Daily AI Brief

Thursday, November 12, 2026

Top stories

Google launches a low-cost API for AI decisions

Google introduced the Gemini Decisions API for bounded tasks such as choosing tools, routing requests, and checking content against rules. It evaluates up to 1,000 questions against a shared input and returns probabilities rather than generating conversational answers. Google lists pricing at $0.10 per million input tokens, with no output-token charges; its speed and performance figures are vendor-reported.

Why it matters: Routine automation may become cheaper when developers use a specialized decision model instead of a full conversational model.

Source: Google — Gemini Decisions API · November 11, 2026

DeepMind releases AlphaGenome 2 for cell-specific genetics research

Google DeepMind announced AlphaGenome 2, which predicts how DNA variations affect particular human cell types. Trained on more than 100 million cell profiles, it identified nearly twice as many gene-expression-linked variants as its predecessor in the company’s cited evaluation. Access is available for noncommercial research, with commercial access in private preview; the model is not validated for clinical diagnosis.

Why it matters: Researchers could better prioritize experiments by identifying which cells a genetic variation is likely to affect. That is a research opportunity—not evidence of a new treatment or diagnostic test.

Source: Google DeepMind — AlphaGenome 2 · November 11, 2026

Databricks expands Genie One from answering questions to doing work

Databricks announced expanded Genie One capabilities combining enterprise search, generated dashboards and presentations, and actions across connected workplace tools. Administrators can configure permissions, approval requirements, tracing, and spending controls. Availability varies by feature, so the announcement should not be read as universal immediate access.

Separately, Databricks announced plans to invest approximately $3 billion across Europe and the Middle East over three years, including engineering expansion and AI training.

Why it matters: The enterprise competition is shifting toward assistants that complete workflows—with governance becoming as important as model capability.

Source: Databricks — Genie One · November 11, 2026

OpenAI’s reported revenue jump comes with substantial losses

Reporting based on investor documents puts OpenAI’s third-quarter revenue at approximately $10 billion, up nearly 50% from the previous quarter. Adjusted operating losses reportedly reached approximately $10.9 billion, despite an improved operating margin. These are reported investor figures, not publicly audited results.

Why it matters: Commercial demand can grow rapidly without resolving the cost of delivering AI. Revenue growth alone does not establish a sustainable business.

Source: Forbes — OpenAI revenue and losses · November 11, 2026

Bottom line

Today’s developments point toward more specialized AI: inexpensive decision engines, research-specific models, and workplace agents with explicit controls. The practical test remains measurable results—not broader capability claims.

Compiled with AI assistance and reviewed by Dan Kardell.

← All briefs