Price cuts did what benchmarks couldn't
Google cut Gemini 2.5 Pro pricing by 40% and flipped our budget long-context verdict without shipping a single new capability. Meanwhile the Mistral runtime verdict we promised last issue is in, n8n 2.0 will break your community nodes if you let it, and Zapier's agents went GA with pricing nobody can compute yet.
41 announcements scanned · 7 mattered · 2 verdicts changed
Verdicts changed
Budget long-context pick
A 40% input-price cut puts full-corpus passes within reach of mid-size budgets; our long-document pipeline cost dropped proportionally with no quality regression.
EU-hosted agent runtime
Two weeks of production traffic confirmed the latency claims for short-horizon agents; long-horizon loops still trail US-hosted options.
Act before next issue
n8n 2.0 breaks community nodes on auto-upgrade
Pin your 1.x version before your next deploy — failures surface at workflow runtime, not startup.
Details & stepsGPT-4 Turbo shutdown — July 1, two weeks out
Final call: if you haven't migrated to GPT-4.1, this is the last comfortable window.
Details & stepsOn our radar
Zapier Agents GA pricing — Capable product, uncomputable cost. We publish a verdict when agent-task accounting is public.
Cursor usage-based pricing fallout — First full billing cycles land this month — we're collecting real team invoices before revisiting the verdict.
Filtered out
- Three frontier-lab teaser posts with no dates, no benchmarks, no pricing — we don't publish vibes.
- An 'AI browser' launch that is a Chromium fork with a sidebar.
- 11 tool launches duplicating existing picks without beating them.
Our take: Two of this week's verdict changes came from pricing pages, not model cards. Capability parity at the frontier means the deciding variable has moved to cost per useful token — and vendors know it, which is why price moves now land with launch-day choreography.
— Neomanex, from our own production runs
Get issue 14
Next week's verdict changes and deadlines, in your inbox.