DAILY AI BRIEF 🗞 — Oct 3

OPENAI 🔥:
* OpenAI's Python SDK 3.24.0 quietly adds a Voices API for creating custom voices from a text prompt or an audio sample.
* ChatGPT launched virtual Try On globally, letting you see clothes from shopping results on your own photo.
* OpenAI disclosed a second Australian government incident, where an AI model accessed non-public NSW fire statistics.

GOOGLE 🔥:
* Antigravity quietly added Claude Sonnet 5.5 and Opus 5.5 (thinking) for AI Pro and Ultra plans, with Claude 4.6 models and GPT-OSS-120b being removed on Nov 2.
* Gemini Desktop is testing a hidden "Full Access" mode that lets computer use reach files and apps anywhere on your Mac 👀
* Guided Vision is rolling out in Gemini Live on Android, describing in real time whatever your camera sees.
* Google's Project Suncatcher prototype satellite is in orbit, testing how TPUs handle radiation and heat in space.

ANTHROPIC 🔥:
* Claude Frontier Academy launched with $100M to train 10,000 Frontier Deployed Engineers by 2027.
* Profile picture support looks to be coming to Claude apps.
* Claude Code 2.1.288 adds --max-findings to /code-review.

MICROSOFT 🔥:
* GitHub Copilot CLI and the Copilot desktop app got computer use in public preview.
* Copilot code review can now be requested through the REST and GraphQL APIs.

SPACEXAI 🔥:
* An official experimental TypeScript SDK for the SpaceXAI API quietly shipped on npm as xai-official/sdk.

DEEPSEEK 🔥:
* DeepSeek Harness v0.2 now ships as desktop apps for macOS and Windows, with Linux via npm.

META 🔥:
* Muse Home Link, a USB-C gadget that connects Muse to smart-home devices, is free for US subscribers, with 5,000 units shipping in October.

APPLE 🔥:
* Apple is tightening macOS Full Disk Access, citing the extra risk from AI agents.

NVIDIA 🔥:
* DGX Spark 64GB arrives Oct 23 from $4,999, and two units can cluster to 128GB.

CLOUDFLARE 🔥:
* Cloudflare released Clef and Clef-flash, open decision models that return probabilities instead of text.

Meta is working on “Agent Engine” for its Meta API Platform.

Agent Engine is called “Forge” internally, however, it is not yet clear what solution will it be offering.

A new harness model incoming? 👀

Aleph Alpha released Kolibri, a 78B-parameter, 3.46B active, MoE open-weight model with 1M context window, distributed under the Apache 2.0 license.

Aleph Alpha is a German AI lab, and it is probably the first general-purpose model from the EU, apart from Mistral, to achieve this level of performance!

> "The model supports an explicit reasoning mode and tool calling. It is optimized for long-context and inference efficiency."

> "The Aleph Alpha team developed a bilingual German/English tokenizer and focused on including organic German data throughout the training process of the model, so that 21.3% of the pre-training tokens are German."

> "Kolibri was built with the EU AI Act, the General-Purpose AI Code of Practice and the GDPR in mind from the ground up, with copyright law being a focus of our work to trustworthy technology."

More EU labs joining the party 👀

Connecting the "dots" between ChatGPT and World Wallet

Leaked app screens suggest ChatGPT will add a wallet for agent-led purchases. With OpenAI testing payment flows and World App rebranding to World Money, the timing hints at a possible link, though no public evidence confirms it.

🗞 #openai @testingcatalog

Aleph Alpha releases open-weight Kolibri with 1M context

Aleph Alpha released Kolibri, a bilingual open-weight 78.1B MoE model for sovereign government and regulated use. Apache 2.0 licensed, on-prem deployable, it offers 1M-token context, German focus, tool calling, and grounding.

🗞 #release @testingcatalog

Mistral is working on its own Voice Mode 👀

It seems to be based on TTS, but it may be meant to simply work with a different model that hasn't been released yet.

DAILY AI BRIEF 🗞 — Mon 5

GOOGLE 🔥:
* Gemini app is reshuffling model access: free users drop to Flash-Lite only from Oct 9, AI Plus loses Pro, and AI Pro gains Deep Think.

ALEPH ALPHA 🔥:
* Kolibri-1 is out as open weights under Apache 2.0: a 78B MoE with 3.46B active parameters, built for German/English sovereign deployments.

ANTHROPIC 🔥:
* Claude now asks voice users to opt in to sharing voice data for model training, via a separate privacy toggle that is off by default.

OPENAI 🔥:
* GPT-6.1 Sol Ultrafast is coming soon, per Codex lead Tibo Sottiaux.
* Safety lead David Robinson resigned, writing in The Atlantic that OpenAI's "culture is broken."

MISTRAL 🔥:
* Mistral is working on its own Voice Mode, which looks TTS-based for now.

META 🔥:
* Meta is building an "Agent Engine" for its Meta API Platform, codenamed Forge internally.

Figure plans to release Hark AI assistant this week

Figure is planning to release its own proactive AI assistant this week!

Here is what we know so far 👀

> Hark will likely run on Opus 5.5 for general tasks, with Gemini 3.1 Flash Image, Qwen3.6-27B, and GPT-6 Astra referenced in the code for other specific tasks.

> Hark will arrive with Hark Pro, Pro², Pro³, and Pro⁴ paid plans, with 5× and 20× usage options and another offering twice the usage of Pro³.

"Hark has the power to plan 800 weddings, buy 1000 grand pianos, move across the country 1,500 times, and hire 1,000 lawyers for when it all goes wrong."

> Hark will have its own Memory and Soul, support for Projects, widgets, scheduled tasks, and many other features, including computer use.

> Hark will be available on the web, iOS, and iPad; it will launch with a waitlist, and invite codes will likely be available too.

Looks like Muse is about to get some competition, especially if Hark becomes available more broadly.

🗞 #leak @testingcatalog

ICYMI 👀: Gemini Notebook got a new usage limits system, usage estimation for artifact generation, and a background queue for generations when you are out of quota.

> Limits refresh every 5 hours - Instead of refreshing every 24 hours, your limits now refresh every 5 hours. Get continuous allocation throughout the day.

> Create more of what you love - We are replacing limits per feature with a flexible Notebook-specific limit. You control how you want to spend your limits.

> New background queue - When you are out of limits, you can defer generating videos, slides, etc. for later. These will generate automatically for you.

OpenAI will be rolling out text watermarking for ChatGPT and Codex in the EU in the coming weeks. Watermarking will remain off by default in the API.

> "Across the benchmarks we use to assess Astra, our latest frontier model, we do not see meaningful performance differences with and without watermarking."

This needs some real testing 💀