AI News
Snorkel AI Triples Valuation to $3.5B as Training Data Demand Explodes
The data startup raised a $350 million Series E led by Insight Partners and S32,...
Qualcomm's Snapdragon 8 Elite Gen 6 Chips Push More AI Onto the Phone
Qualcomm's new flagship processors can run small models on a sensing hub, handle...
Anthropic Ships Opus 5.5 With Lower Prices and Fable-Level Performance
Anthropic's new Opus 5.5 beats its larger Fable model on many benchmarks, costs ...
OpenAI's GPT-6 Sol and Luna Arrive at Half the Price of Their Predecessors
OpenAI expands the GPT-6 family with refreshed Sol and Luna models, cutting API ...
Copilot on Windows 11 Is Getting Its Own Built-In Browser
Microsoft will roll out in-app browsing to the Copilot app broadly in late Novem...
OpenAI Forms Math Advisory Group as Its Models Claim 100-Plus Open Problems
The new advisory group, hosted at the Institute for Advanced Study, gives mathem...
Google's $899 Googlebook Opens for Preorder With Gemini Wired Into the Cursor
Google's AI-native laptop runs Android with a desktop Chrome browser, features a...
Amazon Blocks Meta's Muse Agent From Shopping on Its Marketplace
Amazon has cut off Meta's Muse AI assistant from making purchases on Amazon.com,...
Meta's Muse Hits 1.8 Million iOS Downloads in 12 Days, Outpacing ChatGPT's Debut
New third-party estimates show Meta's Muse AI app beating ChatGPT's early mobile...
Context Windows Keep Growing - Here Is What Actually Breaks
Bigger context does not mean better recall. Models still lose information in the...
AI Summaries in Search Results: What They Change for Site Owners
Answer boxes reduce clicks on informational queries but reward pages that state ...
Why Open-Weight Models Keep Winning Specific Jobs
Open-weight releases rarely top general leaderboards, but for classification, ex...
How AI Products Charge: Seats, Tokens and Everything Between
Token metering was supposed to be the answer. In practice most vendors blend sub...
Why Enterprise AI Pilots Stall, and What Gets Them Moving
Most stalled pilots fail on data access, unclear ownership and evaluation gaps r...
Large Language Models: A Plain Explanation
A large language model predicts the next token from everything it has seen so fa...
Retrieval-Augmented Generation: How It Works
RAG fetches relevant passages from your own documents and places them in the pro...
Fine-Tuning versus Prompting versus RAG
Most teams reach for fine-tuning too early. Prompting and retrieval solve the ma...
Evaluating AI Output: Metrics That Mean Something
Automated scores are useful for tracking regressions, not for deciding whether s...
AI Note-Taking Apps Compared: Which One Fits Your Meetings
We looked at how AI meeting tools handle speaker separation, action items, searc...
AI Coding Assistants: A Month of Real Project Work
Inline completion is commoditised. The real divide is how well each assistant ha...