DAILY AI BRIEF 🗞 — Sept 15

AI SLOWDOWN 🔥:

> Dario Amodei: pace the frontier, not halt it — embedded evaluators, lab coordination, then a global deal.

> Elon, All-In: peer-review rival models before release. “Dario is right.”

> Sam Altman agrees, will match evaluators, and says pacing means slower, not stopped.

> Demis Hassabis: Amodei’s essay “points towards the right path,” details TBD, plus DeepMind’s standards-body plan.

> Satya Nadella welcomes deliberate pacing and embedded evaluators; superintelligence without human control is “not worth pursuing.”

> Alexandr Wang: alignment is fundamental and may gate scaling near the frontier, with no join on a coordinated slowdown.

> Trump rejects a slowdown: takeover talk is a “hoax,” control is a “strong and smart president,” and a “sick conspiracy” against AI only helps China.

> China MFA’s Guo Jiakun: slowdown talk is “fear-mongering”; confrontation “serves no one’s interest.”

MODEL FORECAST 🔥:
-‌ Today? Gemini 3.8 Live and Live Extended Thinking slugs appeared on the Cloud quota page; not public.
-‌ This week? Grok 4.7 should land near Opus 5.0, not 5.1; multimodal still needs work.
-‌ This week? Testers say some Claude Code Opus 5 traffic is routing to Opus 5.2 — faster and less lazy, still unofficial.
-‌ Soon: Grok 4.8 is a 2.5T model on a new C++ stack; training wraps this week, then RL.
-‌ Not soon: Grok 4.9 is “probably Astra/Fable class”; Grok 5 is “maybe better than anything.”

PERPLEXITY 🔥:
-‌ Portable Computer is live on Windows PCs with NVIDIA RTX GPUs and 24GB+ VRAM, including local MCP and scheduled tasks.

OPENAI 🔥:
-‌ Bought Glass Imaging for $300M, per WSJ — an ex-Apple smartphone-camera startup, after the earlier Opal webcam stake.
-‌ Codex — got an official Arch Linux installer with pacman updates.

GOOGLE 🔥:
-‌ Interactive Reports incoming on Gemini Notebook: embed mind maps, slide decks, flashcards, and quizzes.
-‌ Gemini Live can run Deep Research in the background and ping you when the report is ready.
-‌ API slug antigravity-preview-09-2026 spotted: hosted Linux sandbox agent, expected to be powered by Gemini 3.8 Flash.

P.S. All predictions in the Model Forecast section are unconfirmed and may change anytime.

P.S.S. We've used Grok to compose this brief, cherry-picking the news and doing some post-editing.

Claude Code Hub? Users will be able to see all their sessions across various places in Claude Desktop.

> Model, effort, and permission mode selectors are configurable for each.
> Both local and cloud sessions are listed there so it is easy to switch between them.

GOOGLE 🔥: Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are rolling out on APIs, Google AI Studio, Gemini Live, and other apps.

Gemini 3.8 Live Extended Thinking (High) is taking the 1st spot on most benchmarks, overtaking GPT Live 1.

Grok Imagine users can now precisely edit text on generated images. Users can manually configure color, size, font, and text alignment.

> The feature arrived in Beta, and the SpaceXAI team is collecting feedback.

> This feature would be particularly usefull for marketing and creative tasks on text-heavy images.

I have a good test in mind for this feature for tomorrow.

Testing time 👀

Google rolled out Gemini 3.8 Live and Extended Thinking

Google launched Gemini 3.8 Live and Extended Thinking, voice models for near real-time dialogue, reasoning and task execution. Live serves high-volume, multilingual work; Extended Thinking handles complex workflows and leads key speech benchmarks.

🗞 #google @testingcatalog

Meta launches Meta One globally with 50+ paid features

Meta One rolls out globally with paid tiers across Instagram, Facebook, WhatsApp and Meta AI. It adds 50+ features for consumers, creators and businesses, while core apps and standard Meta AI access remain free.

🗞 #meta @testingcatalog

GOOGLE 🔥: Gemini Notebook on mobile is getting Voice Mode and Voice Recording! Interactive Reports on Gemini Notebook are rolling out as well.

Gemini Notebook users will soon be able to capture their thoughts or record lectures.

> Voice Recorder will start rolling out next week for all users in English and other supported languages.

> Realtime Voice Mode on Gemini Notebook is rolling out to Ultra subscribers this week and is expected to land for Pro users "soon".

> Interactive Reports are rolling out to all users in the coming weeks. Besides that, new Qwiz formats have been added too.

DAILY AI BRIEF 🗞 — Sept 16

GOOGLE 🔥:
- Gemini 3.8 Live and 3.8 Live Extended Thinking are out. 97-language auto-detect, near real-time vision, background tool calling.
- Live is in Search Live plus Gemini API public preview. Extended Thinking is in Gemini Live, with Pro/Ultra getting it in Docs, Gmail, and Keep.
- Gemini Notebook Voice Mode hits Ultra this week, Pro soon. Mobile voice recorder starts next week for all users, English first.
- Interactive Reports for Gemini Notebook roll out to everyone in the coming weeks, plus new quiz formats and 60-second video overviews.

OPENAI 🔥:
- Sam declared a big ship week, then a much larger wave for DevDay. GPT-6 Sol and Luna are the expected drops.
- GPT-5.5 leaves ChatGPT, Work, and Codex on Oct 14. Switch to GPT-5.6 Sol or GPT-6 Astra; the API keeps 5.5.

XAI 🔥:
- Grok Imagine can now edit text on any image in beta — color, size, font, alignment.
- Grok Build 1.0.33: structured MCP JSON, in-UI memory deletes, and long-session checkpoints that survive cleanup.

ANTHROPIC 🔥:
- Salesforce in Claude is in beta. 37 pre-built sales skills for pipeline, forecasts, and call prep.

ALSO TRENDING 🔥:
- TypeSafe AI — Jev is out: a non-LLM decision model at ~150ms, $42 per billion input tokens, output free.
- Salesforce — Koa is its first CRM reasoning model for Agentforce, built on NVIDIA Nemotron.

* Used Grok to compose this brief, cherry-picking the news and doing some post-editing.

Gemini Notebook adds Voice Chats and Interactive Reports

Google is updating Gemini Notebook with real-time voice chat, mobile recording, interactive study reports, new quiz types, and short video overviews, while expanding free higher-limit plans for eligible college students.

🗞 #google @testingcatalog

Apodex launches Apodex v1.1 Deep Discover model

Apodex 1.1 shifts from answering questions to completing analytical work across files, with code execution, traceable outputs, mid-task recovery, and multi-agent coordination, posting large benchmark gains over 1.0.

🗞 #sponsored @testingcatalog

Leaks: Google testing math-focused DeepThink V3 model

Leaked screenshots point to Google testing a math-focused DeepThink V3 variant, “Mathematica,” tuned for heavy calculation and symbolic reasoning, with a 1M-token context. Public release remains unconfirmed pending internal tests.

🗞 #google @testingcatalog

Mistral models now power Firefox Smart Window

Mistral and Mozilla are bringing private, multilingual AI to Firefox Smart Window, with regional language tuning, zero data retention, and support for search, tab discovery, and recalled content, starting in France and North America.

🗞 #mistral @testingcatalog

GPT-6 Astra Fast Max will be a beast 👀

OpenAI is preparing a Codex Replay feature that lets users test task execution from any imported conversation thread.

> "Start an independent Codex Replay controller on an available loopback port. In both Codex Desktop and Codex CLI, prefer an available Codex in-app browser and otherwise use your system browser."

> "Select one or more historical Claude threads, choose shared Codex models, and start their isolated implementations. Historical state and configuration are detected separately for each thread, with details available when needed."

> "View finished comparisons individually or in aggregate while remaining threads continue. Available GPT-5.6 Sol, Terra, and Luna models are selected by default."

> "Multiple replay sessions can run in parallel. Selection, execution, verification, evaluation, and results stay in the browser controller instead of a guided chat workflow."