus 4.8 remains our pick for review-heavy work. This is a split, not a dethroning.
Best budget model for agents
Gemini 2.5 Pro
Native tool use landed and held up in our reliability runs, at half the per-call cost.
Act before next issue
GPT-4 Turbo shutdown: July 1, three weeks out
Migrate to GPT-4.1. Drop-in for most uses; function-calling payloads need one change.
Details & stepsOn our radar
Mistral's agent runtime beta
Flips our EU-hosting verdict if latency claims hold. Testing now, verdict next issue.
Cursor pricing change
Per-seat → usage-based. Could change our team-tooling math for small teams.
Filtered out
- MCP hits 100M downloads: milestone, no decision attached.
- Two "GPT-5.5 next month" rumors. We don't publish rumors.
- 14 new tool launches that duplicate existing picks without beating them.
Our take: The coding-model race has stopped being about raw capability and started being about who fails less on hour three of a task. That's why this week's verdict change matters more than any launch since February.
Neomanex, from our own production runs
Get issue 13
Next week's verdict changes and deadlines, in your inbox.