Anthropic's policy change of disallowing their Pro/Max subscribers to use it for third-party harnesses such as Openclaw/Hermes was the big news event capping week 14.
Other news was the release of Qwen-3.6 Plus , Gemma 4, GLM-5V-Turbo, 3 in-house models MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 from Microsoft and compendium of material about voice dictation tools including Wispr Flow's March updates
Readers who think others in their family, friends and acquaintances who are curious in knowing more about rapidly evolving AI tools/services/use cases and would benefit from being subscribers of this weekly newsletter are encouraged to share this publication link to them and invite them to subscribe
The following messages were posted on the 'All Things AI' Telegram group from Sunday 29th Mar 2026 to Saturday 4th Apr 2026
Boris Cherny [creator of Claude Code] shared 15 tips for Claude Code include /loop for recurring tasks like PR reviews, /batch for parallel refactors, /teleport for remote sessions, and voice mode by holding the spacebar.
Here goes.
via Alex
Copilot adding ads to PRs
https://mastodon.social/@danluu/116317069604398190
via Yat Siu
just shared this, thoughts, critics or otherwise of course welcome:
Alibaba Launches Qwen3.5-Omni Multimodal AI Model Series
Alibaba's Qwen team released the Qwen3.5-Omni model series, including Plus, Flash, and Light Instruct versions. The models support native text, image, audio, and video understanding with 256k context length, over 10 hours of audio input, and more than 400 seconds of 720P video. Key features include Audio-Visual Vibe Coding for generating runnable code from video sketches, speech recognition in 113 languages, speech generation in 36 languages, real-time interaction capabilities, and 215 state-of-the-art scores across multimodal tasks.
Qwen3.5-Omni is here! Scaling up to a native omni-modal AGI.Meet the next generation of Qwen, designed for native text, image, audio, and video understanding, with major advances in both intelligence and real-time interaction.
A standout feature:
Audio-Visual Vibe
OpenAI Launches Codex Plugin for Anthropic's Claude Code
OpenAI released the 'codex-plugin-cc' , letting users integrate its Codex coding agent into Anthropic's Claude Code via a simple marketplace install. Key commands like /codex:review, /codex:adversarial-review, and /codex:rescue handle code checks, tough challenges, and stalled tasks, all powered by models like GPT-5.3-Codex and requiring just a ChatGPT subscription or API key.
Today we’re open sourcing a plugin for it! You can call Codex from Claude Code with your ChatGPT subscription.
We love an open ecosystem!
Use the Codex plugin for Claude Code to delegate tasks to Codex or have Codex review your changes using your ChatGPT subscription.
Start by installing the plugin:
github.com/openai/codex-p…
/plugin marketplace add openai/codex-plugin-cc
Try it out today with:
/codex:review for a normal read-only Codex review
/codex:adversarial-review for a steerable challenge review
/codex:rescue to let codex rescue your code
Vercel publishes guidance on responsible use of AI coding agents like Opus 4.5
Vercel released a blog post outlining internal practices for using AI coding agents responsibly, emphasizing safeguards for security, durability, and availability in mission-critical infrastructure. Key advice includes applying the litmus test of whether one would own a production incident from the agent's pull request, coding quickly but reviewing PRs slowly, and implementing executable guardrails. The post warns that AI-generated code passing CI checks can still pose dangers due to LLM flaws and overconfidence.
Knowing the inherent flaws and over-confidence of LLMs, we sent a clear message to our teams. Vibing and mission-critical infrastructure don’t go together.
We’re sharing
Agent responsibly - Vercel
There's a difference between leveraging AI and relying on it. A framework for shipping agent-generated code with the judgment and guardrails it requires.
Union Square Ventures USV argues that modern AI agents let teams shape software the way MIT’s Building 20 occupants shaped their physical space: continuously modifying tools to fit how they really work, starting from one concrete pain point and iterating from there. They describe a move from “build something people want” (mass-market software) to “build something you want” (custom agents and internal apps tuned to your organization’s processes and tacit knowledge).
A collection of news tidbits about voice dictation products and the rapid rate of adoption of voice dictation amongst enterprises
====
March 2026 Wispr Flow product updates
Willow did a recent April's Fool prank to bring attention to their voice dictation product with a launch video of their new Atlas model The premise was that the new model is actually a call center, and they had humans listening to you as you dictate.
Independent of this prank Willow has a reasonable userbase and people who like its quality.
[Available on Mac, Windows, iOS with Android expected soon]
OTOH, my Instagram feed these days is filled with Mr Beast's promoting Typeless voice dictation software [it's available on Mac, Windows, iOS, Android]
Wispr Flow's CEO Tanay Kothari with a LinkedIn post which mentions speed of adoption of voice dictation at companies
a16z speedrun Andrew Chen with an interesting hardware choice of a USB foot pedal to activate voice dictation in his endeavor to go 'all-in' on voice dictation
AMA
Qwen 3.6 Plus dropped and it's free on OpenRouter.
Specifically designed for agentic workflow with 1M context window. 65K output tokens. 3x faster than Claude Opus 4.6 in early tests.
https://openrouter.ai/qwen/qwen3.6-plus-preview:free
Mercor AI Suffers Data Breach Exposing 939GB of Source Code and 4TB of Training Data from Major AI Labs via LiteLLM Supply Chain Attack
Mercor AI experienced a data breach through a supply chain compromise in the LiteLLM dependency. The incident exposed 939GB of source code and 4TB of state-of-the-art training data belonging to multiple major AI laboratories.
Athena and Aphrodite are code names
Consolidating a flurry of news into one post
Zhipu aka ZAI launches GLM-5V-Turbo: Vision Coding Model a next-generation foundation model specifically engineered to bridge the gap between visual perception and software engineering.
Unlike standard LLMs that rely purely on text prompts, GLM-5V-Turbo natively ingests design drafts, screenshots, videos, and UI layouts to generate fully runnable code.
OpenAI Acquires Tech-Industry Talk Show TBPN from John Coogan. . TBPN will remain editorially independent, retaining control over programming, guest selection, and production, while winding down its advertising business. The TBPN team will join OpenAI's strategy organization to assist with global communications, marketing, and encouraging constructive conversations about AI.
Cursor 3 launched which introduces the Agents Window, where multiple agents run in parallel across local machines, remote SSH, or cloud setups, handling coding tasks and producing editable diffs with full editor support.
Google launches Gemma 4 with permissive Apache 2.0 license and SOTA performance for their size
Microsoft launches in-house AI models MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2, built by its superintelligence team, as it pursues “AI self-sufficiency”
The show is staying the same and we’ll continue to go live at 11am pacific every weekday.
This is a full circle moment for me as I’ve worked with @sama for well over a decade. He funded my first company in 2013. Then helped us fix a serious
Meet the new Cursor · Cursor
Cursor 3 is a unified workspace for building software with agents.
- MAI-Transcribe-1 dropped today, the most accurate transcription model in the world across 25 languages according to FLEURS WER benchmark.
- MAI-Voice-1 sets a new standard for
via Boris Cherny , creator of Claude Code
===
Starting tomorrow at 12pm PT, Claude subscriptions will no longer cover usage on third-party tools like OpenClaw.
You can still use these tools with your Claude login via extra usage bundles (now available at a discount), or with a Claude API key.
You can still use these tools with your Claude login via extra usage bundles (now available at a discount), or with a Claude API key.
Below is my personal website which aggregates links to many of my socials as well as the various content and community that I curate. Feel free to share this link to others who you think may find this content/community useful to them
The cover image of this newsletter via generated via the Google Nano Banana 2 model within the Freepik tool via the following prompt
photorealistic-realistic architectural rendering of a modern futuristic multiple house within concrete giant blocks and greenery and tall rounded windows with minimalist interior, foggy city location, stylish, generative design, cinematic lighting, micro unreal engine 5 effects, 8K sharp focus qualities