Reviewers see Google as a fast, accurate default for finding information and getting everyday work done, with repeated praise for how smoothly Search, Gmail, Drive, Maps, Docs, and other tools work together across devices. The main complaint is familiar: privacy, data collection, ads, and broader trust concerns. A smaller but clear frustration is weak direct support and some rough edges in specific products, like dated map data or clunky Drive actions. Founders also highlight Googleโs strong AI and developer ecosystem, from search and image models to Chrome and Drive integrations.
Hey Product Hunt ๐
"AI coworker" has been one of the hottest categories on PH all year, and today Google Cloud shipped its own take. At Gemini at Work 2026 they announced the Gemini agent: one agent for questions, knowledge work, images and media, and code, all from a single prompt box.
You don't give it instructions - you give it an objective. Then you close your laptop, and it keeps working.
1. Describe the outcome you need (a deck, a report, a pipeline fix)
2. Gemini plans the work and picks the right skills, tools, and model
3. It pulls context from your systems: Workspace, Microsoft 365, Slack, Jira, Salesforce, BigQuery, Snowflake or any MCP server
4. It hands back finished work inside your docs, inbox or dev environment
Some announced features:
Coworker agents get their own @ agents email, calendar, Drive and directory presence, and only see what you share with them
Runs in the cloud, so long jobs keep going for hours or days, with one memory across web, mobile, desktop and the command line
Multi-model under the hood: Gemini models plus Claude from Anthropic, with smart routing to keep costs down
Works inline in Gmail, Docs, Sheets, Slides, Chat and Calendar, and proactively suggests tasks you can hand off
Every agent gets its own identity, audit trail and sandbox, plus real-time spend caps per project
Industry versions in preview for Financial Services and Legal
Really excited to try it out, you can learn more details here.
the part that'd worry me running this for real is "spins up coworker agents with their own email and calendar" plus "hard spend caps" in the same sentence. what happens when a job hits the spend cap mid-task, does it downgrade to a cheaper model and keep going quietly, or stop and surface that it ran out of budget. silent downgrade is the dangerous version because the output still looks finished, it's just worse, and nothing in the deliverable tells you the budget ran out before the task did