🧬 AlphaGenome 2 doubles DNA variant detection for disease research 🧬
Free non-commercial access to sharper, cell-specific DNA variant predictions.

Google DeepMind has released AlphaGenome 2, a new version of its genomics model that predicts how changes in DNA affect specific cell types. In the lab's tests, AlphaGenome 2 doubles DNA variant detection for disease research, identifying twice as many variants as the original model. It can also adapt to new cell states without retraining, so researchers can study tissues or conditions the model has never seen without building a custom version. The model is available for non-commercial use, opening it to academic labs and research institutes worldwide, and that capability jump gives smaller labs a sharper lens on disease mutations. For drug discovery and rare-disease research, faster variant interpretation could shorten the path from sequencing data to testable hypotheses.

⚡️ Gemini Decisions API ships for developers ⚡️
Fast, calibrated image-and-text decisions with simple API integration.

Google has launched the Gemini Decisions API, a developer tool built to return fast, calibrated decisions from combined image and text inputs. With this release, the Gemini Decisions API ships for developers who need structured answers rather than open-ended chat replies. Google's own example shows the system looking at a photo of a spill, assessing its severity and recommending cleanup actions, a judgment call common in facilities management, retail and manufacturing. "Calibrated" means the model signals how confident it is, which matters when outputs trigger real-world actions, and it raises competitive pressure on vision-AI startups selling similar decision tools. Expect early adoption in operations-heavy industries where speed and reliability count more than eloquence.

🐾 Scammers use AI-faked photos to target lost pet owners 🐾
A new scam turns the panic of losing a pet into an opportunity for fraud.

According to Digital Trends, scammers are now using AI image generators to create fake photos of missing pets, targeting owners at their most vulnerable moment. Public "lost pet" posts on social media give fraudsters everything they need: a photo of the animal, a description and a desperate owner ready to respond. With generative AI, producing a convincing image of that same pet is now a matter of seconds. It is a stark example of how cheap, realistic AI imagery is lowering the barrier for emotional scams. The practical takeaway: treat any unsolicited "we found your pet" photo with caution, ask for a live video call and never send money upfront.

INTERESTING TO KNOW

🌐 Google Antigravity agents get browser use 🌐

Google has started the rollout of browser use in Antigravity, its agentic coding platform, letting agents research, write and verify code live in a browser. With this update, Google Antigravity agents get browser use and can run multiple tasks at once, while user-controlled authentication and download approvals keep humans in charge of sensitive steps. The move adds competitive pressure on rival coding agents like Cursor and Claude Code.

🤝 Grok Bot 1.0 lands in Slack, Teams 🤝

SpaceXAI has opened access to Grok Bot 1.0, a proactive agent that acts without waiting for prompts and keeps a shared organizational memory across a company. Grok Bot 1.0 lands in Slack, Teams and video meetings, and adds publishable Team Bots plus admin controls for IT departments. Its workplace distribution puts it head-to-head with Microsoft Copilot and Slack AI, signaling a serious push into enterprise.