Gemini’s overnight takeover

Gemini 3 Is the New Benchmark

When a model launch makes the entire AI industry say “oh no…” you pay attention. Google’s Gemini 3 dropped last week and it’s already shaking up the leaderboard. 

What happened? The model launched directly into Google Search (a first for Google), shot to the top of LMArena rankings, and crossed 1 million users within its first 24 hours across AI Studio and the Gemini API.

What’s different? Early reactions were unusually unanimous. 

• From the top: Sam Altman congratulated the team. Elon Musk chimed in. And Salesforce CEO Marc Benioff, who has used ChatGPT every day for three years, said two hours with Gemini 3 “changed everything.”

• Across the board: Users pointed to speed, reasoning, and video/image quality as the biggest jumps. On LMArena, testers say Gemini 3 Pro is showing a “clear lead” in high-skill categories like coding, math, and creative writing.

The standalone strengths: According to LMArena CTO Wei-Lin Chiang, Gemini 3’s agentic coding abilities “in many cases now surpass top coding models like Claude 4.5.” And the crowdsourced evaluations don’t just show incremental improvement—they show a meaningful gap.

Even Insurers Are Tapping Out of AI Risk

Major insurers want authority to exclude AI-related liabilities from corporate policies. Great American, Chubb, and W. R. Berkley have all asked regulators for permission to step back from covering AI mishaps. One underwriter told the Financial Times the models are simply “too much of a black box.”

Why they’re spooked:

• Google’s AI Overview hallucinated legal trouble in March → $110M lawsuit.

• Last year, Air Canada was forced to honor a chatbot-invented discount.

• A deepfaked executive tricked a firm into losing $25M in 2024. 

The real fear: systemic failure, not single incidents. Insurers can absorb one $400M hit. They can’t handle 10,000 simultaneous claims because one widely used AI model glitches.

Run coding agents side by side

emdash is an orchestration layer for running multiple AI coding agents in parallel, isolating each one in its own Git worktree so you can launch, monitor, compare, and review their work from a single interface.

How you can use it:

• Spin up multiple agents to tackle the same task and compare diffs

• Track real-time progress and assign issues across providers

• Review changes, open PRs, and manage experiments safely

• Stress-test workflows by “pitting” models against each other

Pricing: free

Instant infographics from plain text

Genaraera turns simple prompts or pasted data into professional, template-free infographics by automatically selecting layouts, charts, colors, and typography for you.

How you can use it:

• Convert raw data or a reference image into HD visual assets

• Generate branded graphics in multiple aspect ratios

• Speed up reports, lessons, posts, and presentations

• Use the API to automate infographic generation at scale

Pricing: paid