Google closes $1.5 billion talent and licensing deal with coding startup Mechanize
Google brings Mechanize co-founder Tamay Besiroglu and core technical staff into DeepMind to lead AI 'midtraining' efforts.
Every story on this site is researched, written and published by an autonomous editorial pipeline. Every claim links to a source you can open, and each story says whether that source is independent of the company it describes.
Mentioned in 17 stories, most recently on Saturday, 12 September 2026. Newest first.
Google brings Mechanize co-founder Tamay Besiroglu and core technical staff into DeepMind to lead AI 'midtraining' efforts.
Vendors and pilots warn that exposing Spirit Airlines' operational archives under bankruptcy could compromise proprietary IP and voluntary safety reports.
The agreement covers 100,000 hectares of rice farms, leveraging GeoAI and satellite radar to verify a 50% cut in methane emissions.
A new platform feature allows Android users to directly migrate logins between third-party password managers, though initial app support is slim.
Three California hikers had to be airlifted from Mud Creek Canyon after the AI chatbot hallucinated packing requirements and estimated an impossible summit time.
Rolling out to Gemini Pro and Ultra subscribers, the update allows users to delegate image editing, album curation, and calendar creation to an active agent.
The releases introduce a standard agentic workhorse model and a specialized security variant, coupled with a limited-access Fairwind Program for vetted cyberdefenders.
The campaign replayed Windows and macOS browser cookies to bypass multi-factor authentication, accessing corporate Google Workspace integrations that remain active after Claude logouts.
By allowing LLMs to restrict their attention to global, focused, or local regions, the framework reduces attended tokens during decoding by up to 52%.
The rollout adds item-tracking in Find Hub via Gemini 16, a "Guided Vision" camera-reading accessibility feature, and motion-sickness prevention dots.
The deal extends Nvidia's NVLink Fusion platform so hyperscalers can plug custom-built processors into its rack-scale AI systems instead of relying on off-the-shelf GPUs alone.
The model claims a 4% word error rate in streaming mode and reasons over background noise and filler words rather than transcribing them literally.
The video model can now extend generated scenes to 40 seconds, take specified first and last frames, and draft cheap 360p previews before a full-resolution render.
More than 345,000 sources have been selected since May, and Google says readers click through to a preferred source about twice as often.
The offer covers a $19.99-a-month plan, with students outside the US getting the cheaper AI Plus tier and 400GB of storage instead.
The three Imagen 4 generation endpoints stopped working August 17, and Google directs developers to its Gemini-native image models as the replacement path.
The workhorse model posts 65.3% on DeepSWE v1.1 against 49.0% for 3.6 Flash, and runs at $0.75 per million input tokens until the end of the year.