In partnership with
AIHOY and happy weekend, AInauts!
Maybe you missed some of last week's most interesting AI news, tools, and hacks - or perhaps you only just joined us. No need to worry!
We have rounded up our highlights from the past week:
We will finish with the most important Quick News - so you can catch every relevant development in one click. Ready? Let's go!
Your AI prompts are leaving out 80% of what you're thinking.
When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning (constraints, edge cases, examples, tone) and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.
Used by teams at Microsoft, Amazon, and Notion. Try Wispr Flow free on Mac, Windows, iPhone, and Android.
One prompt in, finished video out? We tested it. Claude Code built us a 50-second clip with nine scenes in 24 minutes. And it was genuinely good.
There is no new video model behind it. Claude writes the animations as code, handles sound and music, and renders the result as an MP4. The full article includes the prompts so you can recreate it.
The agentic era needs a different CRM. That’s Attio.
Parallel, Turbopuffer, and Wordsmith run their entire GTM motion on Attio, with agents that chase every buying signal, build pipeline, and move deals forward, 24/7.
Zuck does not want to keep your AI assistant trapped inside a chat window. At Meta Connect, the focus was on bringing Muse to glasses, Mac, and more devices. The agent is meant to understand what you need and help wherever the task happens.
Our take: An assistant that helps you shop, book, and order sits directly at the checkout. We expect Meta to want a cut of those transactions. The interesting question is whether Muse will recommend what is best for you or what earns Meta the most.
We attended the launch of Seth Godin's book The Knot in person, and it was quite an experience. One image stayed with us: his warning about a distorted division of labor with AI, the reverse centaur.
The idea is simple. The AI makes the suggestion, you click "Looks good," and then work through the list the machine wrote for you. The AI thinks while you merely execute. At that point, it is worth asking who is really doing whose job.
We currently have two Gamma subscriptions. Not for much longer - we cancelled both.
The trigger was a presentation for our new Second Brain course for the German market: 88 slides created by Claude from the concept, design guidelines, and our knowledge. The slides remain editable, with comments and collaboration built in. Nice!
OpenAI calls its new permanent assistant Dot. It gets its own cloud computer, connects to your apps, and tracks tasks across conversations. Instead of merely writing a text, it keeps working through open items while you do something else. Sounds useful. Still, we are taking a closer look before handing the cute new colleague our entire digital key ring.
You do not necessarily need a new model to make a better decision. Sometimes a method from 1985 is enough: Edward de Bono's Six Thinking Hats.
We turned the method into a prompt that walks through each perspective with you. The AI can help you think, but the decision still belongs to you.
AI News Quickie: This week's HAI-lights
AI never sleeps!
OpenAI is putting half a toolbox into ChatGPT, Google is replacing Gems with Skills, and LLMs keep getting cheaper. Plus, a few changes you may actually notice in everyday work.
🎯 OpenAI: New engine, new colleagues, new prices
GPT-6.1 Sol is launching in ChatGPT Work, Codex, and the API, but not yet in regular ChatGPT. OpenAI says it comes closer to Astra on demanding work while costing only one-fifth as much per token. And we can confirm: it works.
With Dots, ChatGPT gets a persistent agent with its own cloud computer. Important for us: Pro launches without the EEA, Switzerland, or the UK.
The Pro 500 plan costs - you guessed it - $500 per month. It includes Astra Ultrafast, which burns through your token budget at remarkable speed. The allowance on the $200 Pro plan is shrinking. Same price, less included. Hopefully GPT-6.1 Sol helps it last longer.
With Sign in with ChatGPT, you can log in to participating services and use your Plus or Pro AI allowance there too.
Space and Pages bring shared documents, files, and AI work together. You can write and comment on the same page with others and let ChatGPT make changes. Available for Pro, Business, and Enterprise.
Shared presentations in ChatGPT have also been announced. Multiple people and agents will be able to work on the same deck. This one is still coming.
ChatGPT in Slack and Teams can be mentioned directly in a conversation and use connected tools. On Business and Enterprise, colleagues can collaborate in the same thread even without their own ChatGPT license.
The Meetings plugin creates notes and next steps in Space. The audio recording is deleted after processing and cannot be replayed. The Mac beta starts for Pro and Business; open the plugin here.
Codex Cloud keeps tasks running on a cloud computer even when your laptop is asleep.
Plugin extensions get their own interfaces next to the chat. You can edit forms or files directly instead of describing every change in text. Supported plugins can also trigger automations when something happens in a connected app.
The new Decisions API selects from predefined answers, such as which team should handle a support request. It looks rather familiar to us from Jev. OpenAI's advantage: it can work directly with images too.
ChatGPT is getting a virtual try-on. Upload a selfie, view clothes or accessories on your own image, and save favorites. The reference photo stays available for future try-ons until you replace or delete it.
The camera in ChatGPT combines several scanned pages into a PDF. The rollout starts on iOS, turning loose pages into one file.
💎 Google, Anthropic, Microsoft & Grok
Gemini 4 Argon has been announced but is launching first with selected cybersecurity partners. If you are searching the app, you have not missed it.
Google is turning Gems into Skills in Gemini: reusable instructions you call with / or combine. Existing Gems will migrate automatically, while Labs Gems and Opal will not.
With Guided Vision, Gemini Live describes camera images and helps you aim the camera, for example to read fine print or find an object.
Sonnet 5.5 is here and designed for everyday tasks, documents, and manageable coding jobs. Anthropic promises more than 30 percent higher speed.
Claude Code Mods allow deeper assistant customization, such as your own safety checks before risky changes. But be careful: these mods run with Claude Code's permissions. Only install them when you trust the source.
Microsoft is launching new speech models: live transcription in 60 languages and voices in 23 languages. They are intended to make phone assistants, captions, and dictation more fluid.
Grok's Team Bots share work knowledge and app connections with an entire team. Personal conversations stay separate, while everyone in a shared Slack channel can see the answers. The public beta is available on Team and Enterprise plans.
🌍 Everyday work: Building stores, directing voices, saving money
Shopify is introducing Canvas. You see multiple store pages on one shared workspace and let Sidekick make changes. The rollout is happening over the next few days; the existing editor remains available for now.
Runway Ads is designed to create ads, run them on Meta, Google, and TikTok, and improve them based on results. Human approval is enabled by default. For now, it is only a pilot for selected companies.
Edit precisely instead of regenerating everything: FLUX 3 Image changes individual parts of an image, places elements deliberately, handles up to ten reference images, and outputs up to 4K. The rest of the image is meant to stay intact.
Ideogram 4.5 is designed to preserve colors, textures, and details better across repeated edits. High-resolution areas can be edited precisely and placed back into the original.
Eleven v4 gives you more control over emphasis, mood, and dialogue in more than 90 languages. The Turbo version targets fast voice assistants. Both are available, and ten seconds of audio is enough for rapid voice cloning.
With Speech, Suno generates spoken voice and custom background music together in one audio track. You provide text or an idea and describe the voice and music style. The beta is open to everyone, although pronunciation, accents, and pauses can still miss the mark.
Mercury Voice is meant to help phone agents respond faster and is now available to enterprise customers. It handles the reasoning and wording behind the conversation, not the complete phone system.
🛡️ Big deals, clear boundaries
AMD plans to acquire World Labs, Fei-Fei Li's company for spatial AI and 3D worlds. The all-stock deal is valued at around $8.2 billion and is expected to close by year-end, subject to approval. Chips and models are moving even closer together.
ElevenLabs is now valued at $22 billion following an employee share sale.
NVIDIA's Open Agent Safety Platform is designed to constrain agents outside the language model, log their actions, and stop them when they cross boundaries. An important approach for companies.
Anthropic tested GLM-5.3 for cyber risks. In isolated trials, the model could develop attacks, and its safeguards were easy to bypass.
Done! But no reason to be sad. AInauten will be back soon with more fresh material.
See you Monday for another round of news, hacks, and insights!
Your AInauten
Fabian & Reto
