All sections
News
507 stories · page 1 of 57
SecurityCloudflare: 363% traffic surge to US election sites, but zero major DDoS attacks on Election DayAccording to Cloudflare, traffic to government election websites rose to about 602 million requests on 3 November, while vulnerability exploit attempts against election sites climbed from 200,000 a day in January to 2 million a day in October. On Election Day, Cloudflare detected no major DDoS attack.FrontierGoogle ships Tool Retrieval in the Gemini API: the agent keeps its tool library on standby and loads a schema only when askedGoogle announced Tool Retrieval in preview in the Gemini Interactions API: parameter schemas load on demand instead of the full tool library riding along in every prompt. It works with Gemini 3.6 and later, including all Gemini 4 models.FrontierGenie 4 is now in the Gemini API and AI Studio: 720p worlds at 24fps with synchronized audio and a SynthID watermarkGoogle announced on 19 November that Genie 4 is available to developers through the Gemini API and Google AI Studio. From a text prompt or a reference image, it generates continuous 720p worlds at 24fps with synchronized audio, and every output carries an imperceptible SynthID watermark.ScienceVoyager 1 reached one light-day from Earth, and no science data has come from the probe since 14 NovemberNASA said Voyager 1 reached one light-day from Earth on 18 November. Science and engineering data from the probe have not been received since 14 November, and the cause is still being analyzed.AudioReson8 open-sources Rumble-2-Flash: a speech model with 0.5 billion parameters for 40+ languages, and on phone calls it is on par with the 30B Nemotron, according to Reson8On 18 November Reson8 released Rumble-2-Flash under the Apache 2.0 license. According to Reson8, on phone calls its error rate is 7.86% against 7.78% for Nemotron 3.5 Transcribe. The post does not give the full list of languages.FrontierMicrosoft's Aion Instruct: a small 2.3 billion parameter model for Windows, under MIT, with eight languages and no BulgarianOn 18 November Microsoft published the small model Aion Instruct on Hugging Face: 2.3 billion parameters, MIT and a 32K token context, built to run on the device itself. Microsoft plans for an optimized version to become the language model of Windows by the end of January 2027. Bulgarian is not among the eight languages.FrontierGoogle releases Gemini 4 Neon: $3 per million input tokens, and by Google's figures the cost per task is 32% below Carbon'sGoogle released Gemini 4 Neon on 18 November: $3 per million input tokens, $15 per million output tokens and a 1 million-token context window. By Google's figures, at its default thinking level the model costs 32% less per task than Carbon.ControlOpenAI described its defenses from training to deployment: models trained without live internet, and a Control Room that stops output within 35 secondsOn 18 November OpenAI laid out in one place how it protects its models at every stage. In its words, frontier models are trained with no live internet, and Control Room stops their output within 35 seconds of a halt instruction. There is no outside check.
507 stories