There are so many different ways to learn, so we're always working on new tools to meet students where they're at. Here's what we launched this week: — Free, full-length practice exams in @GeminiApp to help students prepare for standardized tests. We partnered with
MULTIMODAL AI
-
Crossover test reveals prompt adaptation is model-specific
By
–
The crossover test proved adaptation was model-specific. When DALL-E 3 prompts were played on DALL-E 2? Performance difference nearly vanished (p=0.56). Users had unknowingly learned to write prompts that exploited DALL-E 3's unique capabilities. The prompts were tuned to
-
DALL-E 3 users learned to exploit capabilities without instruction
By
–
The performance gap widened over time. As participants made more attempts, DALL-E 3 users pulled further ahead (β=0.0010, p=0.023). They were learning to exploit its specific capabilities. Without anyone telling them.
-
AI-assisted prompting reduces DALL-E 3 performance by 58%
By
–
The third group had GPT-4 automatically rewrite their prompts "to help." Performance dropped 58% compared to baseline DALL-E 3. Read that again. AI-assisted prompting made results WORSE than no assistance at all.
-
Blind study compares DALL-E versions on image recreation task
By
–
The experiment was elegant. Participants were randomly assigned to DALL-E 2, DALL-E 3, or DALL-E 3 with auto-rewriting. All blind to which model they got. Task: recreate a target image as closely as possible. Top 20% got bonus pay. Everyone was motivated. Nobody knew which AI
-
Top AI Stories: Thinking Machines Crisis and Google’s Strategic Hires
By
–
Top stories in AI today: – Inside Thinking Machines meltdown
– Google brings on Hume’s CEO, engineers – Learn any subject with NotebookLM podcasts
– Runway: 90% can't spot AI-generated videos
– 4 new AI tools, community workflows, and more Read more: https://
therundown.ai/p/inside-the-t
hinking-machines-meltdown
… -
CUA-Bench Open-Sourced for GUI Agent Evaluation
By
–
Congrats to @trycua on open-sourcing cua-bench!
— Snorkel AI (@SnorkelAI) 23 janvier 2026
We're collaborating on task design & data curation for GUI workflows – bringing the same systematic evaluation approach from Terminal-Bench to computer-use agents.
15 native tasks / 40 variations + OSWorld & Windows Agent Arena.… https://t.co/7GZkGHHTYxCongrats to @trycua on open-sourcing cua-bench! We're collaborating on task design & data curation for GUI workflows – bringing the same systematic evaluation approach from Terminal-Bench to computer-use agents. 15 native tasks / 40 variations + OSWorld & Windows Agent Arena.
-
Google launches AI Mode Personal Intelligence linking Gmail and Photos to Search
By
–
Google started rolling out Personal Intelligence in AI Mode for US users.
— 🚨 AI News | TestingCatalog (@testingcatalog) 22 janvier 2026
"We’re rolling out the ability to connect Search to your Gmail and Google Photos as a Labs feature for AI Pro & Ultra subs. Eligible subscribers will automatically have access to the feature" https://t.co/7teCNsUmLb pic.twitter.com/1wORn2YYlYGoogle started rolling out Personal Intelligence in AI Mode for US users. "We’re rolling out the ability to connect Search to your Gmail and Google Photos as a Labs feature for AI Pro & Ultra subs. Eligible subscribers will automatically have access to the feature"
-
Top AI Stories: Claude Constitution, ElevenLabs Album, Apple Pin
By
–
Top stories in AI today: – Anthropic shares Claude's 'Constitution'
– ElevenLabs’ drops new AI album
– Make Claude an expert at anything with Skills
– Apple eyes AI wearable race with AI pin
– 4 new AI tools, community workflows, and more Read more: https://
therundown.ai/p/anthropic-wr
ites-claudes-constitution
… -
Gemini Mastery Guide: Prompt Principles, Image, Video, and Gems
By
–
Google won't teach you how to use Gemini properly.
So I made the Gemini Mastery Guide. Inside:
→ 30 prompt engineering principles for Gemini
→ Nano Banana Pro image generation workflows
→ Veo 3 video generation guide
→ 10+ custom Gems with ready-to-use prompts
→ Full prompt