AI News
Daily Tech Reader - AI Edition
Podcast 🎧 • Video 📽 • Short 📽
Claude
Claude Code Adds Self-Hosted Runner Environments — Teams Can Now Run Sessions on Their Own Machines
The latest Claude Code update introduces self-hosted environments via a new CLI command, letting Team and Enterprise users run Claude Code sessions on their own machines or containers rather than Anthropic's infrastructure.
Sources: Anthropic · Releasebot
Claude Code Removes Subagent Spawn Cap — Expands Dynamic Workflows and Nested Subagents
Alongside self-hosted runners, the update removes the prior cap on subagent spawning, adds archive plugin installs via HTTPS zip, and fixes a range of MCP, sandbox, and remote control bugs across CLI and accessibility flows.
Sources: Anthropic · Releasebot
Claude Opus 4.7 Loses Fast Mode — Requests Must Now Use Standard Speed or Migrate to Opus 5
Anthropic removed fast mode for Claude Opus 4.7; requests using the speed fast parameter now return an error and no longer fall back to standard speed, unlike Opus 4.6. Developers are directed to migrate to Claude Opus 5 or Opus 4.8 to continue using fast mode.
Sources: Anthropic
ChatGPT
ChatGPT Surpasses 1 Billion Weekly Users — OpenAI Drops Text Chat Limits for All Tiers
OpenAI announced it is removing limits on text-based chats across all ChatGPT tiers after the platform crossed 1 billion weekly users, with unlimited text access and a new Think button rolling out to Free and Go users starting next week.
Sources: Dataconomy · OpenAI · Releasebot
GPT-5.6 Luna Becomes Default for Free Tier — Sol Upgraded for Plus and Pro Users
GPT-5.6 Luna replaces GPT-5.5 as the default model for Free and Go users, while Plus and Pro subscribers get an upgraded GPT-5.6 Sol with a new reasoning-depth slider and 68% fewer factual errors than GPT-5.5-Instant.
Sources: OpenAI · The AI Insider
OpenAI's First Hardware Device Is a Donut-Shaped Smart Speaker — Priced at $300 to $400
Bloomberg reports OpenAI's debut consumer device will be a high-quality metal smart speaker with moving parts, designed with Jony Ive's LoveFrom studio and targeting a 2027 release — significantly pricier than Amazon's competing lineup.
Sources: Bloomberg · TechCrunch · American Bazaar
Atlas Browser Agent Being Retired August 9 — OpenAI Folds Capability Into ChatGPT and Codex
OpenAI announced Atlas will stop working August 9 as it integrates browser-based agentic capabilities directly into ChatGPT and Codex; users must manually export bookmarks, passwords, and tabs before the cutoff.
Sources: OpenAI Help Center
Gemini
Demis Hassabis Posts Farewell Message to DeepMind — Calls Team "AI Engine Room of Google"
In an internal message now published on Google's blog, Hassabis praised DeepMind's role powering Gemini across AI Mode, AI Overviews, and the Gemini App — which now has over 950 million monthly users — as he hands day-to-day leadership to incoming SVP Koray Kavukcuoglu.
Sources: Google · The Globe and Mail
SemiAnalysis: "Gemini Is Cooked" — GCP Seen as the Real Winner of DeepMind Shakeup
SemiAnalysis argues the biggest beneficiary of Hassabis's exit and Jeff Dean's departure is Google Cloud, predicting GCP revenue growth will accelerate meaningfully now that the internal compute fight between Gemini and GCP has resolved in Kurian's favor.
Sources: SemiAnalysis
Copilot
Grok 4.5 Now Available in GitHub Copilot — Selectable Across VS Code, CLI, and Cloud Agents
xAI's Grok 4.5 is live inside GitHub Copilot for millions of developers via the model picker in VS Code, the Copilot CLI, and cloud agents; enterprise organizations must first enable the model from Copilot settings.
Sources: xAI · Releasebot
Sign in With ChatGPT Launches in Beta — Airtable, GitLab, HubSpot, Notion, Supabase, Vercel First
OpenAI is rolling out Sign in with ChatGPT across select partner plugins, letting users authenticate on external platforms using their ChatGPT account — a move that extends OpenAI's identity footprint into the broader developer ecosystem.
Sources: OpenAI · Releasebot
Grok
Grok 4.6 Launches Today — 1.5 Trillion Parameters, Significantly Improved SFT and RL
xAI released Grok 4.6 on its targeted August 7 date, delivering post-training upgrades — improved supervised fine-tuning and reinforcement learning — on the same 1.5T V9 foundation as Grok 4.5, with Grok 4.7 at 2.1T scheduled a few weeks behind.
Sources: xAI · Roic News · American Bazaar
Grok Build Gets Dashboard and CLI Fixes — Better Prompts, Formatting, and Large-Session Performance
The latest Grok Build update adds collapsible Skills sections in the extensions modal, auto theme detection over SSH and inside tmux, reflowing markdown tables on narrow panes, and performance improvements for large sessions.
Sources: xAI · Releasebot
China AI
DeepSeek Seeks $8 Billion in Renewed Funding Push — Talent Retention Cited as Primary Driver
DeepSeek is restarting its funding round targeting $8 billion, according to Bloomberg's China Show, primarily to retain AI researchers from poaching by rival labs offering higher valuations — not because of a cash shortage at the hedge-fund-backed startup.
Sources: Bloomberg · Whatfinger
Alibaba to Charge Big Users of Next Open-Source AI Model — Qwen3.8-Max Targets 1M Token Context
Reuters reports Alibaba plans to charge high-volume users of its upcoming Qwen3.8-Max model, a 2.4-trillion-parameter open-weighted release with a 1-million-token context window scheduled for release next week.
Sources: Reuters · American Bazaar
Goldman Sachs Raises China AI Revenue Forecast to $13 Billion — Up From $10 Billion Estimate
Goldman Sachs boosted its year-end annualized recurring revenue projection for mainland Chinese AI models, citing stronger-than-expected enterprise adoption and the continued momentum of open-source offerings from DeepSeek and Alibaba.
Sources: South China Morning Post
Big Story
OpenAI Hits a Billion Users and Drops Text Limits — The Era of Uncapped AI Begins
Crossing 1 billion weekly users in under four years, OpenAI's decision to remove text chat caps marks a structural shift: the scarcity model for frontier AI access is effectively over for everyday use, with usage limits now pushed to files, images, and voice.
Sources: OpenAI · Dataconomy · The AI Insider
Grok 4.6 Day Arrives as xAI Bets on Post-Training Over Scale — 4.7 at 2.1T Follows in Weeks
By launching Grok 4.6 on the exact date Musk publicly targeted, xAI signals it can execute on a compressed release cadence — and the two-step strategy of SFT-optimized 4.6 followed quickly by larger-scale 4.7 is a direct challenge to OpenAI and Anthropic's model timelines.
Sources: xAI · Roic News · Tbreak
Business
OpenAI to Pay $3.2 Million to Settle DOJ Visa Hiring Discrimination Case
OpenAI agreed to a $3.2 million settlement with the Department of Justice over allegations that the company discriminated against non-U.S. citizens in its hiring practices, according to reporting from August 5.
Sources: American Bazaar
Hark AI Raises $700 Million Series A at $6 Billion Valuation — Targets "Personal Intelligence" Hardware
Hark, the AI hardware lab founded by Brett Adcock, closed an oversubscribed $700M Series A in May to develop proprietary AI models paired with custom hardware designed as a universal interface between humans and machines.
Sources: American Bazaar · TechCrunch
ChatGPT Outage Reported Friday Morning — DownDetector Logs Surge Around 2:10 A.M. Eastern
DownDetector India registered a spike in ChatGPT outage reports around 12:40 p.m. IST (2:10 a.m. Eastern) on Friday, with users across social media reporting access and response issues.
Sources: DownDetector · DesignTAXI Community