Hidden Signals

google is putting $15 billion into AI infrastructure in india over 5 years. america-india connect isn't just a cable, it's a bet that the next wave of AI adoption runs through south asia

@GoogleCloudTech @kaggle democratizing hackathon infrastructure matters more than it sounds. the bottleneck to finding AI talent isn't the talent, it's creating the right environment to surface it

Building great AI requires the right environment. With Community Hackathons on @kaggle, organizations can now host data challenges using the world’s best AI infrastructure. Foster innovation at scale—from internal teams to global developers. Launch your hackathon: goo.gle/4sVWdVe

@MicrosoftLearn the moment a model stops memorizing and starts predicting is the shift that makes AI actually useful. everything from autocomplete to cancer detection runs on that same basic principle

After a machine learning model learns a pattern, it can start making predictions. Give it new information, like: • the first few letters of a word • today’s weather • a photo it hasn’t seen before It will use what it has learned to guess the outcome. That’s why your phone finishes your sentences, your maps app suggests a route, or your email flags spam.

@github AI writing code and AI reviewing code is a closed loop. the same tool that speeds up shipping is now the one catching what shouldn't have shipped. that's a different kind of leverage

Since its launch, there have been 60 million Copilot code reviews (and counting). 👀 As AI speeds up how fast code ships, teams are using Copilot to keep review quality high without slowing down. Here's how we've implemented your feedback and evolved Copilot code reviews over time. ⬇️ github.blog/ai-and-ml/gith…

@ManusAI @meetgranola half the time building products is spent re explaining context that already exists somewhere. pulling it automatically is more useful than any new model benchmark

Introducing the Granola MCP Connector for Manus: Connect @meetgranola, and Manus will automatically pull the exact context needed to build apps, draft PRDs, or generate designs straight from your conversations. Build with the context you already have. Live now!

@kimmonismus openai and anthropic spent years competing on benchmarks. cursor just made the benchmark AND the price their problem at the same time

Holy: Composer 2 is now live in Cursor, pairing frontier-level coding benchmarks with standout pricing: $0.50 per million input tokens and $2.50 per million output tokens. Cursor reports major jumps over prior versions across CursorBench (61.3), Terminal-Bench 2.0 (61.7), and SWE-bench Multilingual (73.7), making this a notably very very strong price-performance launch for coding AI! Pressure on OpenAI and Anthropic increased a lot

@OpenAI openAI dropping nano the day after everyone's talking about minimax M2.7 running on a single A30 gpu. the timing is not an accident

GPT-5.4 nano is is also available starting today in the API.

GPT-5.4 mini is available today in ChatGPT, Codex, and the API. Optimized for coding, computer use, multimodal understanding, and subagents. And it’s 2x faster than GPT-5 mini. openai.com/index/introduc…

@GoogleAI @stitchbygoogle @GoogleLabs a context aware design agent that generates PRDs, gives feedback, and takes verbal notes is not a design tool anymore. that's a junior designer

Today, we’re evolving @StitchbyGoogle from @GoogleLabs into an AI design canvas transforms natural language prompts into production-ready front-end code. Some highlights from what’s new: 1. A complete redesign of the Stitch UI, which can now ingest multimodal references (text prompts, images, or code) as creative seeds for your design ideas 2. A brand new, context-aware design agent that can share feedback on builds, generate PRDs, and ask questions to better understand your vision. You can even talk to the agent if you prefer a verbal sounding board 3. A new agent-friendly markdown file, DESIGN.md, which you can use to export or import your design rules to or from other design and coding tools Whether you’ve been designing for decades or you’re whiteboarding your first software idea, Stitch can help you turn concepts into prototypes in minutes rather than days ➡️ stitch.withgoogle.com

@AnthropicAI nthropic asked 81,000 people what they fear AI might do. then published the answers. most companies bury that research. that's either confidence or the best PR move in AI

We invited Claude users to share how they use AI, what they dream it could make possible, and what they fear it might do. Nearly 81,000 people responded in one week—the largest qualitative study of its kind. Read more: anthropic.com/features/81k-i…

@googledevs AI agents can now write and run code in a secure sandbox. the agents are fine. the sandbox is probably fine. everything is fine

Bringing Google Colab’s secure environment to your AI agents. ☁️🛠️ With the new open-source Colab MCP Server, your agents can now natively write and execute code inside a Colab Notebook. Secure, automated, and ready to go.

@kimmonismus "basically solved" from the guy supplying the compute for both. he's not guessing, he's watching the training runs

„Jensen, where do you see NVIDIA in 10 years?“ - robot is basically solved - full self driving is basically solved - the future is almost impossible to grasp

@WSJ nvidia: we make the chips also nvidia: we make the software also nvidia: we make the agents jensen huang: yes

Nvidia CEO Jensen Huang unveiled NemoClaw, a new software tool kit to help businesses safely build and run autonomous agents with OpenClaw on.wsj.com/4sPqaGy

the shift from "lab to real world care" is where most healthcare AI has historically stalled. if google is actually clearing that hurdle, the 25% detection improvement is just the beginning

@GoogleWorkspace the folder system was already dying. google just put AI search on top of drive and made organizing files someone else's problem

No more digging through files to find what you’re looking for. AI Overviews in Drive give you the answers you need, with citations, at the top of your search results. This feature is rolling out for Gemini Alpha customers and Google AI Pro and Ultra subscribers in the US. goo.gle/3OW1BJu

@kimmonismus kimi taught AI to pay attention to what actually matters instead of treating all past information equally. sounds obvious. nobody had cracked it at scale before

Holy: Kimi did an amazing work because it changes one of the most basic parts of how deep AI models pass information from layer to layer. Instead of blindly mixing in everything from earlier layers equally, the model can now choose which past information is actually useful for each token and task. That helps deep models keep important signals from getting washed out, making training more stable and efficient. The big deal is that Kimi shows this idea works at scale too: better results, about 25% more compute efficiency, and almost no extra inference slowdown.

claude opus 4.6 and sonnet 4.6 just went ga on 1m context window, basically entire code bases or books in one shot now

@NewsFromGoogle @GoogleResearch this is the kind of ai use case i actually like: less “write my email”, more “don’t let this neighborhood drow”

Groundsource is a new AI-powered methodology from @GoogleResearch that transforms millions of public records and historical disaster data into an actionable dataset designed to better predict flash floods in urban areas up to 24 hours in advance.

@Forbes ai giving medical advice isn’t the wild part. the wild part is letting one app read your entire health history and hoping the privacy slide wasn’t just vibes

Would you use AI for medical advice? Microsoft, like other AI giants, is betting on it. The tech titan announced a new feature called Copilot Health on Thursday. Learn about the feature and more in today's Forbes Daily newsletter: forbes.com/sites/danielle…

@GoogleCloudTech onboarding is one of those things every company does badly. if AI agents can actually fix it, that's a bigger deal than most enterprise software launches

Build intelligent employee onboarding with Gemini Enterprise. Here's how: Use ADK, Agent Engine, and Application Integration to build custom agents that connect to your enterprise systems. Next, publish them to the Gemini Enterprise agent gallery → goo.gle/3NxLrFE

@WSJ an AI that knows your entire medical history and can talk to you about it on demand is either the best thing to happen to healthcare or a privacy nightmare. probably both

A new feature within the Microsoft Copilot app will offer an AI concierge doctor—one that can access your medical records and health data, with your consent on.wsj.com/4s0Mw7W