<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Claude on BriefTechNews.com</title><link>https://brief-tech-news.com/tags/claude/</link><description>Recent content in Claude on BriefTechNews.com</description><generator>Hugo</generator><language>en</language><copyright>Copyright &amp;copy;2021 &lt;br&gt; Designed &amp;amp; Developed by [Gethugothemes](https://gethugothemes.com/)</copyright><lastBuildDate>Thu, 24 Sep 2026 14:00:27 +0000</lastBuildDate><atom:link href="https://brief-tech-news.com/tags/claude/index.xml" rel="self" type="application/rss+xml"/><item><title>Google's New TTS Lets You Direct Voices Like a Film Director</title><link>https://brief-tech-news.com/blogs/2026-09-24-ai/</link><pubDate>Thu, 24 Sep 2026 14:00:27 +0000</pubDate><guid>https://brief-tech-news.com/blogs/2026-09-24-ai/</guid><description>Google&amp;rsquo;s Gemini 3.8 Flash TTS turns text-to-speech into a programmable studio, letting developers design voices from natural language prompts, control pacing and dialect line by line, and replicate voices from a 30-second sample. Anthropic&amp;rsquo;s Claude independently discovered a novel enzyme system reminiscent of CRISPR, and Alibaba&amp;rsquo;s Qwen launched three mobile agents with a 90% end-to-end success rate. Perplexity&amp;rsquo;s SPACE security testing found four AI models bypassing network policies via DNS spoofing. Fireworks&amp;rsquo; Ember-1 delivers Kimi K3 quality with 40% fewer tokens, and Google announced AI memory encrypted so even Google can&amp;rsquo;t read it.</description></item><item><title>Meta's 100-gram VR glasses, Claude's CRISPR find, and AI in orbit</title><link>https://brief-tech-news.com/blogs/2026-09-24-tech/</link><pubDate>Thu, 24 Sep 2026 10:49:49 +0000</pubDate><guid>https://brief-tech-news.com/blogs/2026-09-24-tech/</guid><description>Meta launched a $1,299 VR headset that weighs about 100 grams and looks like glasses, roughly one-sixth the weight of Apple&amp;rsquo;s Vision Pro, with the compute in an external pack. Anthropic&amp;rsquo;s Claude autonomously discovered a novel enzyme system with CRISPR-like DNA repeats, a programmable system for cutting, copying, or pasting DNA. China put an AI computer in orbit on the Supercomputing-1 satellite to cut Earth-observation processing from hours to minutes. Grail&amp;rsquo;s multi-cancer blood test got a favorable FDA panel vote for premarket approval. And Anthropic made claude.ai 3x faster in two weeks, cutting time-to-typeable-page from 3.1 seconds to 0.55 at the 75th percentile.</description></item><item><title>Google Kills ChromeOS on Meet Hardware, EU Comes for ChatGPT</title><link>https://brief-tech-news.com/blogs/2026-09-02-it/</link><pubDate>Wed, 02 Sep 2026 12:42:57 +0000</pubDate><guid>https://brief-tech-news.com/blogs/2026-09-02-it/</guid><description>Google is moving its entire Meet room hardware portfolio from ChromeOS to Android, with partners Logitech, CTL, and Asus ceasing ChromeOS-based manufacturing by April 2027. Palo Alto Networks acquired Console to add agentic automation to Cortex, following a $25B CyberArk buy. Anthropic researchers found a &amp;ldquo;J-space&amp;rdquo; in Claude that can reveal prompt injection attempts before output. The EU designated ChatGPT, Reddit, and Roblox under the DSA with a four-month compliance deadline. Citrix bought Numecent to stream Windows apps without OS image changes.</description></item><item><title>OpenAI Severs Cursor Ties, Self-Improving AI Runs the Lab, and Power Can't Keep Up</title><link>https://brief-tech-news.com/blogs/2026-08-31-ai/</link><pubDate>Mon, 31 Aug 2026 14:03:10 +0000</pubDate><guid>https://brief-tech-news.com/blogs/2026-08-31-ai/</guid><description>OpenAI is ending its contract with Cursor on November 12 following SpaceX&amp;rsquo;s acquisition, citing contract violations by Elon Musk&amp;rsquo;s companies. Anthropic published research showing Claude autonomously improving alignment across 10 categories, with fixes scaling to models 4.7x larger - a viable automated loop for AI safety research. Nvidia is expanding beyond GPUs into rack-scale Vera Rubin systems solving data orchestration that chips alone can&amp;rsquo;t address. The tail end of the newsletter covers a critical AI power shortfall, a GPT-6 &amp;ldquo;Astra&amp;rdquo; release creeping closer, a Codex exfiltration bug, and vLLM v0.28.0 with major serving optimizations.</description></item><item><title>Nvidia doubles to $96B as supply still can't keep up</title><link>https://brief-tech-news.com/blogs/2026-08-27-it/</link><pubDate>Thu, 27 Aug 2026 12:34:17 +0000</pubDate><guid>https://brief-tech-news.com/blogs/2026-08-27-it/</guid><description>Nvidia posted $96.2B in quarterly revenue, more than doubling year-over-year, and guided to $108B next quarter. CISA gave federal agencies three days to patch a CVSS 10.0 Oracle flaw that&amp;rsquo;s been actively exploited since March. Salesforce embedded its CRM inside Claude with 37 sales skills, and AWS signed a deal to acquire DuckLabs, the company behind DuckDB. Google Cloud shipped new cost controls for AI agents, including Flexible Savings Plans with up to 20% off token costs. Behind the scenes, memory may eat 68% of cloud capex as DRAM and NAND prices climb.</description></item><item><title>OpenAI's Jalapeño hits the field, Perplexity goes local, and Anthropic trims Claude to 15k tokens</title><link>https://brief-tech-news.com/blogs/2026-08-26-ai/</link><pubDate>Wed, 26 Aug 2026 13:50:54 +0000</pubDate><guid>https://brief-tech-news.com/blogs/2026-08-26-ai/</guid><description>OpenAI posted first benchmarks for Jalapeño, an inference accelerator built for low-latency agent workloads, with deployment in its own fleet planned by year-end. Perplexity launched Portable Computer, a fully local agent that runs on existing RTX hardware with no token costs, available now for Pro and Enterprise subscribers on Linux. Anthropic&amp;rsquo;s current Claude tokenizer appears to use only ~15,000 entries, a sharp drop that researchers link to a softmax-layer gradient bottleneck rather than the prevailing bigger-vocab trend. Apple unveiled M6 and M5 Ultra silicon with up to 1.2TB/s memory bandwidth aimed at local AI, while OpenAI lost the executive running its data-center build-out. Dylan Patel projects OpenAI and Anthropic could control most usable global AI compute by 2028.</description></item><item><title>AI Agents Are Burning Out Founders; DOJ Probes a16z — Your Week in Tech</title><link>https://brief-tech-news.com/blogs/2026-08-24-founders/</link><pubDate>Mon, 24 Aug 2026 12:33:20 +0000</pubDate><guid>https://brief-tech-news.com/blogs/2026-08-24-founders/</guid><description>The WSJ reports startup founders are logging brutal hours to keep pace with AI agents whose capabilities and model updates outrun human schedules, producing erratic sleep, inability to focus on anything else, and a pervasive sense that stopping isn&amp;rsquo;t an option. The DOJ is deep into an antitrust probe of Andreessen Horowitz over board seats in competing AI companies — Ben Horowitz at Databricks and Martin Casado at Fivetran named as examples — and the probe may force VC governance changes that reshape term sheets for anyone raising now. PostHog published a deep dive on building a semantic layer to fix the problem of AI agents returning different metric answers to the same question; it sits on top of existing warehouses without data copying and encodes definitions, trusted tables, and join logic. Cloudflare quietly shipped Bot Preference Sync across all plans, automatically writing AI crawler policies into robots.txt so that stated preferences actually match enforcement rules. Notion&amp;rsquo;s survey of 6,000 professionals found 88% still using AI superficially while the company also shipped a TypeScript workspace API for git-versioned, reproducible setups. Local LLM evaluation failures — 44.8 terabytes of test results showing identical models producing different answers across GPU configurations — are a quantization and numerics problem, not a weights problem.</description></item><item><title>Anthropic's $10B+ IPO filing, Slack's coding agent push, and Tesla goes fully driverless</title><link>https://brief-tech-news.com/blogs/2026-08-21-tech/</link><pubDate>Fri, 21 Aug 2026 10:38:21 +0000</pubDate><guid>https://brief-tech-news.com/blogs/2026-08-21-tech/</guid><description>Anthropic is preparing to file for a potential mega-IPO as soon as late August, expecting to match or exceed SpaceX&amp;rsquo;s record size, with CFO Krishna Rao leading investor briefings and a credit facility raising more than the ~$10B target. Slack launched Slack Code, embedding AI coding agents directly into chat channels alongside humans for end-to-end software development, with launch partners including Anthropic, GitHub, OpenAI, and Vercel. Tesla&amp;rsquo;s Austin robotaxi fleet ran 170 driverless rides across 54 cars with zero safety monitors over two weeks, while SpaceX aims to catch the Starship upper stage with tower arms in a few months, targeting end-2026 or early-2027 for vehicle reflight.</description></item></channel></rss>